Erwin de Wolff

Researcher of Cognitive Modelling at the Donder's Centre of Cognition.

[Erwin de Wolff @Erwin_de_Wolff]

@PratYosef @DanajaRutar @IrisVanRooij @JohanKwisthout Btw, I do think that all belief models (including our own) have both cognitive blindness and category conflation at (all) times. In fact, we often *discover* new concepts that show us this was the case in hindsight. This discovery process, we argue, is structure learning proper

[Erwin de Wolff @Erwin_de_Wolff]

@PratYosef @DanajaRutar @IrisVanRooij @JohanKwisthout I do agree that categories are a construct of the mind. However, that does not mean that they are not based on very real features and differences. Blindness and conflation then, are the inability to create new, useful categories despite good reason to do so.

[Erwin de Wolff @Erwin_de_Wolff]

@sir_deenicus @IrisVanRooij Exactly, those are the types of phenomena that are (theoretically) well-represented by particle filters. Interesting avenue for further research

[Erwin de Wolff @Erwin_de_Wolff]

@sir_deenicus @IrisVanRooij If each particle is a full belief model, we would have competing models in our mind, some of which would die out from uselessness, other reinforcing themselves (quite like in memetics I'd argue). But whether this is a solid cognitive theory or a statistical model is unclear to me

[Erwin de Wolff @Erwin_de_Wolff]

@sir_deenicus @IrisVanRooij I have also been thinking about particle filters as a, if nothing else, algorithmic description of how we might discover new concepts 'suddenly'. However, it is not entirely clear how one would select from arbitrary changes, many of which would be quite useless. (1/2)

[Erwin de Wolff @Erwin_de_Wolff]

If this is not worthy of a first tweet, nothing is: my first paper, together with the wonderful @DanajaRutar, @IrisVanRooij and @JohanKwisthout is now out:
"Structure Learning in Predictive Processing needs Revision (https://t.co/QUfh7zxHkG)