Predicting internalizing symptoms with machine learning: identifying individuals that need care.

Objective The current project aims to identify individuals in urgent need of mental health care, using a machine learning algorithm (random forest). Comparison/contrast with conventional regression analyses is discussed. Participants: A total of 2,409 participants were recruited from an anonymous university, including undergraduate and graduate students, faculty, and staff. Methods: Answers to a COVID-19 impact survey, the Patient Health Questionnaire-9 (PHQ-9), and the Generalized Anxiety Disorder-7 (GAD-7) were collected. The total scores of PHQ-9 and GAD-7 were regressed on six composites that were created from the questionnaire items, based on their topics. A random forest was trained and validated. Results: Results indicate that the random forest model was able to make accurate, prospective predictions (R2 = .429 on average) and we review variables that were deemed predictively relevant. Conclusions: Overall, the study suggests that predictive models can be clinically useful in identifying individuals with internalizing symptoms based on daily life disruption experiences.

PyMC: a modern, and comprehensive probabilistic programming framework in Python.

PyMC is a probabilistic programming library for Python that provides tools for constructing and fitting Bayesian models. It offers an intuitive, readable syntax that is close to the natural syntax statisticians use to describe models. PyMC leverages the symbolic computation library PyTensor, allowing it to be compiled into a variety of computational backends, such as C, JAX, and Numba, which in turn offer access to different computational architectures including CPU, GPU, and TPU. Being a general modeling framework, PyMC supports a variety of models including generalized hierarchical linear regression and classification, time series, ordinary differential equations (ODEs), and non-parametric models such as Gaussian processes (GPs). We demonstrate PyMC's versatility and ease of use with examples spanning a range of common statistical models. Additionally, we discuss the positive role of PyMC in the development of the open-source ecosystem for probabilistic programming.

Great minds think alike: New measures to quantify the similarity of recalls.

Given the recent interest in how memory operates in social contexts, it is more important than ever to meaningfully measure the similarity between recall sequences of different individuals. Similarity of recall sequences of different individuals has been quantified using primarily order-agnostic and some order-sensitive measures specific to memory research without agreement on any one preferred measure. However, edit distance measures have not been used to quantify the similarity of recall sequences in collaborative memory studies. In the current study, we review a broad range of similarity measures, highlighting commonalities and differences. Using simulations and behavioral data, we show that edit distances do measure a memory-relevant factor of similarity and capture information distinct from that captured by order-agnostic measures. We answer illustrative research questions which demonstrate potential applications of edit distances in collaborative and individual memory settings and reveal the unique impact collaboration has on similarity.

Facial Emotions Are Accurately Encoded in the Neural Signal of Those With Autism Spectrum Disorder: A Deep Learning Approach.

BACKGROUND: Individuals with autism spectrum disorder (ASD) exhibit frequent behavioral deficits in facial emotion recognition (FER). It remains unknown whether these deficits arise because facial emotion information is not encoded in their neural signal or because it is encodes but fails to translate to FER behavior (deployment). This distinction has functional implications, including constraining when differences in social information processing occur in ASD, and guiding interventions (i.e., developing prosthetic FER vs. reinforcing existing skills). METHODS: We utilized a discriminative and contemporary machine learning approach-deep convolutional neural networks-to classify facial emotions viewed by individuals with and without ASD (N = 88) from concurrently recorded electroencephalography signals. RESULTS: The convolutional neural network classified facial emotions with high accuracy for both ASD and non-ASD groups, even though individuals with ASD performed more poorly on the concurrent FER task. In fact, convolutional neural network accuracy was greater in the ASD group and was not related to behavioral performance. This pattern of results replicated across three independent participant samples. Moreover, feature importance analyses suggested that a late temporal window of neural activity (1000-1500 ms) may be uniquely important in facial emotion classification for individuals with ASD. CONCLUSIONS: Our results reveal for the first time that facial emotion information is encoded in the neural signal of individuals with (and without) ASD. Thus, observed difficulties in behavioral FER associated with ASD likely arise from difficulties in decoding or deployment of facial emotion information within the neural signal. Interventions should focus on capitalizing on this intact encoding rather than promoting compensation or FER prostheses.

Persistence of false memories and emergence of collective false memory: collaborative recall of DRM word lists.

Information and misinformation are proliferating on social media. A rapid rise in the use of these platforms makes it important to identify psychological mechanisms that underlie the production, propagation, and convergence of false memories in groups. Websites and social media platforms vary in the extent of restrictions placed on interactive communication (e.g., group chats, threaded or disabled comments, direct messaging), prompting questions about the impact of different interaction styles on false memory production. We tested this question in a laboratory analog of interaction styles and compared two well-known procedures of collaboration, free-for-all and turn-taking. To expose participants to information known to promote recall of both true and false information, we used the Deese-Roediger-McDermott (DRM) word lists (Roediger & McDermott, 1995). Participants recalled these words using free-for-all collaboration, turn-taking collaboration, or individually. Next, all participants individually recalled the studied items. Turn-taking produced more false memories in group recall than did free-for-all collaboration, replicating past findings. Novel findings showed that former group members exhibited social contagion following both interaction styles, where they produced more false information in later individual recall and exhibited collective false memories. We discuss the implications for the emergence and convergence of true and false memories among users online.

Online Survey of the Impact of COVID-19 Risk and Cost Estimates on Worry and Health Behavior Compliance in Young Adults.

The novel coronavirus COVID-19 pandemic is associated with elevated rates of anxiety and relatively lower compliance with public health guidelines in younger adults. To develop strategies for reducing anxiety and increasing adherence with health guidelines, it is important to understand the factors that contribute to anxiety and health compliance in the context of COVID-19. Earlier research has shown that greater perceived risk of negative events and their costs are associated with increased anxiety and compliance with health behaviors, but it is unclear what role they play in a novel pandemic surrounded by uncertainty. In the present study we measured (1) perceived risk as the self-reported probability of being infected and experiencing serious symptoms due to COVID-19 and (2) perceived cost as financial, real-world, physical, social, and emotional consequences of being infected with COVID-19. Worry was assessed using the Penn State Worry Questionnaire (PWSQ) and health compliance was measured as endorsement of the World Health Organization (WHO) health directives for COVID-19. Our results showed that greater perceived risk and costs of contracting the COVID-19 virus were associated with greater worry and while only costs were associated with greater compliance with health behaviors. Neither self-reported worry nor its interaction with cost estimates was associated with increased engagement in health behaviors. Our results provide important insight into decision making mechanisms involved in both increased anxiety and health compliance in COVID-19 and have implications for developing psychoeducational and psychotherapeutic strategies to target both domains.

The social contagion of temporal discounting in small social networks.

Decisions often require a tradeoff between immediate and long-term gratification. How individuals resolve such tradeoffs reflects constructs such as temporal discounting, the degree that individuals devalue delayed rewards. Recent research has started to focus on temporal decisions made in collaborative contexts (e.g., dyads, small groups). Results suggest that directly interacting with others leads to revisions in preferences, such that decision makers become more similar to their collaborative partners over time (e.g., more patient following collaboration with a patient other). What remains to be seen is whether this social influence extends to indirect social effects, such as when an individual influences another's preferences through a shared collaborative partner. In the current study, the focus was on decisions regarding hypothetical monetary rewards. Groups of three participated in a collaborative decision-making chain, in which network member X collaborated with member Y, who then subsequently collaborated with member Z. Though network members X and Z never directly interacted, a significant indirect link was observed between member X's pre-collaborative decision preferences and member Z's post-collaborative decision preferences. These results demonstrate that temporal decision preferences can be transmitted through intervening connections in a small social network (i.e., social contagion), showing that indirect social influence can be empirically observed and measured in controlled environments.

Pattern Recognition to Objectively Differentiate the Etiology of Cognitive Decline: Analysis of the Impact of Stroke and Alzheimer's Disease.

BACKGROUND: Undetected Alzheimer's disease (AD) and stroke neuropathology is believed to account for a large proportion of decline in cognitive performance that is attributed to normal aging. This study examined the amount of variance in age-related cognitive change that is accounted for by AD and stroke using a novel pattern recognition protocol. METHOD: Secondary analyses of data collected for the Health and Retirement Study (N = 17,579) were used to objectively characterize patterns of cognitive decline associated with AD and stroke. The rate of decline in episodic memory and orientation was the outcome of interest, while algorithms indicative of AD and stroke pathology were the predictors of interest. RESULTS: The average age of the sample was 67.54 ± 10.45 years at baseline, and they completed, on average, 14.20 ± 3.56 years of follow-up. After adjusting for demographics, AD and stroke accounted for approximately half of age-associated decline in cognition (51.07-55.6% for orientation and episodic memory, respectively) and explained variance attributed to random slopes in longitudinal multilevel models. DISCUSSION: The results of this study suggested that approximately half of the cognitive decline usually attributed to normal aging are more characteristic of AD and stroke.

Enhanced Error Decoding from Error-Related Potentials using Convolutional Neural Networks.

Error-related potentials are considered an important neuro-correlate for monitoring human intentionality in decision-making, human-human, or human-machine interaction scenarios. Multiple methods have been proposed in order to improve the recognition of human intentions. Moreover, current brain-computer interfaces are limited in the identification of human errors by manual tuning of parameters (e.g., feature/channel selection), thus selecting fronto-central channels as discriminative features within-subject. In this paper, we propose the inclusion of error-related potential activity as a generalized two-dimensional feature set and a Convolutional Neural Network for classification of EEG-based human error detection. We evaluate this pipeline using the BNCI2020 - Monitoring Error-Related Potential dataset obtaining a maximum error detection accuracy of 79.8% in a within-session 10-fold cross-validation modality, and outperforming current state of the art.

Implicit predictions of future rewards and their electrophysiological correlates.

Information that is motivationally relevant to an organism's survival demands preferential attention. Affective mechanisms facilitate attentional shifts and potentiate action to allow organisms to respond appropriately to motivationally relevant information. Previous work has demonstrated that the late-positive potential (LPP) is an event-related potential elicited by inherently emotional stimuli. For example, the LPP typically is evoked by images of weapons or erotica. The present study investigates stimuli that are not inherently emotional, but that implicitly (without participants' awareness) predict future monetary gains and losses. Results indicate that, relative to non-predictive cues, these predictive cues elicited frontally distributed positive potential. These results suggest that prediction of future rewards evokes neural responses that are similar to those evoked by inherently emotional stimuli. Results also indicate that monetary gains and losses elicit a frontally distributed LPP.

Are intertemporal preferences contagious? Evidence from collaborative decision making.

Prior research has provided substantial insight into individuals' intertemporal preferences (i.e., preferences about delayed rewards). In the present study, we instead investigated the preferences of small groups of individuals asked to express collective intertemporal decisions. The paradigm consisted of three phases. During the precollaboration and postcollaboration phases, participants completed an intertemporal decision task individually. During the collaboration phase, participants completed a similar task in small groups, reaching mutually-agreed-upon decisions. The results suggest that group preferences were systematically related to the mean of the group members' precollaboration preferences. In addition, collaborative decision making altered the group members' intertemporal preferences. Specifically, individuals' postcollaboration preferences converged toward the preferences of their respective groups. Furthermore, we found that individuals' postcollaboration preferences were independently related to both their precollaboration preferences and the preferences of the other group members, suggesting that individuals' postcollaboration preferences represented a revision of their precollaboration preferences based on the preferences observed in other group members. In Experiment 2, we demonstrated that similar patterns of results were found whether participants were making matching judgments or binary choices.

Time-Preference Tests Fail to Predict Behavior Related to Self-control.

According to theory, choices relating to patience and self-control in domains as varied as drug use and retirement saving are driven by generalized preferences about delayed rewards. Past research has shown that measurements of these time preferences are associated with these choices. Research has also attempted to examine how well such measurements can predict choices, but only with inappropriate analytical methods. Moreover, it is not clear which of the many kinds of time-preference tests that have been proposed are most useful for prediction, and a theoretically important aspect of time preferences, nonstationarity, has been neglected in measurement. In Study 1, we examined three approaches to measuring time preferences with 181 users of Mechanical Turk. Retest reliability, for both immediate and 1-month intervals, was decent, as was convergent validity between tests, and association was similar to previous results, but predictive accuracy for 10 criterion variables (e.g., tobacco use) was approximately nil. In Study 2, we examined one other approach to measuring time preferences, and 40 criterion variables, using 7,127 participants in the National Longitudinal Survey of Youth 1979. Time preferences were significantly related to criterion variables, but predictive accuracy was again poor. Our findings imply serious problems for using time-preference tests to predict real-world decisions. The results of Study 1 further suggest there is little value in measuring nonstationarity separately from patience.

Memory Transmission in Small Groups and Large Networks: An Agent-Based Model.

The spread of social influence in large social networks has long been an interest of social scientists. In the domain of memory, collaborative memory experiments have illuminated cognitive mechanisms that allow information to be transmitted between interacting individuals, but these experiments have focused on small-scale social contexts. In the current study, we took a computational approach, circumventing the practical constraints of laboratory paradigms and providing novel results at scales unreachable by laboratory methodologies. Our model embodied theoretical knowledge derived from small-group experiments and replicated foundational results regarding collaborative inhibition and memory convergence in small groups. Ultimately, we investigated large-scale, realistic social networks and found that agents are influenced by the agents with which they interact, but we also found that agents are influenced by nonneighbors (i.e., the neighbors of their neighbors). The similarity between these results and the reports of behavioral transmission in large networks offers a major theoretical insight by linking behavioral transmission to the spread of information.

Direct and Indirect Influence of Altruistic Behavior in a Social Network.

Prior research has suggested that recipients of generosity behave more generously themselves (a direct social influence). In contrast, there is conflicting evidence about the existence of indirect influence (i.e., whether interacting with a recipient of generosity causes one to behave more generously), casting doubt on the possibility that altruistic behavior can cascade through social networks. The current study investigated how far selfish and generous behavior can be transmitted through social networks and the cognitive mechanisms that underlie such transmission. Participants played a sequence of public goods games comprising a chain network. This network is advantageous because it permits only a single, unambiguous path of influence. Furthermore, we experimentally manipulated the behavior of the first link in the chain to be either generous or selfish. Results revealed the presence of direct social influence, but no evidence for indirect influence. Results also showed that selfish behavior exerted a substantially greater influence than generous behavior. Finally, expectations about future partners' behavior strongly mediated the observed social influence, suggesting an adaptive basis for such influence.

Reputational concerns, not altruism, motivate restraint when gambling with other people's money.

People may behave prosocially not only because they value the welfare of others, but also to protect their own reputation. We examined the separate roles of altruism and reputational concerns in moral-hazard gambling tasks, which allowed subjects to gamble with a partner's money. In Study 1, subjects who were told that their partner would see their choices were more prosocial. In Study 2, subjects were more prosocial to a single partner when their choices were transparent than when their choices were attributed to a third party. We conclude that reputational concerns are a key restraint on selfish exploitation under moral hazard.

Fairness influences early signatures of reward-related neural processing.

Many humans exhibit a strong preference for fairness during decision-making. Although there is evidence that social factors influence reward-related and affective neural processing, it is unclear if this effect is mediated by compulsory outcome evaluation processes or results from slower deliberate cognition. Here we show that the feedback-related negativity (FRN) and late positive potential (LPP), two signatures of early hedonic processing, are modulated by the fairness of rewards during a passive rating task. We find that unfair payouts elicit larger FRNs than fair payouts, whereas fair payouts elicit larger LPPs than unfair payouts. This is true both in the time-domain, where the FRN and LPP are related, and in the time-frequency domain, where the two signals are largely independent. Ultimately, this work demonstrates that fairness affects the early stages of reward and affective processing, suggesting a common biological mechanism for social and personal reward evaluation.

The predictive accuracy of intertemporal-choice models.

How do people choose between a smaller reward available sooner and a larger reward available later? Past research has evaluated models of intertemporal choice by measuring goodness of fit or identifying which decision-making anomalies they can accommodate. An alternative criterion for model quality, which is partly antithetical to these standard criteria, is predictive accuracy. We used cross-validation to examine how well 10 models of intertemporal choice could predict behaviour in a 100-trial binary-decision task. Many models achieved the apparent ceiling of 85% accuracy, even with smaller training sets. When noise was added to the training set, however, a simple logistic-regression model we call the difference model performed particularly well. In many situations, between-model differences in predictive accuracy may be small, contrary to long-standing controversy over the modelling question in research on intertemporal choice, but the simplicity and robustness of the difference model recommend it to future use.

Evidence for online processing during causal learning.

Many models of learning describe both the end product of learning (e.g., causal judgments) and the cognitive mechanisms that unfold on a trial-by-trial basis. However, the methods employed in the literature typically provide only indirect evidence about the unfolding cognitive processes. Here, we utilized a simultaneous secondary task to measure cognitive processing during a straightforward causal-learning task. The results from three experiments demonstrated that covariation information is not subject to uniform cognitive processing. Instead, we observed systematic variation in the processing dedicated to individual pieces of covariation information. In particular, observations that are inconsistent with previously presented covariation information appear to elicit greater cognitive processing than do observations that are consistent with previously presented covariation information. In addition, the degree of cognitive processing appears to be driven by learning per se, rather than by nonlearning processes such as memory and attention. Overall, these findings suggest that monitoring learning processes at a finer level may provide useful psychological insights into the nature of learning.

Adaptive intertemporal preferences in foraging-style environments.

Decision makers often face choices between smaller more immediate rewards and larger more delayed rewards. For example, when foraging for food, animals must choose between actions that have varying costs (e.g., effort, duration, energy expenditure) and varying benefits (e.g., amount of food intake). The combination of these costs and benefits determine what optimal behavior is. In the present study, we employ a foraging-style task to study how humans make reward-based choices in response to the real-time constraints of a dynamic environment. On each trial participants were presented with two rewards that differed in magnitude and in the delay until their receipt. Because the experiment was of a fixed duration, maximizing earnings required decision makers to determine how to trade off the magnitude and the delay associated with the two rewards on each trial. To evaluate the extent to which participants could adapt to the decision environment, specific task characteristics were manipulated, including reward magnitudes (Experiment 1) and the delay between trials (Experiment 2). Each of these manipulations was designed to alter the pattern of choices made by an optimal decision maker. Several findings are of note. First, different choice strategies were observed with the manipulated environmental constraints. Second, despite contextually-appropriate shifts in behavior between conditions in each experiment, choice patterns deviated from theoretical optimality. In particular, the delays associated with the rewards did not exert a consistent influence on choices as required by exponential discounting. Third, decision makers nevertheless performed surprisingly well in all task environments with any deviations from strict optimality not having particularly deleterious effects on earnings. Taken together, these results suggest that human decision makers are capable of exhibiting intertemporal preferences that reflect a variety of environmental constraints.

Discounting of delayed rewards is not hyperbolic.

Delay discounting refers to decision-makers' tendency to value immediately available goods more than identical goods available only after some delay. In violation of standard economic theory, decision-makers frequently exhibit dynamic inconsistency; their preferences change simply due to the passage of time. The standard explanation for this behavior has appealed to the nature of decision-makers' discount functions, specifically positing a hyperbolic discount function. Though this explanation has been largely accepted, there has been surprisingly little work examining whether preference reversals are actually consistent with hyperbolic discounting. The current study holds hyperbolic discounting to the same empirical standard that exponential discounting has been held to and finds that choice behavior is not consistent with hyperbolic discounting. Despite the overwhelming focus placed on hyperbolic discounting, the current findings cast doubt on hyperbolic discounting as an explanation of decision-makers' undesirable preference reversals and as an explanation of delay discounting behavior in general.