Why do some think organisms using Shannon information affirms it is computed by brains? Brian Skyrms’ account shows that meaningful convention-based signalling can be a purely behavioral and situational matter.

1. Agents exist in states

2. Agents have probabilities of taking actions in states

3. Signals index states discriminated by senders

4. Signals alter probabilities of actions taken by receivers

5. By trial and error, agents propagate info with signals by altering action probabilities relevant to states.

At no point does an agent process information per se, rather it adapts its behavior conditional on outcomes of behaviors experimentally tracking different random signals.

Shannon info is not known to agent, it just measures behavior being however altered in likelihood by signal.