Sustained effects of corrupted feedback on perceptual inference

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Neuroscience

Background:

• Feedback is crucial for learning and adapting perceptual systems.
• The reliability of feedback directly influences the accuracy of learning and inference.
• Understanding how unreliable feedback impacts perception is key to cognitive and learning theories.

Purpose of the Study:

• To investigate the impact of corrupted feedback on perceptual inference within a Bayesian framework.
• To test the hypothesis that unreliable feedback compromises sensory processing and biases inference towards prior information.
• To examine the sustained effects of corrupted feedback on the balance between sensory evidence and prior beliefs.

Main Methods:

• A computational simulation was conducted to model the effects of feedback corruption.
• Two behavioral experiments involving visual detection and discrimination tasks were performed.
• Intervention sessions with corrupted or uncorrupted feedback were compared with pre- and post-intervention tests, including auditory cues to induce prior beliefs.

Main Results:

• Both simulation and behavioral experiments demonstrated decreased performance with corrupted feedback.
• Participants showed an increased reliance on prior beliefs after exposure to corrupted feedback.
• Exploratory analyses revealed reduced decision confidence during corrupted feedback delivery.

Conclusions:

• Corrupted feedback induces sustained changes in perceptual inference, shifting reliance from sensory evidence to prior beliefs.
• The findings highlight the critical role of feedback reliability in shaping perceptual learning and decision-making.
• These results have implications for understanding learning deficits and designing effective feedback mechanisms in various domains.