Abstract
Introduction Anti-Social Personality Disorder (ASPD) and Borderline Personality Disorder (BPD) are
psychopathologies characterized by profoundly aberrant interpersonal behaviour. They strongly affect
the lives of both individual patients and those around them. BPDs and ASPDs share many behavioural
characteristics, such as impulsivity and difficulty in sustaining long-lasting social relationships, and
although they are considered different disorders, the distinction between them is sometimes blurry1. Amongst clinicians and scientists there is a growing interest in novel approaches to psychiatric disease
classification that allow for better understanding and distinction of psychiatric diseases, based on
potentially quantifiable traits. Here, we use a Social Hierarchy decision-making paradigm and apply
computational models to the obtained behaviour to understand better the differences between BPDs,
ASPDs and controls, and how these diseases impact social decision making. Methods As part of a largescale
computational psychiatry program, 159 BPDs, 12 ASPDs and 103 healthy controls (total n=274)
participated in a multi-round social exchange paradigm, the Social Hierarchy (SH) game. In each round of
the SH game one of the players (‘alpha’) receives a given sum of money and decides how to split it with
the other player (‘beta’). Then, the beta player has the chance to ‘challenge’ the alpha’s position. If
there is a challenge, both alpha and beta decide by how much to challenge (beta) or defend (alpha).
Whoever spends the most assumes the alpha position in the next round. The SH game is played for 30
rounds, always with the same opponent (here a computer algorithm mimicking a real person). Results
We found that ASPD participants spent significantly more on average on contests and left their
opponent with lower final earnings compared to both controls and BPDs. For the BPDs, model agnostic
approaches did not show them to differ significantly from healthy controls, but a logistic regression
model relating challenge decision to amount transferred showed BPDs had a higher baseline tendency
to challenge and were more sensitive to the amount of money transferred by the opponent. We also
started doing cross-validated model-based analyses of participants? behaviour and are obtaining
interesting results. Conclusions Our results suggest that ASPDs value social dominance more than
healthy controls and BPDs, whilst the latter are particularly sensitive to social signals. These results
suggest that ASPD and BPD may be two actual separate diagnoses, and offer promising avenues by
which different psychiatric disease states could be distinguished at the mechanistic level.
