Remix.run Logo
Confidence Scores for Exam Questions(nomagicpill.substack.com)
6 points by surprisetalk 3 days ago | 5 comments
clickety_clack 44 minutes ago | parent | next [-]

It would make more sense to just use IRT for grading the responses than trying to add more complexity to the answers themselves.

vmilner an hour ago | parent | prev | next [-]

I seem to remember some medical related multiple choice tests in the UK use a mechanism of +1 for correct , 0 for unanswered , -1 for incorrect.

CGMthrowaway 41 minutes ago | parent [-]

A system like that seems especially appropriate for a practice where the foundational principle is "do no harm."

bee_rider a minute ago | parent [-]

Would probably be applicable to engineers as well, or any other field where the practitioner has an obligation to be aware of the limits of their competency.

esafak an hour ago | parent | prev [-]

https://en.wikipedia.org/wiki/Calibration_(statistics)

https://en.wikipedia.org/wiki/Scoring_rule