You can agree or disagree, as you say, with SoM (Marvin was my undergrad advisor so my opinions are...complex) but that book can in many ways be considered a summation of decades of research in the field. To reinforce your point: even if you disagree with modeling cognition as a set of competing/cooperating "agents", there's enormous ahistorism in the paper. Which, TBF, is quite common in CS.

Also the thermostat has been extremely usefully used in an exploration of agency by Daniel Dennett, in his famous "The Intentional Stance"

(I accidentally typed "Intensional Stance" which really would be a fascinating book to read!)