Computational analysis of value learning and value-driven detection of neutral faces by young and older adults

Abstract

The rapid detection of neutral faces with emotional value plays an important role in social relationships for both young and older adults. Recent psychological studies have indicated that young adults show efficient value learning for neutral faces and the detection of “value-associated faces,” while older adults show slightly different patterns of value learning and value-based detection of neutral faces. However, the mechanisms underlying these processes remain unknown. To investigate this, we applied hierarchical reinforcement learning and diffusion models to a value learning task and value-driven detection task that involved neutral faces; the tasks were completed by young and older adults. The results for the learning task suggested that the sensitivity of learning feedback might decrease with age. In the detection task, the younger adults accumulated information more efficiently than the older adults, and the perceptual time leading to motion onset was shorter in the younger adults. In younger adults only, the reward sensitivity during associative learning might enhance the accumulation of information during a visual search for neutral faces in a rewarded task. These results provide insight into the processing linked to efficient detection of faces associated with emotional values, and the age-related changes therein.