Testing computational models of learning from social, real, and fictive feedback in human and nonhuman primates

该联盟将开发学习和决策的计算模型，并在人类和猕猴两个生物系统中进行测试。这些模型应考虑到使用不同的行动结果信息来源来指导未来的行为：关于实际结果的反馈，关于如果选择不同的行动就会获得的结果的虚构反馈，以及关于在社会环境中的其他行为者所看到的关于行动结果的观察性反馈。我们预计相同的计算原则适用于所有信息源，但学习参数在反馈类型和物种之间存在数量上的差异。模型适用于行为，基于模型的神经数据分析将揭示计算变量的大脑关联。我们希望在监测不同反馈信息来源的过程中发现解剖学和功能上的分离，并随后在实施未来行为变化的单一机制上趋同。在两个灵长类物种中使用互补的方法将导致更好的模型普适性，更好地理解潜在的神经机制，以及相互了解的记录点和分析的选择。

The consortium will develop computational models of learning and decision making and test them in two biological systems, humans and macaques. The models shall account for the use of different sources of information on action outcomes to guide future behavior: feedback on real outcomes, fictive feedback on outcomes that would have been obtained had a different action been chosen, and observational feedback on action outcomes seen in other actors in social situations. We expect that the same computational principles apply for all information sources but that the learning parameters differ quantitatively between feedback type and species. Model fits to behavior and model-based analyses of neural data will reveal brain correlates of the computational variables. We expect to find anatomical and functional dissociations during monitoring of the different sources of feedback information and later convergence on a single mechanism implementing changes to future behavior. Using complementary methods in two primate species will lead to better generalizability of the models, better understanding of underlying neural mechanisms, and mutually informed choice of recording sites and analysis.