Neural mechanisms of predictive processing: a collaborative community experiment through the OpenScope program

A landmark collaborative paper, authored by 53 leading neuroscientists and coordinated through the Allen Institute's OpenScope program, has laid out a definitive roadmap for experimentally testing the brain's 'predictive processing' theory. This theory posits that the brain is not a passive receiver of sensory data but an active predictor, constantly generating models of the world and updating them based on prediction errors. The review, published on arXiv, consolidates years of research to identify key computational building blocks—like stimulus adaptation, hierarchical processing, and the balance of excitatory and inhibitory neurons—that are believed to underpin this predictive capability.

To move from theory to concrete understanding, the authors propose a series of ambitious, standardized experiments to be run on both mice and primates. These experiments will use advanced techniques like in-vivo two-photon imaging and electrophysiological recordings to measure how neural circuits respond to specific 'mismatch' stimuli, such as omitted sensory events. A core innovation is the use of the OpenScope platform, modeled on astronomical observatories, to execute these experiments and immediately share the vast, standardized neural datasets with the global research community.

This open-science approach is designed to accelerate progress by allowing computational neuroscientists and AI researchers worldwide to test their models against real, high-quality brain data. The initiative aims to resolve long-standing debates, such as whether different types of prediction errors engage shared or distinct neural circuits across sensory modalities. By fostering iterative refinement between experimental data and computational models, the project seeks to finally decode the algorithmic principles the brain uses for prediction—principles that could inspire more efficient and robust artificial intelligence systems.