Learning to make decisions based on feedbacks is fundamental for an organism, but relatively little is known about how we integrate the early visual processing to achieve the final decision when motivation differs. To observe the role of visual fixations, we recorded early gaze behaviors using an eye-tracking experiment, which required subjects to choose between different stimuli that were probabilistically associated with rewards for themselves, another person, or no one. Using computational modeling, we showed that people could learn to make decisions with different motivations and these learning procedures exhibited through eye movements can be explained by reinforcement learning model, which is driven by prediction error that is the difference between the predicted and actual outcome of a choice. We showed that people could learn to obtain rewards for another person and no one but do so more slowly than when for themselves. Gaze data revealed that larger difference of associative value between the target and distractor symbols resulted in more initial saccades to target symbol, more fixation numbers before reaching the first symbol, longer fixation durations on the first symbol, but shorter response time to the final decision. Moreover, subjects needed fewer fixation numbers to reach the first symbol when learning for themselves and another person, consistent with the larger learning rate for both conditions. We thus revealed the early gaze evidence and its computational mechanism during learning with various motivations. This framework can provide insights for understanding learning behaviors under different contexts.
刘彦平，中山大学心理学系特聘副研究员。研究兴趣为语言认知、学习与决策及视知觉等，主要研究方法为行为实验、眼动及计算建模。研究成果发表在Psychological Science，Cognitive Science，Journal of Experimental Psychology: Learning Memory and Cognition，Journal of Experimental Psychology: Human Perception and Performance, Psychonomic Bulletin & Review等著名心理学期刊。