I The Role of User-Generated Content in Human Rights Investigations

As the nongovernmental organization WITNESS notes in its seminal 2011 report, Cameras Everywhere,

Ultimately, WITNESS argues, the ability to access the technology, skills, and expertise needed to analyze these videos will determine “who can participate – and survive – in this emerging ecosystem of free expression.”Footnote ³

A wide range of people produce video content and share it through social media, the Internet, semiprivate communication channels like Telegram and Snapchat, or privately via e-mail or physical storage. Conflict events, protests, riots, and other similar events are increasingly being live-streamed as they happen. Some of the creators of this content have been trained in human rights documentation, while others have not. In many cases, damning video will come from the perpetrators themselves, who use the content to boast of their power and accomplishments or seek funding from sympathetic outsiders.

Courts, tribunals, truth commissions, and other fact-finding (or perhaps fact-generating) bodies, as well as journalists and human rights advocates, need to be sensitive to the wide-ranging quality, completeness, and utility of user-generated content. They cannot assume that the content was intentionally created or that the people represented in this material know that their images and activities are being stored, processed, and analyzed for human rights purposes. Extra care must be taken to ensure the privacy, security, and other basic rights of people who produce such content or appear in it. In the case of perpetrator video, they must assume that the content has public relations goals, and they must take care not to spread messages of hate or extremism. Additionally, it is crucial to keep in mind that many war crimes and human rights abuses will continue to leave few electronic traces.Footnote ⁴ Like all other forms of evidence, video is not a magic bullet or panacea that will put an end to atrocities. Nor does it mitigate the need for eyewitnesses and victims to provide testimony and for investigators to visit the scenes of crimes and conduct thorough investigations.Footnote ⁵

Nonetheless, video has potential value at every stage of a human rights investigation, whether that investigation is designed to feed into advocacy or legal proceedings.Footnote ⁶ Most commonly, video generates leads that can be used to start an investigation. It can also provide evidence to establish that a crime or violation occurred, or it can be used to support a particular factual finding, such as whether a particular weapon was used in a conflict or whether pollution from a particular mining site is polluting a water source. Sometimes, it can also link a particular person, group, government, or company to the violation in question.

A great deal of citizen video provides lead evidence or evidence that a crime or violation occurred. For example, it might show the aftermath of an attack or event, such as destroyed infrastructure, suffering victims, dead bodies, inhumane working conditions, or a damaged environment. Evidence linking a perpetrator to a crime or violation is often harder to come by, but the creative investigator can sometimes mine video collections for this information.Footnote ⁷ Videos might show the presence of particular soldiers in a place where a crime has occurred, as demonstrated by uniforms, vehicles with identifiable symbols, or traceable license plates. Further, weapons and munitions (either spent or unexploded) might be traced to a given military unit, government, or manufacturer. Videos posted to the Internet might also include scenes of government officials or other individuals spurring their followers to commit acts of violence against another group. In the nonconflict context, videos might show violations in places of employment, harassment of minority groups, or environmental harm.

As noted above, the information obtained from video can play a central role in criminal proceedings or truth commissions, but it can also contribute to advocacy campaigns aimed at naming and shaming perpetrators or seeking restitution and justice for victims. Video evidence can help combat standard state or corporate responses to allegations of abuse by making it harder to deny the existence of violations. Video can also be used to raise awareness of and elicit sympathy from those who had not previously been aware of the scope or scale of abuses in a particular context. The recent focus on police brutality against black people in the United States is one example of this phenomenon. Although governments and rights violators can deny the content of written reports or question the methodology used to generate statistical claims, video and other images require the accused to develop a different response. Visual evidence makes it much harder for violators to engage in the tactic that Stanley Cohen calls “literal denial” and requires them to provide an alternative explanation for their actions or claim that their actions were justified.Footnote ⁸

In a case of police or security force abuse, for example, it will be more difficult (but not impossible) for the government to claim that the use of deadly weapons was justified when video evidence suggests that a political protester or criminal suspect was not an immediate threat. In the case of a massacre caught on film, the perpetrators will have to convincingly demonstrate that the videos in question were staged or depict another time or place. Video evidence also makes it harder to engage in the standard repertoire of denial techniques used to refute historical claims; i.e., that “it happened too long ago, memory is unreliable, the records have been lost, no one will ever know what [really] took place.”Footnote ⁹

Video evidence also makes it harder (but, again, not impossible) to impugn the credibility of victims or witnesses who testify against perpetrators because the videos that buttress their claims can be authenticated. Using geolocation techniques developed by the video verification community, it is possible to verify the location of a filmed event by matching features using services like Google Earth, Google Street View, and satellite imagery. Additionally, it is sometimes possible to determine the date and approximate time of day using shadows, weather, sun position, and other climatic clues in the footage.Footnote ¹⁰ Information extracted from video can also be combined with, and corroborated by, forensic evidence (such as autopsy, medical, or pathology reports), other scientific evidence (such as environmental analysis), official records, media reports, intercepted communications, satellite imagery, other open source data, or expert testimony (such as weapons/munitions analysis or political analysis).Footnote ¹¹

II The Challenge of Video Evidence in Law and Advocacy

Video evidence is, of course, not definitive and presents significant challenges for activists and advocates in legal and other contexts. Audiences might grow tired of viewing traumatic video and become desensitized to its effects. On the other hand, they might become less likely to believe accounts that are not supported by convincing audiovisual evidence. The persistent creation of falsified media by states, pranksters, and other nefarious actors could also impugn legitimate content.Footnote ¹² Further, any good defense attorney, government operative, or rights violator can challenge the authenticity of the video or the interpretation of its content (e.g., that victims were indeed dead but were not killed by the claimed perpetrator, or that the victims did not heed warnings to disperse in the moments before the video was shot). They can also offer alternative explanations for what is seen (e.g., that the munitions in question were stolen from the army by a rebel force, or that anti-government forces were dressed up as army personnel). Video evidence cannot always help dispute claims made by perpetrators that their actions were justified on national or internal security grounds, that the victims posed a threat, or that the action captured on video was wrong but the matter has already been dealt with internally.

III Three Case Studies: Syria, Ukraine, and Nigeria

In order to better understand the landscape of video content, this section discusses three recent case studies in which videos were used to support human rights advocacy and accountability. These cases are not fully representative of all of the possible uses of audiovisual evidence in human rights work, but they provide geographically diverse exemplars that, when viewed together, highlight many of the important issues in this domain. They highlight the diverse ways that video content is being used in advocacy and accountability efforts; they also demonstrate the variety of conflict-related information that can be gathered from video, including the distribution of weapons systems, the relative strength of military units or protest movements, and social linkages among various actors in a conflict or political movement. These uses go beyond the kind of secondhand witnessing and visual representation of an event that have traditionally characterized video evidence and raise important questions about the representativeness and reliability of audiovisual accounts. These cases also illustrate the way in which video can become the source of other forms of evidence (e.g., forensic studies of blood stains or wounds that have not been directly examined by an expert, or ballistics reports on weapons that are seen only on video), as well as how it can complement or add credence to expert testimony about physical evidence.

IV Discussion

The three case studies presented above do not represent the entire spectrum of uses of video in human rights advocacy and accountability efforts. They do not illustrate the use of video to protect and promote economic, social, and cultural rights or in American or Western European contexts. The three case studies do, however, demonstrate at least some of the ways in which video can be used in human rights and conflict documentation.

Especially in the cases of Ukraine and Nigeria, video is treated as one of many forms of evidence to be corroborated and integrated into a coherent package of evidence. In these two cases, video does not stand on its own in an investigation. The Carter Center takes a somewhat bolder approach, making claims about conflict dynamics – in terms of both the size and influence of particular armed groups and the weapons they possess – based on information extracted from social media networks and social media content.

While it is impossible to definitively test the Carter Center’s claim that the online presence of armed groups is roughly similar to their actual presence (with the notable exception of the Islamic State), the Center’s analysts have strong personal networks in Syria and in neighboring Turkey and are able to take advantage of human intelligence to point out errors and omissions in their open source intelligence. Even with this confirmation, the Carter Center does not claim that its findings are completely accurate or statistically valid, only that they provide insight that can be used in conflict analysis and in decision-making concerning diplomacy and humanitarian aid. The Carter Center recognizes that definitive quantitative claims cannot be made using social media data without significant additional corroboration and statistical analyses of uncertainty. Like most other forms of evidence, social media is a convenience sample – i.e., data that is relatively easily available to the investigator, as opposed to data collected through some form of systematically randomized or cluster sampling – and they acknowledge the limitations this creates.

The Ukraine case shows that governments are increasingly paying attention to the results of open source investigations, and sometimes go to great lengths (including falsifying data) to call their legitimacy into question. At the same time, it also demonstrates that in an age of open source intelligence, even ordinary people can contribute to the monitoring of conflict and human rights violations. Data about conflict and human rights abuse that was once only available to military or intelligence officials can now be found on the same platforms where cat videos, sports highlights, and other cultural ephemera are routinely shared.

The widespread availability of this content has numerous positive aspects, but it also creates significant challenges. At the most basic level, repeated viewings of audiovisual depictions of death, abuse, and torture can have a variety of negative impacts at the individual and societal levels. Excessive viewing of such content can traumatize the human rights investigator, journalist, or even the casual viewer. I experienced this vicarious or “secondary” trauma firsthand after seeing a particular image of a dismembered child in a video that depicted the aftermath of a shelling in Syria, and I have spoken to numerous colleagues who have experienced it as well. It is increasingly becoming a topic of analysis in both journalism and human rights work.Footnote ⁵⁷ Indeed, much of the impetus for the research my human rights colleagues and I are conducting on semi-automating video analysis using machine learning and computer vision comes from a desire to limit exposure to images of human suffering and brutality.

A corollary to secondary trauma is that the proliferation of horrific content on the Internet may ultimately desensitize the public to what is being portrayed. At a certain point, many people will learn to tune out the reality of what they see because it is too graphic, while a minority of the public will find some perverse pleasure in it in an exploitative and voyeuristic, but not empathetic, way. There are many websites and YouTube channels that present the violence in Syria and other countries accompanied by intense music rather than historical, cultural, or situational context.Footnote ⁵⁸ More work will have to be done to determine the effects of the availability of this content on public opinion, but there is little indication that it led to an upsurge in demand for humanitarian intervention in the conflict.

Even more concerning than the potential negative impacts of video in the human rights domain is the reality that the proliferation of audiovisual evidence of war crimes and human rights abuses has not yet resulted in greater accountability, at either the domestic or international level. Few high-level government or military officials accused of war crimes or crimes against humanity have been forced to admit culpability, provide reparations, or step down from positions of power due to the existence of video evidence. Nor have corporate decision-makers been held to account for the actions of their companies on communities, environments, or workers due to damning video evidence.

But perhaps this is too much to ask of video or any other form of evidence. Achieving justice and accountability requires political will. There is already ample evidence available to convict high-level perpetrators of war crimes and human rights abuses in many violent situations around the world, but no action is taken due to geopolitical or economic realities.

That said, advocacy efforts, including naming-and-shaming campaigns, often lead to important policy change in the short term, and the preservation of relevant content allows for historical clarification and the possibility of justice and accountability in the long term.Footnote ⁵⁹ One need only to look at Guatemala’s efforts to bring Efrain Rios Montt to trial for crimes committed during his presidency in 1982–83, Argentina’s recent successful conviction of the architects of its Dirty War (including the dictator at the time and his closest associates), or the Extraordinary African Chamber’s recent conviction of Chad’s ex-dictator Hissène Habré for crimes against humanity that took place in the 1980s to see that justice can emerge many years, or even decades, after crimes occur.Footnote ⁶⁰ Moving forward, video – properly understood, analyzed, and contextualized – will undoubtedly play an important role in these processes.

I Big Data Analytics for Monitoring Human Rights: Collection and Use

Assessing the legitimate application of big data analytics in human rights work depends on understanding what the right to privacy protects. Nissenbaum’s framework of “contextual integrity” can provide a way to understand the value of privacy within the human rights context and the situations in which the use of big data might infringe this right. Contextual integrity ties “adequate protection for privacy to norms of specific contexts, demanding that information gathering and dissemination be appropriate to that context and obey the governing norms of distribution within it.”Footnote ⁵ Nissenbaum does not go down the perilous path of trying to define privacy in absolute terms or finding a precise legal definition against the thicket of competing legal regimes, and neither will this chapter. Instead, she discusses privacy in terms of an individual’s right to determine the flow of information about him- or herself.Footnote ⁶ What individuals care about, Nissenbaum argues, is that their personal information flows appropriately depending on social context. This chapter will employ similar terminology, such that privacy is violated when an individual’s information is collected, analyzed, stored, or shared in a way that he or she judges to be inappropriate.

One challenge with examining whether a specific technology may violate privacy is that technology is not a single artifact that exists by itself. Technology is a combination of sociotechnical processes. Thus, it is useful to divide the technologies that comprise big data broadly into two categories: collection and use. As Alvaro Bedoya notes, most questions dealing with data privacy and vulnerable populations focus on how data will be used rather than how it will be collected.Footnote ⁷ Yet collection and use are intrinsically tied together. Disaggregating these categories into their individual processes provides us with a better understanding of potential privacy concerns related to human rights monitoring.

II Privacy Trade-offs, Urgency, and Temporality

Privacy considerations associated with applying big data analytics in the human rights context has received considerably less attention in the literature than issues like representativeness and validity. A recent volume edited by Philip Alston and Sarah Knuckey discusses new advances in technology and big data that are poised to transform the longstanding activity of human rights fact-finding.Footnote ¹¹ In his contribution to that volume, Patrick Ball raises major reservations about using big data for human rights. Ball argues that sampling procedures and the data itself in many big datasets are riddled with biases.Footnote ¹² Regardless of its “big-ness,” data that is not representative leaves out key populations. Incomplete data can hide the very populations most affected by human rights violations. For example, when estimating the number of victims in a conflict based on available data, those who were secretly killed by paramilitary groups and never reported in the media or government records are left out. The models and analyses resulting from such flawed data, Ball argues, can lead to faulty conclusions that could imperil or discredit human rights fact-finding and evidence gathering.

Patrick Meier, on the other hand, suggests an alternative perspective, buoyed by the high-profile applications of big data analytics in humanitarian crises.Footnote ¹³ From crowdsourced maps after the Haiti earthquake in 2009 to millions of tweets related to Hurricane Sandy in New York, such data is important and can help save lives. Meier concedes that big data can be biased and incomplete. Yet data and information on vulnerable populations are almost always lacking in completeness, even more so in the immediate aftermath of a crisis. Thus, big data, for all its flaws, can serve to inform decision-making in real time (i.e., during a crisis event) where comprehensive information does not exist.

One of the core questions that needs to be answered when using big data for human rights purposes is the extent to which the urgency of the need being addressed impacts the decision to use imperfect data and risk privacy violations. Consider, on the one hand, a fact-finding mission to ascertain whether a human rights violation took place in the past. Sometimes the data collection can take months or years in order to produce evidence for use in justice and accountability proceedings or for historical clarification. On the other hand, in a humanitarian response to a disaster or crisis, data collection seeks to intervene in the present, to find those in immediate need. Of course, temporality is not an absolute rule separating human rights and humanitarian domains. Data can be collected both in protracted humanitarian situations and when human rights violations are happening in the present. The issue at hand is whether an urgent situation places unique demands on the flow of personal information, which impacts one’s dignity, relationships with others, and right to privacy.

During humanitarian responses to time-bound crises like natural disasters, decisions must often be made between maintaining privacy and responding quickly by using available data that is often deeply personal. For example, in the response to the Ebola crisis, humanitarian organizations deliberated on how big data could be used legitimately.Footnote ¹⁴ Consider a responder requesting the mobile phone contact list of a person who tested positive for Ebola in an attempt to stop the spread of the disease. Further, consider a response organization asking a mobile phone company for the phone numbers and records of all the users in the country in order to trace the network of individuals who may have become infected. That data would need to be analyzed to locate those at risk, and personal information might be shared with other responders without the consent of the data subjects.

The assumption in such an operational decision is that saving the lives of the person’s contacts and protecting the health of the broader public outweigh the privacy concerns over that person’s personal information. Yet such decisions about trade-offs, particularly in the case of the Ebola response, remain highly controversial due to the potential privacy violations inherent in collecting the mobile phone records of individuals at a national level without the consent of individuals.Footnote ¹⁵ Even in an urgent humanitarian response context, there is little agreement about when it is appropriate and legitimate to limit privacy rights. Applying real-time data collection and analytics to the investigation of historical human rights violations could raise even more concerns.

When, if ever, is it appropriate to limit privacy rights in order to achieve human rights objectives? Big data can support human rights fact-finding by providing a basis for estimating the probability that an individual falls within a group that has been subjected to past rights violations. Data analytics can also personally identify an individual whose rights are being violated in the present. For example, access to millions of mobile phone records in a large city may reveal patterns of calls by individuals that suggest human trafficking for commercial sexual exploitation. An analysis of these “call detail records” can indicate the movement of people to buildings where exploitation is known to take place at unique times and pinpoint calls with known exploiters. Yet to establish whether a specific individual’s rights have been violated, personal information like cell phone numbers, names, and building ownership would need to be identified. Such data can reveal sensitive information (e.g., religion, political affiliation, or sexual orientation) that could lead to physical harm and retribution if shared with the wrong party.

Ultimately, permissible limitations on the right to privacy will vary depending on the urgency and importance of the objective to be achieved. The accompanying graph gives a visual sketch of the potential trade-offs involved in privacy. The right to privacy is not an absolute right, and it may be limited in order to achieve a legitimate objective. The extent to which the right is infringed, however, must be proportional to the objective. Thus, the graph illustrates a scenario involving a decision about the proportional relationship between the urgency of the objective of human security and permissible limits on privacy. In a human rights context, protecting the right to privacy is of high concern since safeguarding an individual's data also protects the right to freedom of expression, association, and related rights. Yet, when there is an increasing threat to human security and safety, such as the imminent danger of an individual’s death due to violent conflict, the concern over privacy may start to decrease. Perhaps a decision must be made about whether to obtain or release an individual’s personal mobile phone’s GPS coordinates without their consent or permission in order to attempt to locate and rescue that person to save his or her life. There may come an inflection point in the situation where the immediate danger to human security is higher than the protection of an individual’s privacy. When there is no longer an immediate threat to human security, there is every reason to uphold the integrity of the individual’s privacy right at a high level. Of course, there are a number of additional factors that might influence this analysis and decision. In some cases the release of an individual's data will expose that person to the very forces threatening his or her safety and security or may put another individual directly at risk. Clearly, more foundational work is needed to begin to understand how these trade-offs may be operationalize in practice when making a real-time decision.

Does this mean that we cannot use big data collection for long-term and continuous human rights monitoring? Long-term monitoring of digital data sources for human rights violations holds the promise of providing unprecedented insight into hidden or obscured rights abuses such as labor violations, sexual abuse/exploitation, or human trafficking. The use of big data analytics for monitoring human rights, even by human rights organizations, carries both known and unknown risks for privacy. Indeed, the very nature of data collection and analysis can conflict with normative expectations of privacy in the human right context. It is unclear whether the uncertain benefits of long-term human rights monitoring and advocacy can outweigh the very concrete risks to privacy that accompany the use of big data.

III Human Rights Monitoring and Surveillant Anxieties

The human rights field has a long tradition of employing methods and techniques for protecting privacy while monitoring rights abuses and violations. Determining “who did what to whom” involves identifying victims and alleged violators, doing an accounting of the facts, and collecting relevant information.Footnote ¹⁶ Ensuring the privacy and confidentiality of sources and victims is a critical concern in human rights fact-finding. A training manual published by the UN Office of the High Commissioner for Human Rights (OHCHR), for example, urges the use of monitoring practices that “keep in mind the safety of the people who provide information,” seeking consultation in difficult cases and maintaining confidentiality and security.Footnote ¹⁷ At the local level, Ontario’s Human Rights Commission states that data collection should include informing the public, consulting with the communities that will be affected, using the least intrusive means possible, assuring anonymity where appropriate, and protecting privacy.Footnote ¹⁸

In reality, though, ethical data collection protocols that protect privacy in the human rights context, such as obtaining informed consent, are extraordinarily difficult to utilize when deploying big data analytics. Big data analytics often relies on “found” data, which is collected without a user’s consent or even knowledge. It also necessarily involves using this information in ways not intended by the individual to whom it relates. Thus, with respect to big data, privacy harms are likely unavoidable.

The use of big data also harms privacy by adding to a growing, although ambiguous, sense of “surveillant anxiety,” in which we fear surveillance to the point that it affects our thoughts, behaviors, and sense of self. Kate Crawford describes this anxiety as “the fear that all the data we are shedding every day is too revealing of our intimate selves but may also misrepresent us… [N]o matter how much data they have, it is always incomplete, and the sheer volume can overwhelm the critical signals in a fog of possible correlations.”Footnote ¹⁹ A fact-finding mission that creates privacy risks is not equivalent to surveillance. And certainly not all surveillance technologies violate privacy in a legal sense. Yet any use of big data analytics to monitor human rights creates concerns and anxieties about surveillance, whether or not the surveillance is intended for “good.”

Thus, the issue that must be addressed is whether human rights researchers using big data analytics would themselves produce this kind of surveillant anxiety in their data subjects in ways that feel similar to traditional government or corporate surveillance. According to the Special Rapporteur on the right to Privacy, surveillance creates privacy risks and harms such that “increasingly, personal data ends up in the same ‘bucket’ of data which can be used and re-used for all kinds of known and unknown purposes.”Footnote ²⁰ Although surveillance for illegitimate purposes necessarily violates privacy, even surveillance for legitimate purposes will do so if the associated privacy harms are not proportional to that purpose. For example, a public health surveillance program that continuously monitors and identifies disease in a population might constitute an appropriate use for a common good shared by many, but could still violate privacy if it produces undue harms and risks. As Jeremy Youde’s study of biosurveillance contends:

In their work on disease surveillance in the United States, Amy L. Fairchild, Ronald Bayer, and James Colgrove document the conflict around public health surveillance during the early days of the AIDS crisis and the struggle to balance the need to collect and share medical records against charges of institutional discrimination against marginalized groups.Footnote ²²

Big data collection and analysis by technology companies, sometimes called corporate surveillance, can produce the same kinds of anxieties. And big data collection by either governments or human rights organizations often rely on technologies that serve as intermediaries to the digital life of the public. Both a government agency and a human rights organization may collect data on the lives of millions of individuals from major social media platforms like Facebook. The very same tools, techniques, and processes in the collection and use of big data can be employed to both violate and protect human rights. The collection of one’s personal data by governments, corporations, or any number of other organizations using big data analytics may contribute to the constant feeling of being watched and curtail privacy and freedom of expression.

The pressing question is whether the use of big data by human rights organizations is permissible, given the risks to privacy involved. It is fair to say that human rights organizations should be held to the same standards that privacy advocates require of government. Yet the best intentions of human rights organizations using big data are not enough to protect privacy rights or automatically justify privacy violations. Furthermore, any organization collecting, classifying, and storing sensitive human rights data needs to address issues like data protection, secure storage, safe sharing, and access controls. If a human rights organization deploys data collection and analytic tools, how can they incorporate safeguards that responsibly address the inherent risks to privacy and minimize the potential harms?

IV Toward Guidelines for Big Data Applications in Human Rights

Because of concerns about privacy, value trade-offs, surveillance, and other potential risks and harms that may befall vulnerable populations, a rigorous assessment of the legitimacy and impact of the use of any big data analytics by organizations in the human rights context is vital. This begs the question of what type of guidelines could help steer such an assessment, particularly given the global proliferation of technologies and the plethora of context-specific harms. As the UN Special Rapporteur on Privacy states, “The nature of trans-border data flows and modern information technology requires a global approach to the protection and promotion of human rights and particularly the right to privacy.”Footnote ²³ Human rights monitoring organizations may need to update their standards for data collection and analysis to take new technologies like big data into account.

New technological applications do not necessarily require entirely new high level principles. For example, the principles of safety, privacy, and confidentiality outlined in the OHCHR Training Manual would still apply to big data, but these principles may need further elaboration and development when they are applied to new information collection regimes. At the same time, new technological challenges need new solutions. For example, confidentiality policies may not be achieved simply through anonymization techniques alone, since the currently accepted fact in computer science is that no dataset can be fully anonymized.Footnote ²⁴

A major advance in addressing the ethical and responsible use of data is the Harvard Humanitarian Initiative’s Signal Code, subtitled “A Human Rights Approach to Information during Crisis,” which focuses on the right to information in times of crises. Bridging the humanitarian and human rights contexts, the code states that the right to privacy is fundamental in crisis situations and that “[a]ny exception to data privacy and protection during crises exercised by humanitarian actors must be applied in ways consistent with international human rights and humanitarian law and standards.”Footnote ²⁵ Other standards in the code include giving individuals the right to be informed about information collection and use as well as the right to have incorrect information about themselves rectified.

Policy work from other fields, such as international development, will also be relevant to creating guidelines for human rights. UN Global Pulse, the data innovation initiative of the UN Secretary General, has published its own “privacy and data protection principles”Footnote ²⁶ for using data in development work. These principles recommend that actors “adhere to the basic principles of privacy,” “maintain the purpose for data,” and ensure “transparency [as] an ongoing commitment.”Footnote ²⁷ Human rights organizations may also learn lessons from civil society’s demands of governments engaging in surveillance. The International Principles on the Application of Human Rights to Communications Surveillance suggests a number of guidelines designed to ensure that government surveillance does not infringe on the right to privacy, including notifying people when they are being watched.Footnote ²⁸

This brings us to the issue of how to ensure the accountability of human rights organizations using big data for interventions or monitoring. Any actor that uses big data to intervene in or monitor human rights, whether a small domestic NGO or a large international organization, should be responsible for the potential risks and harms to the very populations it seeks to help. Yet since human rights organizations often hold governments to account, asking governments to regulate the use of big data by those very organizations will likely provide an avenue for state suppression of human rights monitoring. As such, traditional regulatory mechanisms are unlikely to be effective in this context.

One emerging framework that seeks to regulate any entity engaged in data collection is the EU’s General Data Protection Regulation, which comes into force in 2018.Footnote ²⁹ This regulation requires all organizations collecting data on EU residents to follow privacy directives about data collection, protection, and consent. For example, individuals in the EU would have the right to withdraw consent given to organizations that collect or process their data. Such organizations would be required to alert the authorities if a data breach of personal information occurred. Any organization, including both companies and nonprofits, could incur heavy fines for noncompliance, such as 4 percent of global revenue.

Another possible avenue may lie in encouraging human rights organizations to engage with technology privacy professionals to assess the possible use of new technologies like big data. Professionals with the appropriate technical, legal, and ethical expertise may be capable of conducting privacy impact assessments of risks that may harm vulnerable populations before a new technology is deployed.Footnote ³⁰ Or perhaps large donors can require grantees to follow data privacy protections as a condition of ongoing funding.

At the end of the day, though, the lack of certainty about the effectiveness of these approaches speaks to the pressing need for both more foundational research and context-specific assessments in this area. Addressing questions about context would require researchers to include more direct input and participation of the vulnerable communities themselves; for example, through field research. As Nissenbaum suggests, protecting privacy is about upholding the contextual integrity of the underlying norms governing information flow. Understanding norms around data privacy and consent, for example, should necessarily involve the communities where those norms exist.

Applying big data analytics in human rights work reveals tensions around privacy that need to be resolved in order to guide current thinking and future decisions. Unfortunately, a sustained knowledge gap in this area puts vulnerable populations at greater risk. And the anxieties and trade-offs around norms and interventions will be compounded as the human rights field addresses “newer” technologies such as artificial intelligence (AI).Footnote ³¹ Since AI is fueled by big data collection and analytics the same privacy concerns discussed above would apply. Furthermore, AI entails some form of automated decision making, which creates dilemmas over whether only a human rights expert, rather than a computer algorithm, can decide about proportionality and trade-offs between rights in real time. A research agenda should work toward guidelines, principles, and practices that anticipate the risks, costs, and benefits inherent in each process involved in emerging technological interventions in the human rights context.

I Introduction

In September 2007, The New York Times columnist Nicholas Kristof traveled with Bill Gates to Africa to look at the work the Bill & Melinda Gates Foundation was doing to fight AIDS. In an e-mail to a Times graphics editor, Kristof recalls:

Kristof’s anecdote illustrates the sometimes unexpected power of data visualization: Expressing quantitative information with visuals can lend urgency to messages and make stories more memorable.

Data visualization is the “visual representation of ‘data,’ defined as information which has been abstracted in some schematic form.”Footnote ² The use of data visualization can strengthen human rights work when data is involved, and it does something for the promotion of human rights that other methods don’t do. Combining data and visuals allows advocates to harness the power of both statistics and narrative. Data visualization can facilitate understanding and ultimately motivate action. And within human rights research, it can help investigators and researchers draw a bigger picture from individual human rights abuses by allowing them to identify patterns that may suggest the existence of abusive policies, unlawful orders, negligence, or other forms of culpable action or inaction by decision-makers. As human rights researchers and advocates look for new ways to understand the dynamics behind human rights violations, get their messages across, and persuade target audiences, they are also expanding the epistemology of advocacy-oriented human rights research. By broadening their evidence base and using new methods, human rights advocates come to know different things – and to know the same things differently.

The use of data visualization and other visual features for human rights communication and advocacy is a growing trend. A study by New York University’s Center for Human Rights and Global Justice reviewing all Human Rights Watch (HRW) and Amnesty International reports published in 2006, 2010, and 2014 revealed an increase in the use of photographs, satellite imagery, maps, charts, and graphs.Footnote ³ In some cases, data visuals augment existing research and communications methodologies; in other cases, they represent alternative and even novel tools and analytical methods for human rights NGOs.

While data visualization is a powerful tool for communication, the use of data and visualization holds exciting promise as a method of knowledge production. Human rights researchers and advocates are adding new methodologies to their toolbox, drawing on emerging technologies as well as established data analysis techniques to enhance and expand their research, communications, and advocacy. This chapter introduces ways data visualization can be used for human rights analysis, advocacy, and mobilization, and discusses some of the potential benefits and pitfalls of using data visualization in human rights work. After a brief historical review of data visualization for advocacy, we consider recent developments in the “datafication” of human rights, followed by an examination of some assumptions behind, and perils in, visualizing data for human rights advocacy. The goal of this chapter is to provide sufficient grounding for human rights researchers to engage with data visualization in a way that is as powerful, ethical, and rights-enhancing as possible.

II A Brief History of Statistical Graphics and Advocacy

Visual storytelling has a long and colorful history. Past generations not only created depictions of their reality, but also crafted visual explanations and diagrams to convey and understand the invisible forces governing the visible world and other realms beyond perception. In The Book of Trees: Visualizing Branches of Knowledge, Manuel Lima charts the use of the branching tree as a visual metaphor in charts from Mesopotamia to medieval Europe to the present.Footnote ⁴ In addition to visual storytelling, ancient civilizations developed visual methods to record, understand, and process large numbers. The ancient Babylonians, Egyptians, Greeks, and Chinese used visual systems to record data about vital resources, chart the stars, and map their territories.Footnote ⁵ While visual storytelling was intended for communication, data visualization was used by the ruling powers to interpret and keep tabs on their empires.Footnote ⁶

Along with the development of modern statistics in the late eighteenth century, there emerged graphical methods of quantitative analysis – new kinds of charts and graphs to visually show patterns in data.Footnote ⁷ The Scottish political economist William Playfair was a mechanical engineer, statistician, and activist who authored essays and pamphlets on the politics of the day and helped storm the Bastille in Paris in 1787.Footnote ⁸ That same year, he published the kind of line, area, and bar charts that we routinely use today for the first time in his Commercial and Political AtlasFootnote ⁹ to display imports and exports between Scotland and other countries and territories. He published the first modern pie chart in his 1801 Statistical Breviary.Footnote ¹⁰

In the first half of the nineteenth century, new technologies helped spread and inspire enthusiasm for both statistics and data visualization.Footnote ¹¹ Commercial mechanical devices for counting, sorting, and calculating became popular, and the first successful mass-produced mechanical calculator was launched in 1820.Footnote ¹² Printing technology, particularly lithography and chromolithography, enabled a more expressive range of printing. The Statistical Society of London was founded in 1834 and, by royal charter, became the Royal Statistical Society in 1887.Footnote ¹³ As the psychology scholar Michael Friendly notes, between 1850 and 1900, there was explosive growth in both the use of data visualization and the range of topics to which it was applied.Footnote ¹⁴ In addition to population and economic statistics, mid-nineteenth-century Paris saw the publication of medical and mortality statistics, demographics, and criminal justice data.Footnote ¹⁵

A few examples from this period show how data visualization contributed new findings using spatial and other forms of analysis, and allowed such findings to be made meaningful to a broader public via visual display.

In 1854, Dr. John Snow mapped cholera deaths around the thirteen public wells accessed in the Soho district of London.Footnote ¹⁶ Using this method, he made a dramatic discovery: there was a particular cluster of deaths around one water pump on Broad Street. His findings ran contrary to the prevailing theories of disease. At the time, the medical establishment believed in the miasma theory, which held that cholera and other diseases, such as chlamydia and the Black Death, were caused by “bad air.” Dr. Snow believed that cholera was spread from person to person through polluted food and water – a predecessor to the germ theory of disease. Despite skepticism from the medical establishment, Snow used his map to convince the governing council to remove the handle from the pump, and the outbreak quickly subsided.Footnote ¹⁷

Florence Nightingale is known primarily as one of the founders of modern nursing, but she was also a statistician who used data visualization to campaign for improvements in British military medicine.Footnote ¹⁸ In 1858, she popularized a type of pie chart known as the polar area diagram.Footnote ¹⁹ The diagram divides a circle into wedges that extend at different lengths from the center to depict magnitude. Nightingale used statistical graphics in her reports to members of Parliament about the condition of medical care in the Crimean War to illustrate how improvements in hygiene could save lives: at a glance, one could see that far more soldiers died of sickness than of wounds sustained in battle.Footnote ²⁰ Nightingale persuaded Queen Victoria to appoint a Royal Commission on the Health of the Army, and her advocacy, reports, and the work of the commission eventually led to systemic changes in the design and practices of UK hospitals.Footnote ²¹

After a long and distinguished career as a civil engineer in France, Charles Minard devoted himself in 1851 to research illustrated with graphic tables and figurative maps.Footnote ²² His 1869 visualization of Napoleon’s 1812 Russian Campaign shows the march of the French army from the Polish-Russian border toward Moscow and back.Footnote ²³ The chart was heralded by Minard’s contemporaries and is held up by twentieth-century data visualization critics as a marvel of clarity and data density. The visualization displays six different dimensions within the same graphic: The thickness of the main line shows the number of Napoleon’s troops; the scale of the line shows the distance traveled; rivers are depicted and cities are labeled; dates indicate the progress of the march relative to specific dates; the orientation of the line shows the direction of travel; and a line chart below the route tracks temperature. Reading the map from left to right and back again, another message beyond the data emerges: As the march proceeds and retreats, the horrific toll of the campaign slowly reveals itself as the troop numbers decline dramatically. The graphic not only details historical fact, but serves as a powerful antiwar statement.

In the United States, data visualization was used to sound the alarm on the frequency of racist violence and its consequences. In 1883, the Chicago Tribune began publishing annual data on lynching in the form of a monthly calendar listing victims by date.Footnote ²⁴ The journalist and anti-lynching campaigner Ida B. Wells cited the data in her speeches and articles. The annual publication of state and national statistics fed the public’s outrage, and on September 1, 1901, the Sunday Tribune published a full front-page table of data on 3,000 lynchings committed over 20 years, as well as information about the victims of 101 lynchings perpetrated in 1901 and the allegations against the victims that their killers had cited to rationalize the violence.Footnote ²⁵ Rather than focusing on individual cases, the data, table, and narrative exploration presented a powerful picture of the frequency and scale of the crisis: though lynchings occurred in mostly southern states, they were found in nearly every state of the union. The pages appeal to public opinion to support change, explicitly calling out the failure of state and local law enforcement and demanding congressional action.

These historic charts and graphs are analytical, but also rhetorical, using visual conventions to identify the dynamics of important phenomena and to communicate findings and make an argument and a persuasive case for policy change. We will investigate some of the promises and pitfalls of coupling visual rhetoric with data below, but first we briefly examine datafication and human rights.

III Datafication and Human Rights

As a field, advocacy-oriented human rights research traditionally favors qualitative over quantitative research methodologies. Research is typically driven by interviews with victims, witnesses, and alleged perpetrators, usually supplemented by official documents, secondary sources, and media accounts.Footnote ²⁶ Additional qualitative methods, including focus groups and participatory observation, are used by some groups, as are forensic methods such as ballistics, crime scene investigations, and exhumations. Quantitative methods such as data analysis and econometrics have been very rare until recently. There are many reasons for the traditional emphasis on qualitative methods. Historically, advocacy-oriented human rights research developed out of legal and journalistic traditions.Footnote ²⁷ Ethically, rights advocates are committed to the individual human story. Human rights practice has been defined as “the craft of bringing together legal norms and human stories in the service of justice.”Footnote ²⁸

At the same time, researchers in social science, epidemiology, and other fields have long used quantitative methods for research on human rights related issues. Political scientists have developed cross-national time-series datasets to interrogate the relationships between human rights and major social, economic, and political processes.Footnote ²⁹ Epidemiologists have studied inequalities in access to health care and disparities in health outcomes between social groups.Footnote ³⁰ Research psychologists have examined the way human psychology may limit our ability to respond to widespread suffering such as that arising from genocide and mass displacement.Footnote ³¹

Human rights NGOs are increasingly embracing scientifically based methods of research that involve data and quantification, and they are beginning to use data visualization to reach broader audiences. Using data-driven methods from other fields enables different ways of knowing, of gathering and processing information, and of analyzing findings.

The spread of digital network infrastructure, increased computing speeds, and a decrease in the cost of digital storage have made collecting, sharing, and saving data easy and prevalent. The economic accessibility of mobile technology has made cell phones widely available, even in poor countries.Footnote ³² Smartphones have put Internet access and the production of digital content in the hands of the people – generating an enormous swarm of digital exhaust and big data about many populations across the world. This explosion of new data reflects a democratization of sorts, but it also puts a new means of surveillance at the command of state agents,Footnote ³³ increases the power of private data owners and brokers, and creates pockets of digital exclusion, where communities that do not benefit from the digital revolution are further marginalized.

In this “datified” world, decision-makers seek evidence in the form of data and quantitative analysis. As Sally Merry notes, “quantitative measures promise to provide accurate information that allows policy makers, investors, government officials, and the general public to make informed decisions. The information appears to be objective, scientific, and transparent.”Footnote ³⁴ While Merry’s language suggests that numbers themselves promise to smooth over the messiness of decision-making by appearing scientific, it is, of course, human beings who insist on quantification. Theodore Porter has identified quantification as a “technology of distance” capable of mediating distrust, such as that between governments and citizens.Footnote ³⁵ In the human rights context, quantification sometimes functions as a way of disappearing the judgment-laden practices of monitoring and assessment, where governments may distrust their monitors as much as monitors distrust officials.Footnote ³⁶ In such a context, knowing how many were killed, assaulted, or detained can be seen to satisfy a yearning for objective knowledge in a chaotic and often brutal world.

Metrics are also attractive because they can be weighed against other data and wrapped up into “indicators.” Kevin Davis, Benedict Kingsbury, and Sally Merry define indicators as:

In the human rights realm, indicators have been developed, inter alia, to directly measure human rights violations,Footnote ³⁸ assess compliance with treaty norms,Footnote ³⁹ measure the impacts of company activities on human rights,Footnote ⁴⁰ and ensure that development processes and humanitarian aid are delivered in a rights-respecting manner.Footnote ⁴¹ As Merry notes in The Seductions of Quantification, indicators are attractive in their simplicity, particularly country rankings that have proven effective in catching the attention of the media and the public.Footnote ⁴² While indicators provide a convenient analysis at a glance, they are complicated and often problematic: Data collected may be incomplete or biased, not comparable between countries, or compromised by encompassing metrics of behavior that may not capture a diversity of values or reasons for the behavior.Footnote ⁴³ There may also be a slippage between the norm and the data used to assess the norm, a dynamic in which difficult-to-measure phenomena are assessed using proxy indicators that may become attenuated from the original norm.

In the human rights field, the relationship between the norm and the data can be especially complicated. Human rights data is almost always incomplete and often fraught with bias and assumptions.Footnote ⁴⁴ The imperfect nature of human rights data is a consequence of the challenges facing its collection. There are inherent difficulties in getting a complete or unbiased dataset of anything, but it is particularly challenging when it is in a government’s self-interest to hide abuses and obstruct accountability. Marginalized groups may be excluded from the available information as a result of implicit bias, or even by design.Footnote ⁴⁵ For human rights researchers, there may be dangers and difficulties associated with asking certain questions or accessing certain information. Much of the data gathered about civil and political rights violations is collected through case reports by human rights organizations, making it inherently biased by factors such as the organization’s familiarity and accessibility, the victims’ willingness to report, and the security situation.Footnote ⁴⁶ Data about economic and social rights may seem easier to gather, since there is a plethora of official data in most countries about education, housing, water, and other core rights. This data is not designed to assess rights, however, meaning that it is, at best, proxy data for rights fulfillment.Footnote ⁴⁷

However, even when there are flaws in the data collection or the data itself, the results can sometimes be useful to researchers and rights advocates. For instance, if the methodology for gathering data is consistent year after year, one may be able to draw certain types of conclusions about trends in respect for rights over time even absent a representative sample. If the data in question was collected by a government agency, it can be strategic for activists to lobby the government using its own data despite the flaws it contains, since such a strategy makes the conclusions that much harder to refute.

Further, a great power of statistics is the ability to work with data that is incomplete, biased, and uncertain – and to quantify bias and uncertainty with some measure of precision. Patrick Ball, Megan Price, and their colleagues at the Human Rights Data Analysis Group have pioneered the application of multiple systems estimation and other statistical methods to work with limited data in post-conflict and ongoing conflict zones.Footnote ⁴⁸ The group is often asked to evaluate or correct traditional casualty counts using their experience with statistical inferences.Footnote ⁴⁹ They have contributed data analysis to both national and international criminal tribunals and truth commissions.

IV Challenges of Data

Data is always an abstraction – a representation of an idea or phenomenon. Data is also a product of its collection method, whether it is a recording of a signal, survey, mechanical trace, or digital log. As the scholar Laura Kurgan explains:

Analysts may try to use algorithms and data to limit human bias and preconceptions in decision-making. However, researchers can’t help but cast a human shadow on facts and figures; data is affected by people’s choices about what to collect, when and how it is collected, even who is doing the collecting. Human rights researchers have begun to call attention to these hidden aspects of data gathering and analysis, examining the rights implications of their elision, and the perils and promise in their use.

For example, data-driven policing based on computerized analyses of arrest and crime data has been advanced as a method for making law enforcement less prone to bias.Footnote ⁵¹ However, the use of algorithms and visualization in such “predictive policing” often amplifies existing assumptions and historical patterns of prejudice and discrimination, driving police to increase scrutiny of already over-policed neighborhoods.Footnote ⁵² A human rights critique is needed to assess the use of algorithms in predictive policing as well as other practices, like the use of computer scoring to recommend sentencing ranges in overcrowded justice systems.Footnote ⁵³ Techniques developed in the algorithmic accountability movement are especially useful here: Audits and reverse engineering can uncover hidden bias and discrimination,Footnote ⁵⁴ which could be assessed against human rights norms.

Metadata describes the origin story of data: the time and place it was created, its sender and receiver, the phone used, network used, IP address, or type of camera. As former US National Security Agency General Counsel Stewart Baker hauntingly put it, “Metadata absolutely tells you everything about somebody’s life. If you have enough metadata, you don’t really need content.”Footnote ⁵⁵ Outside the domain of state security or intelligence, metadata can be useful to human rights researchers and activists as well, for instance, to counter claims that incriminating images or video recordings were falsified or to corroborate that a set of photos were taken in the same place, on the same day, by the same camera. Amnesty International used metadata from photos and videos to corroborate attacks on suspected Boko Haram supporters by Nigerian soldiers, thereby implicating them in war crimes.Footnote ⁵⁶ Building on its experience training grassroots activists to use video for advocacy, the NGO WITNESS worked with the Guardian Project to develop a mobile application called CameraV to help citizen journalists and human rights activists manage the digital media and metadata on their smartphones by automatically encrypting and transmitting media files to a secure server, or, conversely, by deleting and obscuring the metadata when it could put activists at risk.Footnote ⁵⁷

In addition to concerns about accuracy, rights groups should be cautious in their approach to privacy and ownership of data, and to its analysis and expression through visualization. Over the years, researchers and lawyers have developed a set of best practices to guide the proper collection and use of data, with particular attention to human subjects research.Footnote ⁵⁸ Questions related to the collection of data go to the heart of what constitutes ethical research methods: Did the subjects give informed consent regarding the way their personal data would be used? Does using, collecting, or publishing this data put anyone at risk? Is the data appropriately protected or anonymized? The rules about data continue to evolve and are not without gray areas and open questions. Universities in the United States and many other countries have review processes in place to provide guidance and ensure that critical ethical questions are raised before research is approved. In fact, these ethical questions and review processes are required under US law for research institutions that receive federal funding. However, the “common rule” underlying these processes is widely seen as out of date when it comes to data ethics – especially big data ethics.Footnote ⁵⁹ Ethical discussions and guidelines about data visualization are almost nonexistent, with a 2016 Responsible Data Forum on the topic a very welcome outlier.Footnote ⁶⁰ The forum brought together academics, activists, and visualization practitioners to discuss issues such as the ethical obligation to ensure that data is responsibly collected and stored before being visualized; representing bias, uncertainty, and ambiguity in data visualization; and the role of empathy and data visualization in social change.Footnote ⁶¹

V Visualizing Quantitative Data

With data and data visualizations, physical phenomena like the impact of disease and the movement of troops can become legible – as can systems like economies, relationships, and networks of power. Using data to examine policies, populations, actions, and outcomes over time, individual cases can be seen as instances of widespread and systematic patterns of abuse. However, possession of data does not constitute knowledge. Data requires interpretation, context, and framing. Graphics are a powerful way to help contextualize and frame data, present interpretations, and develop understanding. Through visualization and analysis, correlations and patterns of structural violence and discrimination as well as the scope or systemic nature of abuses can become clear.

Data visualization is useful not only for explaining patterns in a dataset, but also for discovering patterns. For its 2011 report A Costly Move, HRW used mapping tools to visualize and analyze patterns in a large dataset of more than five million records concerning the transfer of immigrant detainees around the United States.Footnote ⁶² Analyzing twelve years of data, the group found that detainees were transferred repeatedly, often to remote detention centers, a process that impeded their right to fair immigration proceedings. In 2000, HRW and the American Association for the Advancement of Science visualized statistical analyses of extrajudicial executions and refugee flows from Kosovo to Albania in 1999. Instead of random violence, they found distinct surges of activity that suggested purposeful, planned, and coordinated attacks by government forces.Footnote ⁶³

Data is particularly useful to those seeking to understand structural, systemic violations such as abuses of economic, social, and cultural rights. Taking data from development surveys, activists have used data visualization to compare trends in health,Footnote ⁶⁴ education,Footnote ⁶⁵ housing,Footnote ⁶⁶ and other areas against government budgets, tax revenues, and other economic data to paint a picture of progressive realization of rights against “maximum available resources,” as outlined in the International Covenant on Economic, Social, and Cultural Rights.Footnote ⁶⁷

Within the context of communications and advocacy, one powerful characteristic of data visualization is that it is perceived as scientific. In one study, Aner Tal and Brian Wasnick found that including visual elements associated with science, such as graphs, can enhance a message’s persuasiveness.Footnote ⁶⁸ Our research group at New York University also found that when viewers did not already hold strong opinions against the subject matter, graphics presented in diagrams or charts were more persuasive than the same information presented in a table.Footnote ⁶⁹ In another study, on people’s ability to remember visualizations, Michelle Borkin and colleagues found that specific visual elements of a given presentation affected its memorability. Memorable graphics had a main visual focus, recognizable objects, and clear titles and annotations.Footnote ⁷⁰

The persuasive power of charts and graphs may come at a cost: In a report full of text, numbers crucial to making a rights case are a prime target for attack and dispute.Footnote ⁷¹ The currency of human rights work is credibility, and researchers and program staff at human rights organizations carry the additional burden of having to take special care to protect their credibility and the incontrovertibility of the evidence they present. If a number in a report is convincingly challenged, the rest of the report may be called into question. Charts and numbers are also easily taken out of context, with readers understanding representations of data as statements of fact. This is all the more reason to interrogate the methodology and unpack conditions of production of specific datasets before they are used in visualizations. The powerful impact and memorability of data visualization come with a responsibility to put this knowledge to use with care and attention to the potential ethical pitfalls.

Effective data visualization can make findings clear and compelling at a glance. It provides readers with an interface to navigate great quantities of data without having to drill down into the various data points. This can obscure the fact that visualization is only a part of working with data – and often only a small part. The lead-up to the creation of a data visualization can be the key to its usefulness. Acquiring, cleaning, preparing, and analyzing data very often make up the bulk of the work. When exploring a visualization, the sometimes tedious and decidedly unsexy data work that has been done behind the scenes is not always immediately visible. And given its persuasive power, data visualization in polished and final form may gloss over issues with data collection or analysis. This may be especially true with human rights visualization, where analysis includes normative judgments about the fit between a given dataset and the legal standards at issue.

As noted above, the data used in visualization is subject to bias and the underlying assumptions around data collection and processing. In addition to this, the presentation and design of visualization is also susceptible to distortion and misinterpretation. In a series of experiments performed by our research group at NYU, empirical analysis of common distortion techniques found that these techniques did indeed mislead viewers. Distortion techniques include using a truncated y-axis (starting at a number greater than zero when illustrating percentages) or using area to represent quantity (such as comparing areas of circles.)Footnote ⁷² While manipulation of the facts or deception of the reader is usually unintentional in the human rights realm, accidentally misleading visualizations can affect the clarity of the message and could damage advocacy efforts.Footnote ⁷³ As data and visualization command attention, they can also become the focus of criticism. If a misleading visualization is called to account, it could distract from the credibility of the rest of a given project’s research and advocacy, and perhaps even damage the reputation of the organization.

These risks must be borne in mind as advocates have the opportunity to analyze and visualize the increasing quantities of data made available online as a result of open government efforts. While the call for open sharing of scientific data long predates the Internet, connectivity has spurred an explosion in the use, production, and demand for high-quality data, particularly data collected by government agencies. Governments are sharing great quantities of data online, and making them accessible via Freedom of Information or other “sunshine” requests. Open government data has been used to uncover and analyze patterns of human rights abuse in criminal justice data,Footnote ⁷⁴ inequalities in wage data,Footnote ⁷⁵ unequal burdens of pollution,Footnote ⁷⁶ and the impacts of climate change in environmental data.Footnote ⁷⁷ Data created to track human development, like that collected by international demographic and health surveys, has proven to be fruitful for human rights analysis as well.Footnote ⁷⁸ Under the title “Visualizing Rights,” the Center for Economic and Social Rights (CESR) has published a series of country fact sheets that use publicly available data for analysis and visualization to convey patterns of discrimination and the failure to fulfill rights obligations, e.g., the rights to health, food, and education in Guatemala,Footnote ⁷⁹ or with regard to poverty, hunger, and housing in Egypt.Footnote ⁸⁰ The CESR briefs are designed to be read at a glance by a busy audience of policy officials and individuals who staff intergovernmental human rights mechanisms.

VI Visualizing Qualitative Data

A core output of traditional human rights research is the fact-finding report, which tends to rely on qualitative data such as the testimony of witnesses and survivors of human rights violations.Footnote ⁸¹ Visualization can provide useful context for the broader rights messages in such reports. For instance, to provide visual and spatial context for the findings presented in human rights reporting, it is not uncommon to include a timeline of events,Footnote ⁸² a map of the areas affected,Footnote ⁸³ or a map of towns the researcher visited.Footnote ⁸⁴

Qualitative visualization for human rights generally falls into the category of visual storytelling. Techniques like breaking down an explanation into stages and walking the audience through these stages can elucidate the narrative, building an understanding of the sequence of events or the layers of information. Examining Amnesty International and HRW reports, the 2016 NYU study found that the use of visual features nearly tripled between 2006 and 2014, and that the majority of visual features used were qualitative.Footnote ⁸⁵ For example, the study found that the number of reports using satellite images increased from one in 2006 to four in 2010 to seventeen in 2014. Most maps included in reports during this period displayed geographic information (such as places visited by researchers) and were only rarely used for quantitative display or analysis (such as displaying numbers of refugees).

Some of the changes in human rights reporting were made possible by advances in technology and newly available data. In the 1990s, Global Positioning System (GPS) and high-resolution satellite imagery became available for civilian use.Footnote ⁸⁶ Since then, high-quality satellite imagery has become increasingly accessible from vendors and through free applications like Google Earth and other web-based mapping tools. Human rights groups have used GPS and satellite imagery to present vivid pictures of changes brought about by events such as mass violence, secret detention, extrajudicial executions, internal displacement, forced evictions, and displacement caused by development projects.Footnote ⁸⁷ Satellite imagery has proven especially powerful in showing the visual differences before and after an event.Footnote ⁸⁸ It can show the creation or destruction of infrastructure by marking changes in the landscape designed to hide underground weapons development, or migrations of people by tracking changes in the contours of refugee camps. Satellite images provide local activists with a way to contextualize and document a bigger picture than can be seen from the ground, and enable human rights researchers outside of the country to survey places that are difficult or dangerous to access.Footnote ⁸⁹ These techniques are especially crucial for closed states and in emergency contexts, though researchers based outside of the countries of interest should avoid relying solely on geospatial analysis, since it may not include local voices or context. Integrating local voices with satellite imagery provides both the “near” and the “far” and paints a more complete picture of the situation on the ground.Footnote ⁹⁰

Network graphs are another visual tool that can help illuminate human rights reporting and narrative. Network graphs are a special kind of visualization showing relationships between and among entities. A family tree is one simple example of a network graph. Networks relevant to human rights investigations include networks of corruption, formal and informal chains of command, the flow of resources among industries and the government agencies charged with regulating them, and relationships between military and paramilitary groups. Visualization serves as a useful shorthand, a way to illustrate complex networks that would be cumbersome to describe in text. For example, for its 2003 report on violence in the Ituri region of the Democratic Republic of Congo, HRW used a network graph to illustrate the web of training, funding, and alliances among national governments, national militaries, and local paramilitary groups.Footnote ⁹¹ The graph clarifies the complicity of the national governments in local atrocities. The 2007 Global Witness report Cambodia’s Family Trees uses a network graph to illustrate the connections and relationships among more than sixty individuals and family members, companies, the military, and government agencies in a deeply entrenched web of corruption around illegal logging.Footnote ⁹²

Graph theory, the study of networks and their properties, can be used to model the spread of information or influence along social networks. One of Google’s early innovations was analyzing the network structure of the Internet – i.e., determining which pages are linked to from other pages – in order to rank web pages by relevance. Graph theory algorithms that weigh connections among entities to gauge their importance have proven useful to help navigate millions of pages in document dumps such as WikiLeaks and the Panama Papers. Network analysis and visualization can help make these large sets of data navigable and give researchers and the public a starting point toward understanding connections between parties. Like statistics, network analysis is a tool of social scientists that is increasingly being used by human rights researchers. As noted in Jay Aronson’s chapter (Chapter 6), the Carter Center has used network analysis of social media postings by armed groups in Syria to estimate chains of command and track emerging and shifting alliances among groups.Footnote ⁹³

At the nexus of qualitative and quantitative analysis, Forensic Architecture is an international collaboration that is researching incidents around the world through crime scene reconstructions of human rights violations. Founded by the architect Eyal Weizman, Forensic Architecture uses diverse sources, including photos, cell phone audio and video, satellite imagery, digital mapping, and security camera and broadcast television footage, to painstakingly reconstruct the scene of a violation as a virtual three-dimensional architectural model. The team looks at traces and clues in the data sources. For instance, ascertaining the time of an incident from time stamps on digital metadata and even the fall of shadows in imagery and footage allows the team to establish a sequence of events and uncover falsifications or omissions in recordings. The reconstructions go so far as to adjust the virtual camera lens to match the parallax distortion of video, allowing for analysis of things like line of vision or the position of a munitions impact. The spatial data and architectural model become the nexus that stitch together the reconstruction to determine just what happened at a given point in time and space, how it happened, and who was responsible.Footnote ⁹⁴

Recent developments in machine learning have also made possible a kind of qualitative data analysis of imagery by computers: the use of computer vision algorithms to detect patterns and recognize objects depicted in digital image data. This can include faces or pictures of weapons in massive bodies of social media images, or feature detection in footage from camera-enabled drones, closed-circuit video surveillance, or satellite imagery. Applying these techniques to human rights research, the Event Labeling through Analytic Media Processing (E-LAMP) project at Carnegie Mellon University combines computer vision and machine learning for conflict monitoring by searching through large volumes of video for objects (weapons, military vehicles, buildings, etc.), actions (explosions, tank movement, gunfire, structures collapsing, etc.), written text, speech acts, human behaviors (running, crowd formation, crying, screaming, etc.), and classes of people such as soldiers, children, or corpses.Footnote ⁹⁵ Project ARCADE is a prototype application that uses computer vision to analyze satellite imagery in order to automate the detection of bomb crater strikes and determine their origin.Footnote ⁹⁶ In these instances, after its algorithmic processing, the source imagery is often annotated with visual indicators that are more readily interpreted by humans and the image data is made understandable through a layer of visualization. Such applications are likely to become more common in the human rights field, though, as noted above, machine learning is only as good as its input and the assumptions embedded in it.

VII Technical Decisions Convey Meaning

Given the vital role that data visualization can play in analyzing data and delivering a persuasive human rights message, there is great temptation and good reason to incorporate it into the process of human rights research and its outputs, such as reports and other advocacy products. However, the power of data visualization must be harnessed with care. The techniques of data visualization are a form of knowledge production, with constraints and connotations associated with forms and their interpretation. The meanings conveyed by seemingly technical decisions must be unpacked and considered when designing human rights visualizations.

The keystone technique in the field of data visualization is “encoding” to visually identify a specific aspect of a dataset.Footnote ⁹⁷ Visual encoding associates visual properties like location, color, shape, texture, and symbol with data properties like time, category, and amount. More than one variable and visual encoding can be combined within the same representation, such as x-axis for time, y-axis for magnitude, and color for category.Footnote ⁹⁸ Variables need not be strictly visual, either. Data can be encoded using different aspects of sound (tone, volume, pitch) or touch (height, location, texture).Footnote ⁹⁹ A great power of visualization is the ability to combine multiple encodings within the same visual space, enabling rich exploration. Some kinds of data, like geographic data, can also act as a bridge between other kinds of data. For instance, poverty rates in a particular geographic area can be compared against resources and services available in the same geographic area.Footnote ¹⁰⁰ A standard collection of chart styles has emerged as conventions, and these chart styles come preloaded in popular data tools like Microsoft Excel. As one’s audience becomes familiar with certain chart forms, their legibility is reinforced. As the typographer Zuzanna Licko once noted, “You read best what you read most.”Footnote ¹⁰¹ The pie chart, bar chart, and line chart have become visual conventions. However, the form of a visualization and its interface are not neutral; it constitutes a choice, a way of structuring experience and knowledge. Form imposes its own assumptions and arguments, and the most familiar and common forms of data visualization may not always be suitable for data presentation and can, in some cases, obscure or even distort findings. In the human rights context, it is important to reflect on design decisions and visual conventions – especially when designing research-based products for diverse and often international audiences.

Visual tone can also carry connotations. For instance, does a visualization make beautiful something monstrous and tragic? This is an especially important query in a field marked by its “affect-laden conception of humanity.”Footnote ¹⁰² Presenting visualizations in a “neutral” style could downplay the assumptions behind the data and its collection and analysis. Seemingly less weighty design decisions can also affect the way a map or other graphic is perceived; for example, design decisions that affect color contrast and legibility can obscure a graphic’s meaning or, at their worst, create a misleading visualization.

Color can also carry cultural weight. Consider the color orange, which is associated with Hindu nationalists in India and with Unionism and the Orange Order in Northern Ireland, and was used by groups that participated in the 2004–05 “Orange Revolution” in Ukraine. Depending on how color is used in a given visualization, such associations can invoke secondary cultural meanings, particularly when representing geopolitical data across borders.

VIII Access and Inclusion

Particularly for the sake of advocacy, outreach, and transparency, it is important for human rights information to be accessible and inclusive. Though a strength of data visualization is the ability to invite readers to engage with analysis and communications, an ongoing challenge is access and inclusion. How can human rights researchers and NGOs make their visualizations accessible to the populations to whom their data is relevant? Or work with communities to help them access the tools and expertise necessary to generate their own data visualizations?

One challenge for interactive digital visualization is the physical constraints of the screen, particularly of small mobile screens. The popularity of smartphones and widespread mobile Internet access have overtaken desktop and broadband access in much of the world.Footnote ¹⁰³ While the increasing ubiquity of access is promising, the smaller screen size poses a challenge to making complex data visualizations legible and interactive.

The physical attributes of a visualization and its interaction can profoundly affect how it is accessed and understood by users of different abilities. Limited motor control, color blindness, color vision deficiency, restricted vision, or blindness can affect how a visualization is read. The visualization community has made great strides toward awareness of color blindness and color vision deficiency. But while tools are available to check the accessibility of color use, there is still far more work to be done to make visualizations accessible to blind users. Can visual information be accessed through other means as well, such as accessible HTML, that can be processed by automated screen-reader software? Can the visualization be navigated with a keyboard instead of only by a mouse?

In addition to visual encodings, data visualization relies on culturally coded visual metaphors: an “up” symbol or bigger size means “more,” time moves from left to right, clusters indicate similarity, big is important, etc. As a result, reading, interpreting, and understanding data visualizations require a certain degree of cultural and visual literacy. In the case of quantitative data visualizations, numeracy is key. The success of a data visualization for both analysis and advocacy relies not only on the visualization itself, but on its accessibility to the reader. For its advocacy work to promote better health, the design team behind the now ubiquitous Nutrition Facts label tested more than thirty variations of charts and other formats before settling on the organized table we know so well. They found that graphs, icons, and pie charts were more complicated for consumers than they’d originally thought, requiring a relatively high degree of visual literacy to understand.Footnote ¹⁰⁴

IX Critical Mapping

Mapping is a particularly popular form of data visualization being used in human rights research and advocacy today. Critical cartography is a set of practices and critiques based on the premise that maps are not neutral; it holds that visual design decisions about what to include or exclude, what boundaries to show, etc., are political expressions about space and power. Authors choose the data to include or exclude and decide how to highlight it. Marginalized populations may be excluded from maps for a variety of reasons, or maps may privilege spaces of commerce over spaces of community. Commercial vendors may not consider it profitable to digitize the streets and addresses of villages in developing countries. State statistical agencies with limited resources must inevitably prioritize their activities and focus.

Even when focusing on specific evidence, decisions about what to include or exclude and how to represent visual elements can carry political implications; histories can be contentious, particularly where nationalism and national identities are woven into narratives of conflict. The drawing of maps can raise human rights issues in the act of visualization itself. For example, borders and place names can be particularly contentious. The government of China, for instance, takes border demarcations very seriously – confiscating maps that, through their visuals, “violate the country’s positions on issues such as Taiwan, islands in the South China Sea or territory in dispute with India” or that reveal “sensitive information.”Footnote ¹⁰⁵ Lawmakers in India also considered a draft bill threatening fines and jail time for depicting or distributing a “wrong or false” map of its borders.Footnote ¹⁰⁶ Human rights researchers need to take political sensitivities and significance into account when they engage in visualization involving maps – both the visual expression and where sensitive interactive media is hosted.

Mapping for social justice purposes has a long history, from the historical examples above to the current use of digital mapping. A powerful example of “counter-mapping,” the Detroit Geographic Expedition was formed after the 1967 Detroit riot to conduct and publish research on racial injustice in Detroit and offered free college courses on geography and urban planning for inner-city African American students. The group critically challenged plans put forward by the Board of Education, visualized inequities of Detroit’s public spaces for children’s play, and mapped traffic fatalities of children along commuter routes, all of which pointed to patterns of spatial and racial injustice in the built environment.Footnote ¹⁰⁷

To claim the power traditionally held by governments and companies, grassroots organizations and individuals are also using tools for digital mapmaking. For example, in 1993, Daniel Weiner and Trevor Harris worked with communities in the central lowlands of South Africa to develop participatory applications of GIS in support of the redistribution of natural resources in the post-apartheid transition.Footnote ¹⁰⁸ By far the largest current open mapping collaboration is OpenStreetMap, a free editable map of the world. An ecosystem of tools has been developed around OpenStreetMap data, including some specifically designed for supporting humanitarian responses to crises.Footnote ¹⁰⁹ Other projects are using mapping to capture local knowledge in order to assert claims of land ownership. Inspired by their work with indigenous communities in the Amazon, the organization Digital Democracy developed a method for contributing data to OpenStreetMap without having continuous access to the Internet.Footnote ¹¹⁰ The organization Hidden Pockets found that sexual and reproductive health services in Delhi were absent from Google’s map, so set out to track this data and create its own publicly available map.Footnote ¹¹¹

X Mobilization and Outreach

Data visualization can be a powerful vehicle for collaboration and mobilization in human rights outreach and advocacy. It can be used to interface with activist members, donors, and allies. For instance, using visualization, one can illustrate the impact of one’s findings or recommendations to present a compelling vision of what is possible. Using data visualization not only to describe systemic abuses, but also to render a concrete vision of the future and project an alternative vision and message of hope, can be a powerful way to mobilize supporters. Visually mapping the activities of supporters also provides participants with a visual overview of activities and creates a virtuous cycle of feedback as well as a sense of both transparency and solidarity. Making feedback visible can be an effective way of engaging participants to build solidarity and momentum.

Data visualization for advocacy can also be participatory in public space, inserted into the world. For example, in February 2009, a series of large-scale projections were displayed at sites across the center of Bristol, England. Using a powerful video projector, organizers displayed on building facades the line of the future water level anticipated due to climate change.Footnote ¹¹² On a smaller scale, in 1991, the artist Félix González-Torres created an emotionally powerful series of portraits of friends with HIV/AIDS using piles of candy to match their body weight. Viewers were encouraged to take a piece of candy as they passed each portrait, thereby reducing the weight of the pile, performing, in essence, the wasting and loss of weight caused by the illness before each person’s death, and quietly implicating themselves.Footnote ¹¹³

XI Technical Sustainability

Data visualizations are often presented in paper copies of reports and briefs, but they also figure prominently in web-based communications by human rights organizations, thus making an understanding of that technology essential to maintaining best practices with respect to data and data visualizations. A growing number of human rights visualizations have also moved beyond the presentation of a single view of a given dataset, and instead allow users to explore data in online applications. However, online databases sometimes require ongoing maintenance, particularly when they rely on external services such as maps or timelines hosted by third parties. Amnesty International launched a series of interactive data sites in 2007 with “Eyes on Darfur,” followed by “Eyes on Pakistan” in 2010, “Eyes on Nigeria” in 2011, and “Eyes on Syria” in 2012, to map human rights abuses in those regions. Whereas the 2007 “Eyes on Darfur” project hosted satellite imagery on Amnesty’s own web server, the Syria and Nigeria sites used Google Maps to display points of interest. Google Maps provides a low-cost, easy-to-use, web-based interface to map information, which is attractive for human rights organizations with budgetary constraints. However, as Amnesty International’s experience illustrates, reliance on a third-party service provider comes with a long-term cost: By 2016, Google had updated its interface and both “Eyes on Syria” and “Eyes on Nigeria” no longer functioned. Amnesty would be required to update the back-end code for these sites in order to continue to plot information on Google Maps. Though human rights and humanitarian crises continue in these countries, the “interactive evidence” in Amnesty International’s visualization has become inaccessible.

Periodic updating of interactive sites is, in fact, essential to maintaining their accessibility and relevance. While “Eyes on Darfur” continues to function, its use of FlashFootnote ¹¹⁴ to display information makes it inaccessible on tablet and mobile devices, which have become popular since the site’s development in 2007 and do not support Flash. The Pakistan site featured “a geocoded database of more than 2,300 publicly reported incidents occurring between 2005 and 2009.”Footnote ¹¹⁵ A collection of rights-related incidents of that scale represents a treasure trove of possible human rights cases and a powerful baseline by which to judge reports of ongoing abuse. As of 2016, however, eyesonpakistan.org is no longer online and its database is no longer accessible. The apparent failure to prioritize or plan for the demands of changing technology and the lack of ongoing technical support means it is no longer possible for human rights researchers and advocates to use the Syria, Nigeria, or Pakistan data.

One way to mitigate the obsolescence that plagued the Amnesty sites is to make both the data and the source code of a visualization readily available for download by users. This would allow visitors to access the raw data regardless of the technical implementation of the interactive interface. Though still rare among human rights NGOs, this is a growing practice among news organizations. The evolving field of data journalism offers one view of things to come. “Computer-assisted reporting” and “data-driven journalism” use spreadsheets and databases to help find patterns in data by using statistical methods and other techniques from the social sciences. The findings of the investigations are often presented through interactive news applications and data visualizations that can engage readers with data-driven reporting in a richer way, beyond the constraints of print. News organizations are increasingly posting these projects, their data, and tools to code-sharing websites like GitHub for others to download, use, and modify.Footnote ¹¹⁶ Like human rights organizations, journalists face limited resources and technical overhead, but news media are, thus far, more readily embracing the use of data analysis to drive investigations and visualization for effective storytelling. Any effort to share the data behind human rights reporting or visualization will need to carefully grapple with crucial ethical and security challenges, including confidentiality and anonymization, consent, and potential misuse of the data by abusive governments or other opponents.

XII Conclusion

Human rights researchers and advocates are adding new methodologies to their toolbox, drawing on emerging technologies as well as established data analysis techniques to enhance and expand their work. Data visualization holds exciting potential, bringing new techniques of knowledge production, analysis, and communication to human rights research and advocacy. Organizations are increasingly recognizing the power of data visualization to support human rights analysis and arguments to help make a memorable and persuasive case for change.

Enabled by digital technology and engagement with data, effective visualization is a powerful tool for understanding social problems and their potential solutions. While journalism and academic disciplines, including the social sciences, are using data visualization for both analysis and communication, the human rights field is just beginning to tap its potential. As interest grows, human rights organizations will need to struggle with the ethical and practical considerations of producing data visualizations.Footnote ¹¹⁷ More research is still needed on the effective use of data visualization and human rights. Used in a principled way, however, data visualization can benefit human rights researchers and advocates, and those whose rights are in danger. It can help researchers identify patterns and trends; clarify a call to action; make data analysis compelling, understandable, and interactive; rally supporters; and perhaps even visualize the effects of activism itself.

I IntroductionFootnote ¹

The rise of information and communication technologies (ICTs) has captivated many human rights practitioners and scholars. Particular interest, mine included, is focused on the potential of using ICTs to support the pluralism of human rights fact-finding and advocacy.Footnote ² In theory, now anyone with a cell phone and Internet access can document and disseminate evidence of human rights abuses. But what happens when this theory is put into practice?Footnote ³ What happens when ICTs are adopted in empirical realities shaped by unique contexts, distributions of resources, and power relations?Footnote ⁴ I will argue that, while the rise of ICTs has certainly created new opportunities, it has also created new risk – or negative outcomes – for human rights practitioners. This risk is silencing, and unequally so.

In this chapter, I focus on human rights fact-finding and advocacy from the perspective of practitioners at human rights NGOs, while acknowledging that the range of practices and actors involved in human rights work is much broader.Footnote ⁵ These practices form a communication chain: information moves from witnesses on the ground to human rights practitioners during fact-finding, who gather and evaluate this information for evidence of violations. This evidence is then packaged and communicated to audiences such as journalists, policy-makers, and publics as advocacy work designed to impel change through persuasion.Footnote ⁶ At each stage, we can think of successful communication as speaking to and being heard and understood by intended audiences.Footnote ⁷ In other words, it is about audibility (or its equivalent, visibility). Unsuccessful communication, in contrast, involves either audibility to unintended audiences or inaudibility to intended audiences. Successful communication can also have unsuccessful outcomes for the actors involved, as when a message is audible to intended audiences but is misunderstood or turns out to be erroneous, or when a message is received and interpreted as the communicator intended but turns out to be deceptive.

The success of communication matters, of course, for human rights practitioners’ ability to generate accountability for individual cases of human rights violations. It also matters for a value at the core of human rights: pluralism, or the successful communication of a variety of voices. Three types of pluralism are of concern in this chapter. The first is the pluralism of human rights actors vis-à-vis the state or non-state actors they wish to hold to account. The second is the pluralism of individual human rights actors within the human rights world, which, as with all worlds, has hierarchies corresponding to the distribution of power.Footnote ⁸ The third is the pluralism of access by the subjects and witnesses of violations to the mechanisms of human rights accountability, which, of course, cannot act on a violation without first hearing about it.Footnote ⁹

The chapter begins by outlining how risk is entwined with communication in the digital age. Rather than considering risk in isolation, we can think of it as manifesting via “risk assemblages,” or dynamic combinations of actors, technologies, contexts, resources, and risk perceptions.Footnote ¹⁰ In the subsequent two sections, I detail selected types of risk for human rights communication resulting from new combinations of actors and technologies involved in digital fact-finding and advocacy. For fact-finding, these include the risk of surveillance, which has consequences for participants’ physical security, and the risk of deception, which has consequences for their reputational integrity. For advocacy, these include the risk of mistakes, which can in turn risk reputational integrity, and the risk of miscalculations, which can jeopardize precious resources. In the following section, I explain how this materialized risk combines with risk perceptions to create a silencing double bind. Human rights practitioners may be silenced if they don’t know about risk – and they may silence themselves if they do. This silencing effect is not universal, however, but disproportionately affects human rights practitioners situated in more precarious contexts and with less access to resources.Footnote ¹¹ This has consequences for the three types of pluralism outlined above. The chapter finishes by outlining four ways of loosening the risk double bind: educational, technological, reflexive, and discursive approaches to working with risk.

II Communication, Mediation, and Risk in the Digital Age

As communicators, we all do a number of things to increase the odds that our communications are successful. We establish the identities of our communication partners through clues we gather from their appearance and bearing. We supplement our messages with cues such as facial expressions or emoticons to guide interpretation, and we look for cues from our audiences that they have heard and understood us.Footnote ¹² We gather information about our interlocutors’ context – the time and place in which they are communicating – and supplement our messages with information about our own contexts (often referred to as metadata). We adjust our production and reception of content to these clues, cues, and contextual information. Still, even with all of these aids, communication can be unsuccessful, and this risk is exacerbated by the mediation of communication over ICTs.

Mediation is the extension of communication across time and/or space using technology. The closer we are in time and space to our communication partners, the easier it tends to be for us to establish their identities, observe and provide cues, and understand context. Easiest of all is face-to-face communication. By introducing “temporal and spatial distances,” mediation makes all of this more difficult, as we are no longer in the same environment.Footnote ¹³ It is not, however, just this distance that increases the risk of unsuccessful communication, but also the introduction of intermediaries. These intermediaries are not neutral, but rather introduce new technical features and new actors with new motives, as well as new ways for existing actors to intervene with communications.

The technical features of ICTs can diminish, augment, or alter the clues, cues, and contextual metadata associated with a communication. Furthermore, the complexity of these technical features may make it difficult to understand just what has happened. For example, many social media user profiles allow people to communicate with pseudonyms or assumed identities. Twitter’s character limit squeezes nuance out of tweets, though users have introduced the use of hashtags as an abbreviated interpretation cue. In another example, YouTube automatically dates videos according to the day it is in California at the time of upload, no matter where the upload took place. This metadata is widely misunderstood and has contributed to disputes about the veracity of videos.Footnote ¹⁴

A significant proportion of new actors behind ICTs are commercial, governed by profit motives. These motives can shape technical features, like Facebook’s “like” and “share” buttons, which are designed to keep eyeballs on timelines peppered with advertisements. The motives of commercial communication platforms may not necessarily align with the motives of communicators. As discussed further below, this is particularly evident in the algorithms that determine visibility on social media and thus who is seen by whom. These algorithms may make certain communications either more or less visible than their producers intended. The phenomenon of commercial intermediaries controlling public visibility is nothing new – think of the gatekeeping role of mainstream news organizations. What is new is the lack of transparency and accountability when visibility decisions are made by a black-box algorithm instead of a human journalist.Footnote ¹⁵ Just as the technical complexity of ICTs obscures these algorithms and the commercial motives underpinning them, it also hides third-party actors. These include political actors who have a vested interest in human rights communication. The market for digital surveillance is thriving, and hardware and software that allow us to communicate over time and space also create opportunities for eavesdropping. In sum, at the same time as communicators using ICTs usually can glean less about their interlocutors and eavesdroppers than in a face-to-face situation, they must also know more about intermediary technologies that are both complex and opaque.Footnote ¹⁶ Mediation thereby increases the risk of unsuccessful communication and its attendant consequences.

Alongside many other professional worlds of communication, human rights practitioners are considering and adopting new ICTs. This use of ICTs and the mediation they engender supplements other forms of communication, creating a new “interaction mix” characterized by renewal, as fresh technologies proliferate and slightly stale ones become obsolete.Footnote ¹⁷ In terms of human rights fact-finding, this new mix has been described as enabling a new generation of methodologies.Footnote ¹⁸ Traditionally, the gold standard of fact-finding has been the face-to-face interview between civilian witnesses and human rights practitioners. Often facilitated by trusted networks cultivated over time, the interview allows for the co-production of information between the witness and the practitioner. This witness testimony and the accompanying analysis done by human rights practitioners are the cornerstones of the weighty, precisely worded, and highly documented orthodox human rights report.Footnote ¹⁹ These reports, in turn, underpin human rights advocacy, which practitioners traditionally – though not exclusively – communicated to targets via the mainstream media.Footnote ²⁰

Human rights communication has therefore always been mediated, whether information is passed through a trusted network of witnesses or shaped to attract the attention of journalists covering human rights violations.Footnote ²¹ It has also always entailed risk. One only has to dip into the multitude of reports on the conditions of human rights practice to see this – or to consider practitioners’ risk-mitigation tactics, ranging from security training to robust and transparent methodologies to publicity strategies.Footnote ²² But the new mix of human rights fact-finding and advocacy in the digital age has brought about new risk assemblages shaped by technologies, actors, contexts, resources, and risk perceptions.Footnote ²³ Over the next three sections of this chapter, I outline elements of these new risk assemblages and explain how they can hinder successful communication, with implications for the pluralism of human rights communication.

III Digital Fact-finding and Communication Risk

Human rights practitioners have adopted ICTs for fact-finding in a variety of ways, including using high-technology information sources like satellite images, drone videos, big data, and statistics as well as open source social media content.Footnote ²⁴ Given our concern with communication, here I focus on practitioners’ use of digital information that documents human rights violations and has been produced and transmitted by civilian witnesses – “civilian” in contrast with professional to highlight their inexpert status, and “witness” as someone who is purposively communicating experienced or observed suffering.Footnote ²⁵ Civilian witnesses can be spontaneous or solicited.Footnote ²⁶ In the digital age, spontaneous witnesses might use their smartphones to document violations that they then share with broader audiences via social media or messaging apps; sometimes this information is gathered, curated, and connected to human rights NGOs by networks of activists. Solicited witnesses may be answering a human rights NGO’s open call for information made via a digital crowdsourcing project or a digital reporting application.

Digital information from civilian witnesses affords human rights practitioners a number of fact-finding advantages. First, the images and video civilian witnesses produce can provide much more detailed evidence than witness interviews that rely on memory.Footnote ²⁷ Second, consulting civilian witnesses can tap wells of knowledge, particularly expertise relating to local contexts unfamiliar to foreign practitioners. Third, a wider incorporation of civilians via ICTs can fire up public enthusiasm about human rights and thus receptivity to advocacy.Footnote ²⁸ Fourth, and most important for our concern with pluralism, these new sources can support the variety and volume of voices speaking and being heard on human rights. They supplement interviewing’s traditional co-production of information between witnesses and practitioners with both the more autonomous production of spontaneous digital witnesses and new forms of co-production via solicited digital witnesses.Footnote ²⁹ If these witnesses are situated in closed-country contexts or rapidly unfolding events, they might otherwise be inaccessible to human rights practitioners.Footnote ³⁰ Indeed, fact-finding in a number of recent cases has hinged on evidence documented digitally by civilian witnesses. For example, Amnesty International’s research into a 2017 shooting at an Australian refugee detention center in Papua New Guinea used refugees’ photos and videos to challenge both governments’ official version of events, which was that Papua New Guinea Defence Force soldiers fired bullets into the air rather than into the center.Footnote ³¹

These opportunities are all made possible by ICTs’ mediation of communication over time and place. Of course, this mediation, and the intermediaries it requires, also introduces risk. Below, I outline two possible manifestations of communication risk and their consequences arising from the introduction of new technologies into fact-finding, associated new commercial actors, and new opportunities for existing actors to interfere with communications. The first is the risk of surveillance, in which the communication is audible to unintended recipients and generates concomitant risk for the physical security of civilian witnesses and human rights practitioners. The second is the risk of deception, in which the producer of a digital communication engineers the recipient’s misinterpretation of that communication. Misinterpretation creates follow-on risk to the reputational integrity of human rights practitioners and their NGOs. These are familiar categories of risk in the human rights domain but manifested, as explained below, in new ways. Both are made possible by the technical complexity of mediating ICTs, which allows eavesdroppers to hide and deceivers to manipulate metadata.

IV Digital Advocacy and Communication Risk

As with fact-finding, human rights practitioners are incorporating ICTs into their advocacy strategies, venturing far beyond websites into formats including apps, livestreaming, and virtual reality. Because human rights practitioners are paying particular attention to mainstream social media platforms to supplement their traditional advocacy practices, I focus on that medium here.Footnote ⁵⁵ Practitioners are communicating advocacy messages via social media to directly target policy-makers, either publicly or via private messages, and to attract the attention of the mainstream media.Footnote ⁵⁶ They are also using social media to mobilize publics for a variety of reasons, including fundraising, creating visibility for an issue, and building networks between publics and subjects of violations in a show of global solidarity.Footnote ⁵⁷ Many NGOs undertake advocacy over social media through institutional accounts operated by individuals. Though dedicated communications professionals operate these accounts at some organizations, at others the arrangement is more ad hoc, undertaken by existing staff according to interest or availability.

The use of social media affords human rights practitioners a number of advocacy advantages. It can allow them to amplify messages and reach advocacy targets without depending on the mainstream media, whose human rights coverage may be circumscribed by commercial imperatives, censorship, and norms of newsworthiness.Footnote ⁵⁸ The range of communication formats supported by social media enables development of new and captivating ways to represent human rights information, such as data visualization.Footnote ⁵⁹ Additionally, the quantification metrics built into social media platforms, such as numbers of likes and shares, allow human rights practitioners to track engagement with their messages.Footnote ⁶⁰ They can then incorporate these numbers into their campaigns targeted at policy-makers to quantify public support for their advocacy aims.Footnote ⁶¹ A wide variety of human rights advocacy communications over social media exists, such as the 2013 Thunderclap campaign created by EDUCA, an NGO based in Oaxaca, Mexico, to raise awareness about ongoing human rights violations there. Thunderclap is a digital platform that allows users to coordinate their supporters’ automatic participation in a onetime, synchronized mass social media posting of a particular message.Footnote ⁶² EDUCA surpassed its goal of 100 supporters, and its Thunderclap – timed to coincide with the October 23 UN Universal Periodic Review of Mexico’s human rights record – reached more than 58,000 people via social media.Footnote ⁶³

Again, however, the advantages of social media for advocacy are accompanied by risk, and here I detail two types of communication risk and their consequences. Both stem from the introduction of new technologies into advocacy, which in turn introduces new actors with new motives. Human rights practitioners are accustomed to considering the motives of intermediaries and their intended audiences in shaping their advocacy strategies. For example, they cater to the “media logic” of mainstream media outlets, tailoring the tone and theme of their content as well as building their identities as credible sources to meet journalists’ exigencies.Footnote ⁶⁴ The use of social media intermediaries, however, requires them to shape advocacy messages in light of new “social media logics.”Footnote ⁶⁵ These are also commercially driven motives, manifested in new technical features. Like journalists and journalism, these technical features can be inscrutable to human rights practitioners and incompatible with human rights advocacy – but in different ways.Footnote ⁶⁶ Conducting advocacy via these intermediaries thus introduces new facets to existing risk. This risk includes audibility to unintended audiences, which I refer to as mistakes that can have reputational consequences. The second variety of risk addressed below is inaudibility to intended audiences, or advocacy miscalculations that waste resources.

V Risk Assemblages, Pluralism, and Inequality

Returning to the three types of pluralism introduced earlier in the chapter, it is clear that the manifested forms of risk outlined above have silencing effects on the first category – the pluralism of the human rights world vis-à-vis the world of power it aims to hold to account. New mediating technologies, with commercially driven technical features that complicate communication, fuel new communication cultures and allow new spaces for adversaries to intervene. Surveillance, through its consequences for physical security, can stop human rights practitioners from speaking. The susceptibility of practitioners to deception and mistakes, with the repercussions for reputations, may deafen advocacy targets to their communications. Advocacy miscalculations may prevent advocacy targets from hearing those communications at all. In order to understand the effects of communication risk on the other types of pluralism, however, we must further develop our understanding of these new risk assemblages. We must also think about context, resources, and risk discourses.

As materialized risks are always embodied, an individual practitioner’s context and resources matter in understanding how risk impacts the second type of pluralism, namely pluralism within the human rights world. Context – or individuals’ “social risk positions” ranging from their political environments to their positions within relevant social hierarchies – influences exposure to risk.Footnote ⁸⁸ In turn, the resources individuals have at hand influence their ability to mitigate risk. Key here is the resource of expertise, such as digital literacy about computer security, knowledge of digital verification practices, and facility with social media. Also relevant are the resources that can be used to secure expertise, including money, of course, but also the social capital and reputations that can connect practitioners to expertise and convince experts to share it. The same resources can be used to secure physical and digital safeguards. The upshot is that risk curtails pluralism within the human rights world by silencing practitioners unequally.

Inequalities in contexts and resources intersect with the types of risk enumerated above in a variety of ways. The risk of surveillance depends greatly on the proclivities of a practitioner’s political opponents for purchasing surveillance technologies and enacting pro-surveillance legislation. It also depends on the practitioner’s networks, whose resistance to surveillance is only as strong as their weakest links; one member falling prey to malware can unwittingly expose all her communication partners.Footnote ⁸⁹ Security literacy is crucial. As a practitioner at an organization that trains human rights reporters on digital security once told me, “A lot of them don’t know that Facebook is where a lot of people who would target human rights defenders go shopping.” Security literacy is expensive, in terms of money and time, and it is daunting; therefore, it is more accessible to some than to others.Footnote ⁹⁰

Deception via the manipulation of truth-claims is also a risk that human rights practitioners experience differently. Like surveillance, this risk is conditional on the political context, since some governments are particularly inclined to engage in information wars. Associated reputational risks are not isolated, but rather may have repercussions for a practitioner’s networks. This is because human rights organizations build their credibility in part through networks of association with credible peers; one organization’s loss of credibility allows opponents to tar its network with the same brush.Footnote ⁹¹ Some organizations can weather a hit on their credibility better than others. As human rights organizations also build credibility through performance over time, a more well-established NGO would have more reputational capital to counterbalance an instance of susceptibility to deception or a mediation-related mistake.Footnote ⁹²

The risk of advocacy mistakes and miscalculations can be mitigated by human rights organizations’ in-house social media expertise and consequently the money required to acquire this expertise. Funds also allow human rights organizations to buy visibility for their social media communications through targeted advertisements. Those with fewer resources to dedicate to social media advocacy are, unfortunately, more likely to waste resources by engaging in this practice. This is evident in the results of a recent study, which found that, of 257 sampled human rights NGOs, the richest 10 percent had 92 percent of the group’s total Twitter followers, 90 percent of their views on YouTube, and 81 percent of their likes on Facebook. The study also found that social media advocacy does not seem to help NGOs set the agenda in the mainstream media – further evidence that unsuccessful digital communication can curtail the greater pluralism that using ICTs could bring, both within the human rights world and vis-à-vis the world of power.Footnote ⁹³

A major purpose of the first and second forms of pluralism is to support the third form: the pluralism of civilian access to human rights mechanisms. Civilians cannot access accountability without their voices – their accounts – being heard. If the NGOs representing them are silenced, they too may be silenced. So, communication risk restricts civilian access to the mechanism of human rights unequally as well. As this effect maps onto context and resource distributions, this means that civilians in more precarious contexts with relatively few resources – in other words, those who might most need human rights mechanisms – are more likely to be silenced. The networked nature of human rights NGOs, which are characterized by solidarity, information exchange, and international communication, goes some way to counteract this effect, as another organization may be able to pick up the communication chain.Footnote ⁹⁴ Still, while ICTs do create human rights communication channels where none existed before, we must be alert to the possibility that they do not level inequalities of audibility, but rather extend them.

So far, this chapter has looked at materialized risk, but risk perception is just as important for understanding the human rights practitioner’s lived experience of risk.Footnote ⁹⁵ It is also just as important for understanding how the risk accompanying use can impact the pluralizing potential of ICTs. As evident from interviews with some human rights practitioners, in which they qualified their view of ICTs with words such as “terrified” and “scary,” knowing about risk can be distracting and even debilitating. The more complex the risk assemblage, the stronger this effect, as it is more difficult to understand and predict the risk. This knowing but not knowing exactly brings its own anxieties.Footnote ⁹⁶

Risk perception is not necessarily accurate, in part because risks are hard to estimate and because the idea of them can be overwhelming. Furthermore, as explained below, the specter of risk associated with a practice may have been conjured on purpose to prevent people from undertaking that practice; it may be a discourse deployed in the pursuit of power.Footnote ⁹⁷ A full exploration of risk perception, which is outside the confines of this chapter, would consider the practices individuals adopt in anticipation of these risks and would investigate how these practices affect pluralism. For example, some human rights practitioners are renouncing digital communication methods for a return to analog.Footnote ⁹⁸ Some are slow to adopt digital information from civilian witnesses for fact-finding.Footnote ⁹⁹ As mentioned above, practitioners introduce protracted review systems for social media communications and pay for the visibility of their social media messages, and thus the success of their digital communications depends on the resources of time and money. Risk perception can also silence, and unevenly so. Furthermore, as practitioners weigh up pluralism versus security in deciding whether or not to communicate digitally, erroneous risk perception can swing the balance too far to security.

What we have here, then, is a risk double bind – risk is bad for pluralism if you know about it, and it is bad for pluralism if you don’t. If the latter, human rights practitioners are more likely to fall prey to communication risk. If the former, risk perception can prevent them from communicating digitally in the first place. This creates its own follow-on risk, like missing vital pieces of evidence or being dismissed as Luddites in the context of a broader pro-technology zeitgeist that has enthused donors. Though this double bind can make practitioners feel caught between paralysis and propulsion, it is not impervious to resistance. Next, I offer four approaches to loosening the silencing risk double bind.

VI Loosening the Silencing Risk Double Bind

The silencing risk double bind, constructed in part by commercial and political actors, threatens to squeeze the pluralism potential from human rights practitioners’ adoption of ICTs. Political adversaries of the human rights world benefit directly from this silencing effect. The commercial actors of social media companies profit from human rights practitioners shaping their communications to social media logics, which can have silencing consequences. Human rights practitioners – as well as all those involved in the human rights and technology space, such as scholars, technologists, and donors – can, however, counteract these forces. In outlining four approaches to loosening the risk double bind, this chapter moves beyond the techno-pessimistic enumeration of materialized risk, and its potential contribution to silencing risk perception, toward a techno-pragmatic position. The four approaches, which work best in tandem, support the development and adoption of ICTs for human rights pluralism. The first pair of approaches, involving education and technology, are about mitigating materialized risks, while the second pair, involving reflexivity and discourse, relate to the construction and perception of risk. As human rights practitioners know very well, risk is an unavoidable element of their work; the aim is not to eliminate it, but to work alongside risk without it getting in the way.

VII Conclusion

Techno-optimism has surfaced in the human rights world, as in many others, based in part on the perceived benefits of ICTs for the pluralism of human rights communication. These benefits have been realized in a number of cases, but the application of ICTs has also materialized risk. As human rights practitioners consider whether and how to incorporate ICTs into their practices, this chapter has sought to outline some types of risk they may face and associated consequences for human rights pluralism. This risk, I argue, is a product of ICTs’ affordance for mediation, or communication across time and place. This mediation, and the technical features it requires, alters the identity clues, interpretation cues, and contextual information communicators draw upon in order to increase the likelihood that their communication is successful.Footnote ¹¹⁹

Furthermore, the use of ICTs introduces new intermediary actors to the human rights communication chain, and the technical complexity of ICTs makes these actors and their impact on communication more difficult to identify and assess.Footnote ¹²⁰ Of particular note here are new commercial actors with profit motives. To be sure, human rights reporters have interacted with commercial motives before in their communication practices, such as in considering the marketability of newsworthiness decisions.Footnote ¹²¹ Never before, however, have commercial actors been so influential over and yet so hidden in mediation.Footnote ¹²² Cases in point are the commercial-political surveillance nexus, the lucrative gray market for spyware, and the proprietary, revenue-maximizing algorithms of social media platforms. Incorporating ICTs into human rights fact-finding and advocacy contributes to new risk assemblages for human rights practitioners.

The types of risk outlined here are by no means the only ones that ICTs introduce or exacerbate for the human rights world. Others include the risk to human rights practitioners of secondary trauma brought on by exposure to images and videos of violations, or the retraumatization of individuals featured in advocacy material, particularly if the material is re-mediated and re-mixed.Footnote ¹²³ The types of risk detailed here, however, have particular consequences for human rights pluralism. In digital fact-finding, human rights practitioners face surveillance risk that can imperil their physical security and deception risk that can jeopardize their reputational integrity. In digital advocacy, they encounter the risk of mistakes that have negative repercussions for reputations, as well as the risk that miscalculation poses for their resources. Some of these materialized risks and their repercussions silence human rights practitioners and civilian witnesses, while others deafen intended audiences to human rights communication. The perception of these risks can also be silencing, leading to a risk double bind in which both knowing and not knowing about risk can curtail human rights communication.

Acknowledging the silencing risk double bind throws into relief the importance of thinking about risk not in isolation, but rather as socially constructed. These social contexts produce values and connect individuals that could end up on opposite sides of a risk trade-off. In deciding whether or not to speak in the face of risk, human rights practitioners are choosing between the value of pluralism and the value of security. In so doing, they are also choosing between types of follow-on risk: the risk of physical harm and harm to reputations and resources if they choose pluralism, and the risk of ongoing human rights violations if they choose security. This means they are also making choices between risk populations.

The silencing risk double bind can feel unstoppable, part of the “juggernaut” of rapidly advancing technological change – with its associated complexities, inscrutable interconnections, and risk – that characterizes contemporary societies.Footnote ¹²⁴ Yet, silencing is not inevitable. This chapter proposes four approaches to loosening the risk double bind: the educational and technological, which can limit materialized risk, and the reflexive and discursive, which can stay the construction of risk and erroneous risk perceptions. For practitioners, technologists, donors, and scholars, these approaches are useful heuristics for assessing risk. These heuristics also support human rights practices that allow successful digital communication to coexist with risk rather than be dictated by it.

The net impact of ICTs on the pluralism of the human rights world vis-à-vis the world of power it aims to hold to account is difficult to determine. What we can establish, however, is that materialized and perceived risk curtail pluralism unevenly within the human rights world. This dampening effect is stronger for human rights practitioners in more perilous political and social contexts and with less expertise and associated resources. It is not only particular organizations that are more affected by the materialized and perceived risk of digital human rights fact-finding and advocacy, but also the particular populations and particular human rights that they represent. For a world fundamentally concerned with pluralism, this momentum toward the use of technology creates risk for human rights enforcement in general, as it may be reinforcing inequalities around who speaks and gets heard on which human rights.