😂🙏

These computational considerations have implications for understanding the algorithms supporting goal pursuit—and how miscalibration of these mechanisms might relate to mental health

We explore three adaptive reasons for "stubborn goals": (1) efficient use of cognitive resources, (2) shielding decisions from interference/temptation, and (3) scaffolding motivation when rewards are abstract or distant

Thanks Mirela!!

Thanks!!

Ah yes true, they would probably relearn Rule A with feedback but I expect they would retain good generalisation (to untrained test examples from the same task) unlike the splitter group

Thanks! Great question, our task isn’t long enough to tell but I’d love to know this too - my prediction would be that more splitters would become lumpers (ie learn to generalise with time and/or sleep like “grokking”) but very interested if you’d predict otherwise!

Our work bridges psychology & AI, showing that continual learning is shaped by how knowledge is organised, with shared trade-offs across both systems.

We could capture this mixture of behaviour by tweaking the training regime of ANNs, (‘rich’ vs. ‘lazy’) shifting them towards shared representations (enabling transfer/generalisation but at the cost of interference) versus separated representations. (7/8)