17 March 2021: Computational Approaches to Understanding Suicide

Alex Millner
Department of Psychology
Harvard University

This presentation will be divided into two fairly distinct parts. First, I’ll discuss Pavlovian biases in the context of aversive states. Traditionally, aversive Pavlovian biases are associated with inhibition; however, this had only been tested in the context of avoiding punishment. Using a novel behavioral task and computational model, we show that the effect of Pavlovian control depends on the proximity of the aversive state: when escaping an ongoing aversive state, there is a Pavlovian bias for vigorous, active responses whereas when avoiding a potential aversive state, Pavlovian control favors inhibition. Escape-related Pavlovian biases have relevance for many psychiatric disorders. For example, decades of theories and clinical accounts have argued that suicidal thoughts and behaviors are mostly driven by a desire to escape aversive internal states and we show that people with a history of suicidal thoughts and behaviors show an increased Pavlovian bias for escape. In the second part, I will critique current approaches to computational psychiatry (such as my study discussed in the first part) and offer a complementary approach that includes developing formal theories of clinical states, such as suicidal thoughts and behaviors. I will present some very preliminary work in this area, with a focus on outlining the advantages and challenges of this novel approach.

View this recorded session here.