Stahl, D. S. | Search bioRxiv

bioRxiv2018

Amazon Mechanical Turk as a platform for borderline personality disorder research

Researchers investigating the psychological processes underlying specific mental health problems often have difficulties achieving large enough samples for adequately powered studies. This can be particularly problematic when studying psychopathology with low base rates in typical samples (i.e., undergraduate and community). A relatively new approach to recruitment and testing employs online crowdsourcing to rapidly measure the characteristics and behavior of large numbers of people. We tested the feasibility of researching borderline personality disorder (BPD) in this manner using one large crowdsourcing site, Amazon Mechanical Turk (MTurk). Specifically, we examined prevalence rates of psychopathology in a large MTurk sample, as well as the demographic, psychosocial, and psychiatric characteristics of individuals who met criteria for BPD. These characteristics were compared across three groups: those who met criteria for BPD currently, those who met criteria for remitted BPD, and those who had never met criteria for BPD. The results suggest that MTurk may be ideally suited for studying individuals with a wide range of pathology, from healthy to intensely symptomatic to remitted.

animal behavior and cognition↗

bioRxiv2018

Differential valuation and learning from social and non-social cues in Borderline Personality Disorder

BackgroundVolatile interpersonal relationships are a core feature of Borderline Personality Disorder (BPD), and lead to devastating disruption of patients personal and professional lives. Quantitative models of social decision making and learning hold promise for defining the underlying mechanisms of this problem. In this study, we tested BPD and control subject weighting of social versus non-social information, and their learning about choices under stable and volatile conditions. We compared behavior using quantitative models.\n\nMethodsSubjects (n=20 BPD, n=23 control) played an extended reward learning task with a partner (confederate) that requires learning about non-social and social cue reward probability (The Social Valuation Task). Task experience was measured using language metrics: explicit emotions/beliefs, talk about the confederate, and implicit distress (using the previously established marker self-referentiality). Subjects weighting of social and non-social cues was tested in mixed-effects regression models. Subjects learning rates under stable and volatile conditions were modelled (Rescorla-Wagner approach) and group x condition interactions tested.\n\nResultsCompared to controls, BPD subject debriefings included more mentions of the confederate and less distress language. BPD subjects also weighted social cues more heavily, but had blunted learning responses to (non-social and social) volatility.\n\nConclusionsThis is the first report of patient behavior in the Social Valuation Task. The results suggest that BPD subjects expect higher volatility than do controls. These findings lay the groundwork for a neuro-computational dissection of social and non-social belief updating in BPD, which holds promise for the development of novel clinical interventions that more directly target pathophysiology.

neuroscience↗