Computational Psychiatry for Computers

Abstract

Computational psychiatry is a nascent field that attempts to use multi-level analyses of the underlying computational problems that we face in navigating a complex, uncertain and changing world to illuminate mental dysfunction and disease. Two particular foci of the field are the costs and benefits of environmental adaptivity and the danger and necessity of heuristics. Here, we examine the extent to which these foci and others can be used to study the actual and potential flaws of the artificial computational devices that we are increasingly inventing and empowering to navigate this very same environment on our behalf.

(Mal)adaptive Mentalising in the Cognitive Hierarchy, and its link to paranoia

“…Our work highlights how maladaptive DoM are functions of the agent's own beliefs, its environment and the beliefs of other agents. This is consistent with prior observations (46; 47; 48), and is relevant for the maladaptive behaviour of machines (49). It also shows how complex phenomena like scepticism can arise even from optimal Bayesian inference (16; 50) and how what an agent might think of as optimal Bayesian inference can go awry given the confusion about the decision problem or an unfortunate environment (48) The over-attribution of negative social intentions is a central feature in paranoid delusions and borderline personality disorder (32) and hyper-mentalising has been identified as an important transdiagnostic feature in psychopathology more broadly (19; 26).…”

Between prudence and paranoia: Theory of Mind gone right, and wrong

“…Our work highlights how ToM-induced paranoia and its detrimental consequences are not only a function of the agent's own ToM but also of its environment and other agents. This is consistent with venerable observations (Simon, 1990;Bhui et al, 2021;Huys et al, 2015), and is relevant for the maladaptive behavior of machines (Schulz & Dayan, 2020). It also shows how complex phenomena like skepticism can arise even from optimal Bayesian inference (Bhui & Gershman, 2020;Alon et al, 2022) and how what an agent might think of as optimal Bayesian inference can go awry given confusion about the decision problem or an unfortunate environment (Huys et al, 2015) Our work has particular relevance for computational psy-chiatry: Overly vigilant behavior is hypothesised as a generative factor in psychiatric symptoms, such as paranoia or anxiety (McLaren et al, 2022;Sharp et al, 2011).…”

Using cognitive psychology to understand GPT-3

“…Methods from cognitive psychology have also previously been applied to understand other deep learning models' behavior 54 . Therefore, our current work can be seen as part of a larger scientific movement where methods from psychology are becoming increasingly more important to understand capable black-box algorithms' learning and decision-making processes [55][56][57][58] .…”