Week 6: Perception as construction
The lecture where the direction of flow reverses, and most of what you take to be sensing turns out to be generation checked against input.
On predictive accounts the system produces its expectation and passes up the error, which makes illusions predictions of the framework rather than curiosities at its edge. You go on seeing the Adelson checker-shadow wrongly after you understand it, because the inference runs below the level at which understanding operates.
The consequence is the sentence students find hardest: nothing tags which parts of your experience were sensed and which were generated.
I spend the last twenty minutes arguing against the framework I have just presented. The dark room problem is the sharpest objection, and Litwin and Miłkowski’s methodological version is the one I find hardest to answer.
Teaching team
- Tuyet Nabarro — lecturer