Author: Edwin Clayton

Jennifer Siegel
Zuckerman Institute
Columbia University

Borderline Personality Disorder (BPD) is a serious mental disorder characterized by marked interpersonal disturbances, including difficulties trusting others and volatile impressions of others’ moral character, often resulting in premature relationship termination. The ability to build and maintain healthy relationships depends on our ability to use social information to build accurate representations of others and their mental states through social learning. One aspect of social learning that is especially relevant to forming and maintaining relationships is inferring others’ moral character, i.e., whether they are helpful and trustworthy, or harmful and untrustworthy. We introduce a novel computational assay of moral inference to investigate how patients with BPD form beliefs about the moral character of others and incorporate new information into existing beliefs. We find that the computational mechanisms of moral inference differed for untreated BPD patients relative to matched non-BPD control participants and BPD patients treated in a Democratic Therapeutic Community (DTC). In untreated BPD patients, beliefs about harmful agents were more certain and less amenable to updating relative to both non-BPD control participants and DTC-treated participants.The findings suggest that DTC may help the maintenance of social relationships in BPD by increasing patients’ openness to learning about adverse interaction partners. The results provide mechanistic insights into social deficits in BPD and demonstrate the potential for combining objective behavioral paradigms with computational modelling as a tool for assessing BPD pathology and treatment outcomes.

Frederike Petzschner
Institut für Biomedizinische Technik
Universität Zürich und ETH Zürich

While OCD and gambling are often conceptualized as disorders of compulsivity arising from dysfunctional beliefs, the nature of the beliefs leading to symptom manifestation remains unclear. Computational modeling can help disentangle the complex interplay between beliefs and symptoms, identify core components, and suggest targets for novel treatment approaches.

This talk covers three lines of work that center around a computational informed understanding of the origins of compulsive behavior: The first part relates individual structure learning differences to obsessive-compulsive traits and SSRIs – the first-line pharmacological treatment of OCD. The second part delves deeper into the origin of compulsive behavior. While there are a number of theoretical explanatory frameworks mostly centered around belief alterations, there is no clear consensus on what beliefs are necessary or sufficient to cause compulsive behavior.
We used a minimal model approach (based on a POMDP) to address this question and pinpoint to the belief about the success of preventive actions as being centrally important for eliciting and maintaining compulsive behavior. Finally, the third part of the talk highlights that similar beliefs about the efficacy of one’s actions may not only be associated with compulsive behavior in OCD but may also play an essential role in the compulsive aspect of pathological gambling.

View this recorded session here.

Michael J Frank
Carney Center for Computational Brain Science
Brown University

Computational modeling can formally adjudicate between theories and affords quantitative fits to behavioral/brain data. This talk has two parts: in the first I describe an example of how simplified models of brain function can be useful for linking neural circuits to decision making functions. The models provide a mechanistic interpretation of changes in decision making in patient populations. However, for quantitative fitting purposes, the space of plausible generative models considered is dramatically limited by the set of models with known likelihood functions. For many models, the lack of a closed-form likelihood typically impedes Bayesian inference methods. As a result, standard models like the drift diffusion model are often fit to data for convenience even when other models might be superior. In the second part of the talk I will present a new method using artificial neural networks that learn approximate likelihoods for arbitrary generative models, allowing fast posterior sampling with only a one-off cost for model simulations that is amortized for future inference. We show that these methods can accurately recover posterior parameter distributions for a variety of neurocognitive process models. We provide code allowing users to deploy these methods for arbitrary hierarchical model instantiations linking brain mechanisms to behavior and for interrogating alterations in patient populations.

View this recorded session here.

David Zald
Center for Advanced Human Brain Imaging Research
Rutgers University

The emotional attentional blink (EAB) refers to a transient impairment in the ability to detect or discriminate a target when it is presented closely in time to an emotional distractor. The paradigm, which is also referred to in the literature as emotion induced blindness, has provided insight into the nature of bottom-up capture of attention by emotionally salient stimuli in both health and in psychopathology. Although the most dramatic aspect of the phenomena occurs when the target is not perceived at all, recent data indicate that the effects of the emotional distractor are more consistent with a graded impact on target processing than a purely “all or none” phenomena, with emotional distractors often slowing target detection and lowering the subjective vividness of target representations, rather than eliminating all subjective awareness of the target. Such data have implications for computational models of the competition of bottom-up and top-down attentional mechanisms.

Thorsten Kahnt
Department of Neurology
Northwestern University Feinberg School of Medicine

Research across species has shown that the orbitofrontal cortex (OFC) is important for decision making. However, it is less clear what specific computations are carried out in this region that make it so important for this function. Recent work from our lab and others has shown that OFC activity is correlated with expectations about specific outcomes. In this talk, I will present evidence from functional magnetic resonance imaging (fMRI) and transcranial magnetic stimulation (TMS) experiments suggesting that outcome expectations in OFC are required for decisions that are based on inferred or simulated outcomes, as opposed to behavior that can be based on direct experience alone. Because OFC is not directly accessible to TMS, we utilize network-targeted TMS and apply continuous theta burst stimulation (cTBS) to sites in lateral PFC that are individually selected to be functionally connected to the OFC. We show that OFC network-targeted cTBS selectively disrupts choices that require subjects to infer outcomes, without affecting choices that can be based on direct experiences alone. These findings suggest that the OFC contributes to decision making by representing associative relationships that can be used to simulate outcomes when direct experience is missing.

Lauren Atlas
Section on Affective Neuroscience and Pain
National Center for Complementary and Integrative Health

Pain is a fundamental experience that promotes survival. In humans, pain stands at the intersection of multiple health crises: chronic pain, the opioid epidemic, and health disparities. Addressing pain requires multidisciplinary approaches, integrating insights from basic neuroscience to psychology and psychiatry. In this talk, I will focus on the relationship between pain, affect, and learning. I will present new work addressing how instructions and learning shape pain, the role of the ventromedial prefrontal cortex in expectancy effects on pain, and comparing pain with appetitive and aversive taste to address processes that may be unique to pain versus those that may be involved in value-based learning more generally.

Gal Shoval
Sackler Faculty of Medicine
Tel Aviv University

Despite the growing use of antidepressants and the potential grave consequences of inadequate treatment, little is known about the impact of adherence to antidepressant treatment on mortality.
The objective of this line of research was to evaluate the association between adherence to antidepressants and all-cause mortality in a population-based cohort. Data were extracted from the electronic medical record database of the largest health provider in Israel (54% of the nation’s population) on a total of 251,745 patients aged 40 years and above who filled an antidepressant prescription at least once during a 4 year period. The main outcome measure was all-cause mortality during the study period. Adherence was measured as a continuous variable representing possession ratio (duration of filled antidepressant divided by duration of prescribed antidepressant). Proportional hazard Cox regression for multivariable survival analysis was used, adjusting for demographic and clinical variables that affect mortality. We tested our hypothesis in the general population as well as in selected patient populations with certain physical disorders (cardiovascular, stroke, cancer and cancer subtypes, parkinson’s disease). We also looked at the effect of age and gender on antidepressant adherence. The findings will be presented at the talk.

Mateo Leganes-Fonteneau
Center of Alcohol Studies
Rutgers University–New Brunswick

Research on implicit processes has revealed multiple shortcomings of awareness categorization tools, which are most often based on the failure to score significantly above chance level. Moreover, post-hoc awareness categorizations result in regression to the mean effects, by which aware participants are wrongly categorized as unaware. Using Bayes factors to obtain sensitive evidence for participants’ lack of knowledge could prevent regression to the mean effects. Here I present the reliability analysis of a novel Bayesian awareness categorization procedure, comparing it to traditional approaches based on t-tests.

Following a reward conditioning task, we divided participants in different groups (Aware, Unaware, Insensitive) depending on the results of individual Bayesian analysis ran on an awareness measure. We examined the cumulative awareness scores of each group, and found that Unaware participants did score below chance level, contrary to the predictions of regression to the mean effect. This was further confirmed using a resampling procedure with multiple iterations. Conversely, when categorizing participants using t-tests, Unaware participants scored significantly above chance level on the resampling procedure, showing regression to the mean effects.

The results show that using Bayes factors instead of t-tests as a categorization tool it is possible to avoid regression to the mean effects. The reliability of the Bayesian awareness categorization procedure strengthens previous evidence for implicit reward conditioning. I will discuss how Bayes factor can reduce measurement errors to provide a reliable post-hoc awareness categorization tool, improving the quality of research on implicit processes. The toolbox necessary to perform the categorization procedure is available to be incorporated in future experiments.

View the presentation HERE