Nov 9, 2023Liked by Felipe De Brigard

A good read, Felipe! For adversarial collaboration, you're right one way to go is to restrict them to highly informative experiments that terminate further interpretation. But check our new Neuron paper providing a Bayesian approach to adversarial collaborations that can score quite diverse theories against each other, as long as there is any difference in predictions. Then one theory can get ahead in the ongoing Bayesian horse race, allowing the rest of the scientific field to better place their bets (even if the losing theorist keeps flogging their theory). Adversaries then are wise to consider carefully what to say about the opponent's predictions (eg if they are considered banal, then make the same prediction, which renders that prediction uninformative). This moves much focus on to the bridge principles from theory to prediction - which is what you discuss for IIT (silent neurons, grid structure etc). Conceived like this, adversarial collaborations become quite interesting tools for science even for fledling fields like consciousness, or well-established ones like memory.

A. W. Corcoran, J. Hohwy and K. J. Friston. Accelerating scientific progress through Bayesian adversarial collaboration. Neuron 2023 DOI: 10.1016/j.neuron.2023.08.027

Expand full comment

Thank you so much! I am going to read your paper right away!

Expand full comment

Some interesting points of view here. I point you to the article below for a different perspective.

The strength of weak integrated information theory


Expand full comment