Bowers and Baker to lead NeuroAI methods critique at CCN 2026

The Cognitive Computational Neuroscience conference is opening its 2026 program with a public methods fight, and it is worth watching if you care about how brain-inspired models actually get evaluated. The first Group Analysis and Critique of the year is titled "NeuroAI Methods and Theoretical Frameworks", and the coordinators, Jeffrey Bowers of the University of Bristol and Nicholas Baker of Loyola University, are teeing up an argument about whether NeuroAI is using the right methodology to study mind and brain at all.

The framing is direct. As the proposal puts it, "NeuroAI and psychology often adopt different empirical methods and modelling approaches to understand the mind and brain." The NeuroAI camp tends to evaluate model predictions against behavioral and brain responses using naturalistic stimuli, aiming for human-like computational representations. Psychology, by contrast, tests specific hypotheses using controlled experiments with artificial stimuli. Whether those two things are converging on the same object of study, or quietly talking past each other, is the whole question.

The kickoff workshop is on August 4th at CCN 2026, and the speaker list is what makes it a real fight rather than an abstract one. The proposal page names Jenelle Feather from Carnegie Mellon, Anya Ivanova from Georgia Tech, Grace Lindsay from NYU, Aran Nayebi also from CMU, Milton Montero from IT University of Copenhagen, and Martin Schrimpf from EPFL. That is a cross-section of the labs that do the benchmarking work under discussion, which is why a critique from inside the tent lands differently than one from outside it.

The honest caveat is that the page itself is a call for community feedback, not the result of one. What the reporting doesn't give you is which specific benchmarks or published claims the coordinators think are broken, what reforms they intend to argue for on August 4th, or whether the GAC will produce a written output the wider field can cite. Take this as a preview of an argument, not a conclusion.

Still, if your work leans on the "this model matches brain data" line of evidence, whether for a grant, a paper, or a product story, the methods people at CCN are about to get louder about what that line actually shows, and the useful move now is to know which of your claims would still stand if the naturalistic-benchmark framing gets downgraded.