finally published! Optimal metacognitive decision strategies in signal detection theory -- with the brilliant Brian Maniscalco and Lucie Charles #neuroskyence #neuroAI #metacognition #consciousness #psychology link.springer.com/article/10.3... - ThreadSky

meganakpeters.bsky.social • 101 days ago

finally published! Optimal metacognitive decision strategies in signal detection theory -- with the brilliant Brian Maniscalco and Lucie Charles
#neuroskyence #neuroAI #metacognition #consciousness #psychology

https://link.springer.com/article/10.3758/s13423-024-02510-7

Comments

elnaza.bsky.social•101 days ago

Ooh! I actually was looking for this! Published right on time!

meganakpeters.bsky.social•101 days ago

Signal detection theory (SDT) has long provided the field of psychology with a simple but powerful model of how observers make decisions under uncertainty. SDT can distinguish sensitivity from response bias and characterize optimal decision strategies.

meganakpeters.bsky.social•101 days ago

But classical SDT pertains to “type 1” judgments about the world. What about "type 2" SDT: sensitivity for metacognitive or “type 2” judgments about one’s own type 1 processing, e.g. confidence ratings?

Can you look at optimal decision strategies in type 2 SDT? YES!

meganakpeters.bsky.social•101 days ago

Turns out, what you SHOULD do in placing metacognitive criteria depends on what you're trying to optimize for... perhaps unsurprisingly 😉

meganakpeters.bsky.social•101 days ago

We derive formulae for 4 optimal type 2 criteria: maximizing type 2 accuracy, maximizing type 2 reward, calibrating conf to accuracy, & maximizing difference btwn type 2 hit & false alarm rates.

meganakpeters.bsky.social•101 days ago

Our framework can also be used to better understand metacognitive strategies of real observers, like characterizing observers’ spontaneously chosen strategies, assessing ability to fine-tune strategies when instructed, or determining over- or under-confidence relative to an optimal standard.

singhal.bsky.social•94 days ago

#consci

boryslaw.bsky.social•100 days ago

Let me point out a minor issue that I noticed reading the abstract. SDT is a *two* parameter model so it can *barely* describe the "decision process" of a thermostat...

boryslaw.bsky.social•100 days ago

An SDT fit is equivalent to a humble linear regression of the binary response on the binary stimulus because one can be (trivially) recovered *exactly* from the other...

boryslaw.bsky.social•100 days ago

Because this model is so simple, it implies extreme(ly unrealistic) homogeneity of the decision process. The best obe can say about SDT is that it can be used to *describe apparent sensitivity and response bias*, ...

boryslaw.bsky.social•100 days ago

but it cannot possibly provide unbiased estimates of the *latent* sensitivity and decision criterion. It's just way too simple for a structural (=quantitative causal) model of the response process...

boryslaw.bsky.social•100 days ago

Pretending that SDT or any if its identified generalizations are not way too simple to work as structural models because these models fit the data nicely is to confuse interpretability (causality) with predictive performance (correlation) + interventional & structural (=latent) causal properties :-)

Comments

Posting Rules

Reply