To send content items to your account,
please confirm that you agree to abide by our usage policies.
If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your account.
Find out more about sending content to .
To send content items to your Kindle, first ensure email@example.com
is added to your Approved Personal Document E-mail List under your Personal Document Settings
on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part
of your Kindle email address below.
Find out more about sending to your Kindle.
Note you can select to send to either the @free.kindle.com or @kindle.com variations.
‘@free.kindle.com’ emails are free but can only be sent to your device when it is connected to wi-fi.
‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.
Decades of research has debated whether women first need to reach a “critical mass” in the legislature before they can effectively influence legislative outcomes. This study contributes to the debate using supervised tree-based machine learning to study the relationship between increasing variation in women's legislative representation and the allocation of government expenditures in three policy areas: education, healthcare, and defense. We find that women's representation predicts spending in all three areas. We also find evidence of critical mass effects as the relationships between women's representation and government spending are nonlinear. However, beyond critical mass, our research points to a potential critical mass interval or critical limit point in women's representation. We offer guidance on how these results can inform future research using standard parametric models.
We perform a sparse identification of nonlinear dynamics (SINDy) for low-dimensionalized complex flow phenomena. We first apply the SINDy with two regression methods, the thresholded least square algorithm and the adaptive least absolute shrinkage and selection operator which show reasonable ability with a wide range of sparsity constant in our preliminary tests, to a two-dimensional single cylinder wake at $Re_D=100$, its transient process and a wake of two-parallel cylinders, as examples of high-dimensional fluid data. To handle these high-dimensional data with SINDy whose library matrix is suitable for low-dimensional variable combinations, a convolutional neural network-based autoencoder (CNN-AE) is utilized. The CNN-AE is employed to map a high-dimensional dynamics into a low-dimensional latent space. The SINDy then seeks a governing equation of the mapped low-dimensional latent vector. Temporal evolution of high-dimensional dynamics can be provided by combining the predicted latent vector by SINDy with the CNN decoder which can remap the low-dimensional latent vector to the original dimension. The SINDy can provide a stable solution as the governing equation of the latent dynamics and the CNN-SINDy-based modelling can reproduce high-dimensional flow fields successfully, although more terms are required to represent the transient flow and the two-parallel cylinder wake than the periodic shedding. A nine-equation turbulent shear flow model is finally considered to examine the applicability of SINDy to turbulence, although without using CNN-AE. The present results suggest that the proposed scheme with an appropriate parameter choice enables us to analyse high-dimensional nonlinear dynamics with interpretable low-dimensional manifolds.
Prolonged sitting in a fixed or constrained position exposes aircraft passengers to long-term static loading of their bodies, which has deleterious effects on passengers’ comfort throughout the duration of the flight. The previous studies focused primarily on office and driving sitting postures and few studies, however, focused on the sitting postures of passengers in aircraft. Consequently, the aim of the present study is to detect and recognize the sitting postures of aircraft passengers in relation to sitting discomfort. A total of 24 subjects were recruited for the experiment, which lasted for 2 h. Furthermore, a total of 489 sitting postures were extracted and the pressure data between subjects and seat was collected from the experiment. After the detection of sitting postures, eight types of sitting postures were classified based on key parts (trunk, back, and legs) of the human bodies. Thereafter, the eight types of sitting postures were recognized with the aid of pressure data of seat pan and backrest employing several machine learning methods. The best classification rate of 89.26% was obtained from the support vector machine (SVM) with radial basis function (RBF) kernel. The detection and recognition of the eight types of sitting postures of aircraft passengers in this study provided an insight into aircraft passengers’ discomfort and seat design.
The alcohol policy environment was shown to exert a preventive effect on alcohol consumption. However, little is known about the heterogeneity of this effect.
To capture the extent of heterogeneity in the relationship between the strictness of alcohol policy environments and heavy drinking and to identify potential effect modifiers.
Method: Cross-sectional data from 5986 young Swiss men participating in the cohort study on substance use risk factors (C-SURF) in Switzerland was analysed. Self-reported risky single-occasion drinking (RSOD, drinking 6 standard drinks or more on a single occasion at least monthly) in the past 12 months was the outcome of interest. A previously-used index of alcohol policy environment strictness across Swiss cantons was analysed in conjunction with 21 potential effect modifiers. Random forest machine learning and individual conditional expectations captured high-dimensional interaction effects and the heterogeneity induced by the interaction effects and identified potential effect modifiers.
Subject-specific absolute risk reductions ranged from 16.8% to -4.2%, with the latter implying a risk increase. Four prototypical subgroups were evident: “preventive” (alcohol policy environment decreased RSOD risk), “causative” (alcohol policy environment increased RSOD risk), “immune” (no effect due to low RSOD baseline risk), and “doomed” (no effect due to high RSOD baseline risk). Antisocial personality disorder and sensation seeking were major effect modifiers that reduced the preventive effect of stricter alcohol policy environments.
Conclusion: Whereas stricter alcohol policy environments were associated with a reduced RSOD risk, adding selective prevention measures that target high-risk subpopulations is necessary.
There is a significant relation between old-age depression and subsequent dementia in patients aged 50. This supports the hypothesis of old-age depression being a predictor, and possibly a causal factor, of subsequent dementia. The number of people aged 60 years and over has tripled since 1950, reaching 16% in 2050, leading to new medical challenges. Depression is the most common mental disorder in older adults, affecting 7% of the older population. Dementia is the second most common with about 5% prevalence worldwide, but it is the first leading cause of disease burden.
Early detection and treatment is essential in promoting remission, preventing relapse, and reducing emotional burden. Speech is a well established early indicator of cognitive deficits. Speech processing methods offer great potential to fully automatically screen for prototypic indicators of both dementia and depressive disorders.
We present two different methods to detect pathological speech with artificial neural networks. We use both deep architectures, as well as more traditional machine learning approaches.
The models developed using a two-stage deep architecture achieved 59% classification accuracy on the test set from DementiaBank. Our CNN system achieved the best classification accuracy of 63.6% for dementia, but reaching 70% for depressive disorders on the test set from Distress Analysis Interview Corpus.
These methods offer a promising classification accuracy ranging from 63% to 70%, applicable in an innovative speech-based screening system.
Artificial intelligence and machine learning are increasingly being researched within the field of psychiatry to find out what use it might be. With this review, therefore, we would like to assess what literature, if any, exists that answers the question of whether this technology can be useful for providing dementia care. We also wanted to consider the ethical questions of autonomy, consent and privacy when working with this vulnerable group of patients.
To identify and appriase the literature to assess the existing research landscape of the area of machine learning and AI, relating to the care of people with dementia.
A literature search was conducted, searching the PsychInfo, Medline, PubMed and Embase databases. We assessed the quality of the research and considered what overall findings there were in the existing literature.
619 papers were identified, of which 28 related to the use of AI in the care of people with dementia. The papers were divided into categories to show the utility and effectiveness these technologies may have: 1: to alert caregivers to problems 2: to facilitate activities for people with dementia 3: to help plan care for people with dementia 4: to consider the ethical implications of the use of artificial intelligence and machine learning
Despite a paucity of literature in the area, existing studies show potential, if used well, for technologies to be a useful addition to care of people with dementia. The experience of patients and their carers must be integral to their development and use.
Introduction: Adaptive Treatment Strategies warns therapists of patients at risk of treatment failure to prompt an adaption of the intervention. Internet-delivered Cognitive Behavioural Therapy (ICBT) collects a wide range of data before and during treatment and can quickly be adapted by adjusting the level of therapist support. Objectives: To evaluate how accurate machine learning algorithms can predict a single patient’s final outcome and evaluate the opportunities for using them within an Adaptive Treatment Strategy. Methods: Over 6000 patients at the Internet Psychiatry Clinic in Stockholm receiving ICBT for major depression, panic disorder or social anxiety disorder composed a training data set for eight different machine learning methods (e.g. k-Nearest Neighbour, random forest, and multilayer perceptrons). Symptom measures, messages between therapist and patient, homework reports, and other data from baseline to treatment week four was used to predict treatment success (either 50% reduction or under clinical cut-off) for each primary symptom outcome. Results: The Balanced Accuracy for predicting failure/success always were significantly better than chance, varied between 56% and 77% and outperformed the predictive precision in a previous Adaptive Treatment Strategy trial. Predictive power increased when data from treatment weeks was cumulatively added to baseline data. Conclusions: The machine learning algorithms outperformed a predictive algorithm previously used in a successful Adaptive Treatment Strategy, even though the latter also received input from a therapist. The next steps are to visualize what factors contributes most to a specific patient’s prediction and to enhance predictive power even further by so called Ensemble Learning.
The most common medical decision is the prescription of medicines. More than 130 different drugs with proven efficacy are currently available for the treatment of patients with mental disorders.
The aim was to use routine data available at a patient’s admission to the hospital to predict polypharmacy and drug-drug interactions (DDI).
The study used data obtained from a large clinical pharmacovigilance study sponsored by the Innovations Funds of the German Federal Joint Committee. It included all inpatient episodes admitted to eight psychiatric hospitals in Hesse, Germany, over two years. We used gradient boosting to predict respective outcomes. We tested the performance of our final models in unseen patients from another calendar year and separated the study sites used for training from the study sites used for performance testing.
A total of 53,909 episodes were included in the study. The models’ performance, as measured by the area under the ROC, was “excellent” (0.83) and “acceptable” (0.72) compared to common benchmarks for the prediction of polypharmacy and DDI, respectively. Both models were substantially better than a naive prediction based solely on basic diagnostic grouping.
This study has shown that polypharmacy and DDI at a psychiatric hospital can be predicted from routine data at patient admission. These predictions could support an efficient management of benefits and risks of hospital prescriptions, for instance by including pharmaceutical supervision early after admission for patients at risk before pharmacological treatment is established
This work was supported by the Innovations Funds of the German Federal Joint Committee (grant number: 01VSF16009). The funding body played no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscrip
Machine learning has increasingly been applied to classification of psychosis spectrum in neuroimaging research. However, a number of multimodal studies using MRI and electroencephalography (EEG) is quite limited.
To assess the power of multimodal structural MRI (sMRI) and EEG data to provide pairwise discrimination between first-episode schizophrenia (FES) patients, individuals at ultra-high-risk of psychosis (UHR), and healthy controls (HC) using machine learning algorithms.
46 FES male patients, 39 UHR individuals, and 54 matched HC underwent sMRI (3T Philips scanner) and electroencephalography. T1-weighted images were processed via FreeSurfer to obtain cortical and subcortical measures. L2 regularized logistic regression was used to evaluate the efficacy of diagnostic prediction.
The accuracies of pairwise discriminations were: 87% for FES vs HC (specificity 83%, sensitivity 91%); 77% for FES vs UHR (specificity 76%, sensitivity 79%); 75% for UHR vs HC (specificity 77%, sensitivity 73%).
Current findings suggest that the patterns of anatomical and functional variability have potential as biomarkers for discrimination between schizophrenia, UHR, and healthy subjects. Furthermore, results show that the selection and multimodality of feature types are important. Specifically, adding EEG data to morphometric measures improved accuracy rates in FES vs HC and FES vs UHR contrasts, whereas standalone EEG data provided higher accuracy compared with morphometric or multimodal data in UHR vs HC discrimination. Expectedly, predictive power for the UHR was smaller than for the FES due to its intermediate anatomical features, located between those observed in healthy controls and those found in patients. The work was supported by RFBR grant 20-013-00748
Schizophrenia is a chronic and severe mental disorder. While research focus remains mainly on negative outcomes, it is questionable whether we are placing enough emphasis on improving their sense of well-being and functioning. This could be accessed through the study of the quality of life (QoL). To date, QoL prediction models mainly focused on neurocognition and psychotic symptoms, but their predictive power remained limited.
The aim is to accurately predict the QoL within schizophrenia using unsupervised learning methods.
We computed variables from 952 patients from the CATIE study, a randomized, double-blind clinical trial for schizophrenia treatment. QoL was measured using the Heinrichs-Carpenter Quality of Life Scale and potential predictors included almost all available variables: symptoms, neurocognition, medication adherence, insight, adverse effects, etc. By optimizing parameters to reach optimal models, three linear regressions were calculated: (1) baseline predictors of 12-month QoL, (2) 6-month predictors of 12-month QoL, and (3) baseline predictors of 6-month QoL. Adjustments were made to ensure that included variables were not collinear nor redundant with QoL.
Calculated models had adjusted R-squared of 0.918, 0.922 and 0.913, respectively. Best predictors were medication side effects, sociodemographic and neurocognitive variables. Low psychotic and depressive symptoms were also included, as well as lab values suggesting the absence of problems with chloremia and calcemia.
Calculated predictive models explain almost all subsequent QoL. It appears that physical health variables, generally omitted from mental health-related studies, have an important impact on patients’ QoL. Therefore, interventions should also consider these aspects.
Social Media might represent an amazing and valuable source of information on mental health and well-being. Several researches revealed that adolescents aged 13 to 17 years old go “online” daily or stay online “almost constantly”.
The aim of this project is to identify distress in pre-clinical stages using Social media screening methods. The system can be modelled to centre on different several health-related topics.
We created a digital system able to analyse scripts written by adolescents on Twitter. InsideOut works using machine learning techniques and computational linguistic items to catch significant and sense of written messages and it improves its performances with iterations. The system is able to automatically identify semantic information relevant to different topics: in this case “distress in teenagers”.
The task of our system is considered correct when it is able to identify triples of Life Event, Sentiment and Experience of a tweet in agreement with the Gold Standard established among the annotators. The system has around 70% of accuracy in identifying triples. The analysis has been carried out both in Italian and English collecting over 4 million Italian tweets and 30 million English tweets. Comparative analysis with self-report questionnaires show that tweet analysis is able to suggest similar statistics.
This study analyzed contents of messages posted on Social Media Twitter meta-dating them with psychological and health-related information. Using InsideOut, we can plan clinical intervention in district and regions where high levels of uneasiness are revealed.
Different electrophysiological indices have been investigated to identify diagnostic and prognostic markers of schizophrenia (SCZ). However, these indices have limited use in clinical practice, since both specificity and association with illness outcome remain unclear. In recent years, machine learning techniques, through the combination of multidimensional data, have been used to better characterize SCZ and to predict illness course.
The aim of the present study is to identify multimodal electrophysiological biomarkers that could be used in clinical practice in order to improve precision in diagnosis and prognosis of SCZ.
Illness-related and functioning-related variables were measured at baseline in 113 subjects with SCZ and 57 healthy controls (HC), and after four-year follow-up in 61 SCZ. EEGs were recorded at baseline in resting-state condition and during two auditory tasks (MMN-P3a and N100-P3b). Through a Linear Support Vector Machine, using EEG data as predictors, four models were generated in order to classify SCZ and HC. Then, we combined unimodal classifiers’ scores through a stacking procedure. Pearson’s correlations between classifiers score with illness-related and functioning-related variables, at baseline and follow-up, were performed.
Each EEG model produced significant classification (p < 0.05). Global classifier discriminated SCZ from HC with accuracy of 75.4% (p < 0.01). A significant correlation (r=0.40, p=0.002) between the global classifier scores with negative symptoms at follow-up was found. Within negative symptoms, blunted affect showed the strongest correlation.
Abnormalities in electrophysiological indices might be considered trait markers of schizophrenia. Our results suggest that multimodal electrophysiological markers might have prognostic value for negative symptoms.
Major depressive disorder (MDD) is complex and multifactorial, posing a major challenge of tailoring the optimal medication for each patient. Current practice for MDD treatment mainly relies on trial-and-error, with estimated 42%-53% response rates for antidepressant use.
We sought to generate an accurate predictor of response to a panel of antidepressants and optimize treatment selection using a data-driven approach analyzing combinations of clinical and demographic factors.
We analyzed the response patterns of patients to five antidepressant medications in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study and the Pharmacogenomic Research Network Antidepressant Medication Pharmacogenomic Study (PGRN-AMPS), and employed state-of-the-art machine learning (ML) tools to generate a predictive algorithm. To validate our results and confirm the algorithm’s external generalizability outside of its training groups, we assessed its capacity to predict individualized antidepressant responses on a separate validation and test sets consisting of 1,021 patients overall from both studies.
The algorithm’s ML prediction models achieved an average accuracy of 0.6416 (64.16%, SD 4.4) across the analyzed medications, and a cumulative accuracy of 0.6012 (60.12%), AUC of 0.601, sensitivity of 0.6034 (60.34%) and specificity of 0.599 (59.9%).
These findings support applying ML to accumulating data derived from large studies to achieve a much-needed improvement in the treatment of depression. By an immediate analysis of large amount of combinatorial data at the point of care, such prediction models may support doctors’ prescription decisions, potentially allowing them to tailor the right antidepressant medication sooner.
Dekel Taliaz is the founder and CEO of Taliaz and reports stock ownership in Taliaz. Amit Spinrad and Sne Darki-Morag serve as data scientists in Taliaz.
Cost structure is the engineering of economics. Rather than approach economics as a study of the motivation toward maximum utility or profit, we approach it as a structural design problem for which cost structure is the starting point. The main distinction is between fixed and variable costs, a distinction first devised by ceramicist Josiah Wedgwood after a financial crash in the 1770s. We consider Mine Kafon, the wind-powered landmine removal device designed by Massoud Hassani (collected by the Museum of Modern Art), the willfully inefficient production of Lenka Clayton, who makes drawings on a typewriter, the intentionally market-allergic manufacture of Charlotte Posenenske, the advances of technology that change cost structure and artistic production (tube paints and machine learning), and the costs of the Impressionist painters, especially Camille Pissarro. We use the breakeven calculation and the income statement to organize costs and to diagnose the sustainability of operations.
The quality of life and lifespan are greatly reduced among individuals with mental illness. To improve prognosis, the nascent field of precision psychiatry aims to provide personalised predictions for the course of illness and response to treatment. Unfortunately, the results of precision psychiatry studies are rarely externally validated, almost never implemented in clinical practice, and tend to focus on a few selected outcomes. To overcome these challenges, we have established the PSYchiatric Clinical Outcome Prediction (PSYCOP) cohort, which will form the basis for extensive studies in the upcoming years.
PSYCOP is a retrospective cohort study that includes all patients with at least one contact with the psychiatric services of the Central Denmark Region in the period from January 1, 2011, to October 28, 2020 (n = 119 291). All data from the electronic health records (EHR) are included, spanning diagnoses, information on treatments, clinical notes, discharge summaries, laboratory tests, etc. Based on these data, machine learning methods will be used to make prediction models for a range of clinical outcomes, such as diagnostic shifts, treatment response, medical comorbidity, and premature mortality, with an explicit focus on clinical feasibility and implementation.
We expect that studies based on the PSYCOP cohort will advance the field of precision psychiatry through the use of state-of-the-art machine learning methods on a large and representative data set. Implementation of prediction models in clinical psychiatry will likely improve treatment and, hopefully, increase the quality of life and lifespan of those with mental illness.
This study aimed to identify health behaviours that determine adolescent’s adherence to the Mediterranean diet (MD) through a decision tree statistical approach.
Cross-sectional study, with data collected through a self-fulfilment questionnaire with five sections: (1) eating habits; (2) adherence to the MD (KIDMED index); (3) physical activity; (4) health habits and (5) socio-demographic characteristics. Anthropometric and blood pressure data were collected by a trained research team. The Automatic Chi-square Interaction Detection (CHAID) method was used to identify health behaviours that contribute to a better adherence to the MD.
Eight public secondary schools, in Algarve, Portugal.
Adolescents with ages between 15 and 19 years (n 325).
According to the KIDMED index, we found a low adherence to MD in 9·0 % of the participants, an intermediate adherence in 45·5 % and a high adherence in 45·5 %. Participants that regularly have breakfast, eat vegetable soup, have a second piece of fruit/d, eat fresh or cooked vegetables 1 or more times a day, eat oleaginous fruits at least 2 to 3 times a week, and practice sports and leisure physical activities outside school show higher adherence to the MD (P < 0·001).
The daily intake of two pieces of fruit and vegetables proved to be a determinant health behaviour for high adherence to MD. Strategies to promote the intake of these foods among adolescents must be developed and implemented.
The answers that each political community finds to the law reform questions posed by AI may differ, but a near-term threat is that AI systems capable of causing harm will not be confined to one jurisdiction – indeed, it may be impossible to link them to a specific jurisdiction at all. This is not a new problem in cybersecurity, but different national approaches to regulation will pose barriers to effective regulation exacerbated by the speed, autonomy, and opacity of AI systems. For that reason, some measure of collective action is needed. Lessons may be learned from efforts to regulate the global commons as well as moves to outlaw certain products (weapons and drugs, for example) and activities (such as slavery and child sex tourism). The argument advanced here is that regulation, in the sense of public control, requires active involvement of states. To co-ordinate those activities and enforce global ‘red lines’, this chapter posits a hypothetical International Artificial Intelligence Agency (IAIA), modelled on the agency created after the Second World War to promote peaceful uses of nuclear energy, while deterring or containing its weaponization and other harmful effects.
The increasing autonomy of AI systems is exposing gaps in regulatory regimes that assume the centrality of human actors. Yet surprisingly little attention is given to what is meant by ‘autonomy’ and its relationship to those gaps. Driverless vehicles and autonomous weapon systems are the most widely studied examples, but related issues arise in algorithms that allocate resources or determine eligibility for programmes in the private or public sector. This chapter develops a novel typology that distinguishes three lenses through which to view the regulatory issues raised by autonomy: the practical difficulties of managing risk associated with new technologies, the morality of certain functions being undertaken by machines at all, and the legitimacy gap that is created when public authorities delegate their powers to algorithms.
The rule of law is the epitome of anthropocentrism: humans are the primary subject and object of norms that are created, interpreted, and enforced by humans – made manifest in government of the people, by the people, for the people. Though legal constructs such as corporations may have rights and obligations, these in turn are traceable back to human agency in their acts of creation, their daily conduct overseen to varying degrees by human agents. Even international law, which governs relations among states, begins its foundational text with the words ‘We the peoples…’. The emergence of fast, autonomous, and opaque AI systems forces us to question this assumption of our own centrality, though it is not yet time to relinquish it.
As AI systems operate with greater autonomy, the idea that they might themselves be held responsible has gained credence. On its face, the idea of giving those systems a form of independent legal personality may seem attractive. Yet this chapter argues that this is both too simple and too complex. It is simplistic in that it lumps a wide range of technologies together in a single, ill-suited legal category; it is overly complex in that it implicitly or explicitly embraces the anthropomorphic fallacy that AI systems will eventually assume full legal personality in the manner of the ‘robot consciousness’ arguments mentioned earlier in the book. Though the emergence of general AI is a conceivable future scenario – and one worth taking precautions against – it is not a sound basis for regulation today.