Computational Modelling of Behavioural Differences in PTSD

My project focuses on modelling behavioural differences in PTSD, and potentially Complex PTSD (cPTSD), by implementing Bayesian & predictive coding approaches and reinforcement learning. Typically these approaches are combined with behavioural data in order to identify learning (and decision-making) differences between individuals and groups [1][2][3].

I intend to use datasets collected by the Seriès lab, as well as my own behavioural task, to investigate differences in the predictive mechanisms of individuals with PTSD. I.e. is trauma/PTSD symptomatology related to differences in prediction? If so, how are these differences represented within computational models and how are they realised in behaviour?

Computational modelling has been applied to a variety of mental illnesses. Some (e.g. depression) [4] have been studied more than others (e.g. PTSD).

As such, there are a plethora of questions surrounding PTSD, and the hidden mechanisms behind its aetiology and progression are still unknown [5]. By testing falsifiable hypotheses on differences in those with PTSD, we can investigate how the disorder functions, e.g. if individuals diagnosed with PTSD learn differently from aversive outcomes than control individuals.

This approach will uncover mechanisms of the disorder that may be crucial to its formation and progression, allowing us to make inferences on how these differences may impact disorder risk, progression, and treatments. They also lead to a big question - are these differences a cause or a consequence of the disorder?

I have several objectives/milestones: 1) a systematic literature review of computational modelling of PTSD (no such reviews currently exist), 2) analysis of a Seriès lab anxiety dataset from a PTSD perspective, and 3) conception of main project idea.

1) The majority of related literature applies a range of methods implemented with varying modelling quality. A systematic literature review would be incredibly valuable in assessing what previous methods have found and may be used to create guidelines to improve future research in PTSD. This review will include comparisons of models, behavioural tasks, analysis techniques and results across studies.

2) Similar computational approaches have recently been performed on anxiety in the Seriès lab. A behavioural task was carried out on a general population that included a variety of self-report questionnaire scores (including PCL-5, the checklist for PTSD). Analysing this from a PTSD perspective may yield interesting results. The results from this will be useful in the design of my own behavioural task, as well as in defining possible hypotheses.

3) My main project will investigate the metacognitive aspect of the disorder, focusing on self-esteem and self-confidence. Many approaches model the disorder in terms of differences in learning and relate findings to essential features of avoidance, hyperarousal, hypervigilance, and persistent perceptions of heightened current threat [6][7][8][9]. Few studies have investigated other features of negative alterations in mood/cognition, severe problems in affect regulation, and persistent diminished beliefs about oneself - linking closely to self-esteem and self-confidence.

These features have recently been included as criteria for the separate disorder, Complex PTSD [10]. Investigations into the mechanisms underlying these features could provide us with more rigid definitions of PTSD and cPTSD, and may uncover intriguing differences/similarities between the two related disorders. Some studies have investigated the effect of self-esteem in PTSD [11][12][13], but none appear to have implemented the computational modelling approach proposed here.

With studies of metacognition being popular in recent years [14][15], and a theoretical metacognition model of PTSD being introduced in 2004 [16], this project will be make a valuable contribution to research in understanding metacognition in PTSD.