POSTER DETAILS
Decisions are guided by learning and perceptual biases in a 2-alternative-forced-choice task
Elena Menichini, Liang Zhou, Victor Pedrosa, Peter Latham, Athena Akrami
Date / Location: Tuesday, 12 July 2022 / S05-103
Add this poster event to your 🗓
Add this poster event to your 🗓
Traditionally, improved performance is a key hallmark of learning. However, even if the process of learning may eventually lead to near-perfect performance, the learning trajectory depends on many factors, such as trial order or how the task is represented. In addition, a myriad number of biases affect human and animal performance, for instance serial (Fritsche et al. 2017), choice (Busse et al. 2011, Abrahamyan et al. 2016) or sensory history biases (Akrami et al. 2018, Ashourian et al. 2011) in perceptual decision-making tasks. Here, we find that rats and humans show different learning patterns in a 2-alternative forced-choice (2AFC) task involving the categorization of auditory stimuli based on a perceptual boundary. We show that models based only on feedback-dependent learning, including those incorporating statistical decision confidence, are not sufficient to explain data about the effect of a previous stimulus on the current stimulus. Instead, we identify a stimulus-dependent repulsion effect that contributes to learning in this task. This repulsion effect, in tandem with a feedback-dependent component that may depend on confidence, can recreate patterns in the data with high fidelity. From there, we further isolate the stimulus-dependent component in a separate experiment that limits feedback to only a certain portion of trials and find that purely stimulus-dependent learning can account for the data. We conclude that both stimulus-dependent and feedback-dependent processes are necessary to explain patterns of learning.