Excited to see this now out in the world! We identify a computational basis for how persistent underconfidence is maintained in the face of intact performance, finding that it is grounded in impaired updating of global self-beliefs from local metacognition www.nature.com/articles/s41... 1/N - ThreadSky

smfleming.bsky.social • 129 days ago

Excited to see this now out in the world!

We identify a computational basis for how persistent underconfidence is maintained in the face of intact performance, finding that it is grounded in impaired updating of global self-beliefs from local metacognition

https://www.nature.com/articles/s41467-025-57040-0

1/N

Comments

smfleming.bsky.social•129 days ago

Developing these models opens doors for a more systematic computational treatment of global aspects of metacognition

Work led by Sucharit Katyal together with @docqhuys.bsky.social and Ray Dolan

juanbalc.bsky.social•128 days ago

The frequent explanation from the decision-making literature is reduced information sampling linked to anxiety and depression. One other related factor is abnormal learning rates explaining underconfidence.What are your thoughts on those 2 as potential and (interacting) factors in this work?

lenoreblum.bsky.social•128 days ago

Thanks for link to the paper. Just printed it. Also wonder about the opposite. Persistence of global overconfidence grounded in impaired updating of global self-beliefs from local negative feedback. EG in mania.

smfleming.bsky.social•127 days ago

Hi Lenore, thanks! Yes indeed, one future direction is to model/study the opposite bias leading to phenotypes where overconfidence dominates

lenoreblum.bsky.social•127 days ago

Many thx!Our prediction from the CTM would be that,for individuals where overconfidence dominates,confidence will generally over rule globally,even given local neg feedback.Unfortunately, we are suffering that syndrome in the US where local neg feedback seems to provoke global overconfident behavior

lenoreblum.bsky.social•128 days ago

Reason for my query: The CTM model has a parameter d between [-1, +1] which we call disposition. If d>0 (optimistic CTM) then + valenced chunks submitted by LTM processors to the competition for global broadcasting have higher probabilities of winning than neg.valenced chunks.

lenoreblum.bsky.social•128 days ago

The opposite happens if d<0 (pessimistic CTM). In the latter case, even if processors are given positive feedback, negative valenced chunks still rule globally. The extremes cases (+1, mania; -1, deep depression) in CTM are also interesting.The only way to get out of deep depression is to reboot.

Comments

Posting Rules

Reply