VOLUME LXVI | JUN 2022

The Surprising Connection Between Positive Biases and Reinforcement Learning

When it comes to learning and making decisions, humans don’t always play fair. We have this tendency to be biased towards positive outcomes and information that confirms what we already believe. It turns out that these biases are not just limited to ‘high-level’ belief updates, as previously thought. Recent evidence suggests that the same biases can be observed in reinforcement learning tasks. In other words, the way our brains process and update beliefs and values shares some common computational principles with how we learn through positive reinforcement. These findings challenge our understanding of bias in decision-making and highlight the need for further research to unravel the complex relationship between emotions, cognition, and learning. If you’re intrigued by these fascinating connections, dive into the research to explore the computational roots of positivity and confirmation biases in reinforcement learning.

Li Wei Chen

Li Wei is a Chinese-Canadian neuroscientist in Vancouver, studying brain plasticity and lifelong learning. He contributes articles on harnessing neurotechnology to expand human capabilities, drawing from his experiences in cross-cultural innovation hubs.

Humans do not integrate new information objectively: outcomes carrying a positive affective value and evidence confirming one’s own prior belief are overweighed. Until recently, theoretical and empirical accounts of the positivity and confirmation biases assumed them to be specific to ‘high-level’ belief updates. We present evidence against this account. Learning rates in reinforcement learning (RL) tasks, estimated across different contexts and species, generally present the same characteristic asymmetry, suggesting that belief and value updating processes share key computational principles and distortions.
Read Full Article (External Site)