Existing research has developed a working understanding of borderline personality disorder (BPD) patient traits and behav-ior in everyday life, but the subtleties of their cognitive processes during decision-making remains unclear. To understandhow reliance on previous experiences (priors) versus current sensory information (likelihoods) in the decision-makingprocess may differ for those with BPD in comparison to those within neuro-typical population, we implemented a coin-catching behavioral task with varying levels of prior and likelihood uncertainty. We hypothesized that, in accordance totypical BPD characteristics, BPD patients will rely significantly more on likelihood information even when likelihoodinformation is more unreliable than prior information. Analyzing the results using Bayesian statistics, we found evidencesuggesting that both the BPD patient group and the neuro-typical control group utilized prior and likelihood informa-tion similarly in decision-making. We theorize that BPD characteristics that are prominent in social interactions may notexactly replicate in non-social settings.

Substantial evidence has suggested that major depression is associated with a dysregulated dopamine system, which playsa pivotal role in decision-making under uncertainty. Previous research has proposed that dopamine enhances the weightgiven to current sensory information (sensory weight) versus prior beliefs, yet how much this relationship holds true indepression remains a topic under debate. To examine whether depression patients have decreased sensory weight dueto disturbed dopaminergic neurotransmissions, we used a visual coin-catching task in which uncertainty in both priorand sensory information varied. Decision-making strategies during the task were modeled by Bayesian statistics. Ourresults supported that depression patients preserved the ability to learn both prior and sensory information uncertainty,comparable to healthy controls. In contrast to our prediction, depression patients did not show decreased reliance onsensory information compared to controls, suggesting that depression does not induce a universal alteration in decision-making strategies under uncertainty. Our study provides empirical evidence that depression does not always show deficitsin uncertainty processing regardless of its correlation with dopamine dysregulations.

Many people have suffered harm as a result of hate speech on social media. The majority of research has focused on coarse-grained explicit hate speech detection while disregarding fine-grained implicit hate speech classification. It is crucial for more effectively combating hate speech. Although the language used in implicit hate speech may vary greatly, the mental states involved are usually the same. There are rarely similarities and differences between the mental states present in implicit hate speech examined. We create a module to infer mental states from implicit hate speech to close this gap. Mental states primarily refer to the speaker's intent and the reader's reaction. Then, we use them as the positive sample in contrastive learning. This strategy can pull the implicit hate speech which has similar mental states in similar representations and push away different ones. Comprehensive experiment results demonstrate superior classification performance and generalization of the proposed method.