Arai Kensuke, Jacob Suma, Widge Alik S, Yousefi Ali
Computer Science Department, Worcester Polytechnic Institute, Worcester, MA, 01609, USA.
Semel Institute, Child & Adolescent Psychiatry, University of California, Los Angeles, CA, 90210, USA.
Sci Rep. 2025 Apr 29;15(1):14955. doi: 10.1038/s41598-025-95444-6.
Current psychiatric nosology is based on observed and self-reported symptoms. Heterogenous pathophysiological mechanisms may underlie similar symptoms leading to diagnosis not matching up to the neurobiology. Recent research has sought to move away from diagnoses by symptoms, to viewing aberrant mental health in terms of abnormal human neurobehavioral functioning and concurrent deviations in the pathophysiology. Human behavior in a social context is a core neurobehavioral function with large individual variation that may reflect genomic, metabolic or neurobiological variation, whose identification potentially yields more accurate targeting for the development of interventions and biomedical treatments. In this research, we describe an experimental framework that utilizes a zero-sum game of repeated Rock-Paper-Scissors played against an artificial intelligence agent as an assay of social interaction. Human deviation from the Nash Mixed Equilibrium strategy of play, the only guaranteed way to avoid exploitation, can be seen in the sequential dependence of hands. We hypothesize that this deviation represents humans mimicing randomness to avoid exploitation through constant adjustments of behavior, which we analyze in terms of a set of switching heuristic lag-1 conditional response rules. We quantify and interpret the set of rules subjects are able to utilize as mirroring individual traits. Subjects in the study also completed the Autism Quotient Abridged survey, and subscores of the social, imagination and routine factors were found to be predicted by a combination of behavioral features derived from game play.
当前的精神疾病分类学基于观察到的症状和自我报告的症状。不同的病理生理机制可能是相似症状的基础，导致诊断与神经生物学不匹配。最近的研究试图从基于症状的诊断转向从人类异常神经行为功能和病理生理学的并发偏差角度看待异常心理健康。在社会背景下的人类行为是一种核心神经行为功能，个体差异很大，可能反映基因组、代谢或神经生物学差异，识别这些差异可能为干预措施和生物医学治疗的开发提供更准确的靶点。在这项研究中，我们描述了一个实验框架，该框架利用与人工智能代理进行的重复剪刀石头布零和游戏作为社会互动的一种测定方法。人类偏离纳什混合均衡策略（这是避免被利用的唯一保证方式）可以从出手的顺序依赖性中看出。我们假设这种偏差代表人类通过不断调整行为来模仿随机性以避免被利用，我们根据一组切换启发式滞后1条件响应规则对其进行分析。我们量化并解释受试者能够利用的一组规则，将其视为反映个体特征。该研究中的受试者还完成了自闭症商数简版调查，发现社交、想象力和常规因素的子分数可由游戏行为衍生的行为特征组合预测。
