May 28, 2026
Pop Quiz: How sure are you?
Confidence Scores for Exam Questions
The internet says ‘cool idea,’ then immediately points out schools have done this forever
TLDR: The article argues exams should score not just answers, but how confident students are, to punish lucky guessing and reward real knowledge. Commenters mostly replied with a collective eye-roll, saying versions of this have existed for years and debating whether it’s clever reform or just extra hassle.
A proposal to make students rate how sure they are on every exam answer has kicked off the kind of comment-section showdown that turns a nerdy testing idea into full-on forum theater. The basic pitch is simple: instead of only marking an answer right or wrong, students would also say whether they’re 50%, 70%, or 100% sure. The goal? Separate the people who truly know the material from the lucky guessers who stumble into the same score.
And yet the community reaction was less “revolutionary!” and more “my dude, this already exists.” One commenter casually dropped that some UK medical exams already use a harsh little scoring system: right gets a point, blank gets zero, wrong gets you punished. Another piled on with the devastatingly calm reminder that Moodle has had “certainty based marking” for 15 years. Ouch. Then came the classic internet flex: links to calibration and scoring rules, basically saying, “welcome to the literature, pal.”
The hottest split in the comments was over whether this makes tests smarter or just more annoying. One side likes the idea of exposing bluffing and rewarding real confidence. The other side thinks adding a second answer bubble to every question sounds like turning exams into paperwork cosplay. Bonus elite-math cameo: Terence Tao already blogged about a version of this, which gave the thread a very “the final boss has entered the chat” energy.
Key Points
- •The article says traditional multiple-choice and free-response exams do not measure student confidence and can reward correct guesses the same as certain knowledge.
- •It proposes using Brier scores to combine answer correctness with a student’s stated confidence for each question.
- •The article defines a perfect Brier score as 0 and notes that relying on Brier score alone could create incentives to be confidently wrong.
- •The suggested implementation adds a second response field per question for confidence, with levels from 50% to 100%, and could be supported by modified Scantron forms.
- •The article concludes that the idea overlaps with existing literature on confidence-based marking, including work by A.R. Gardner-Medwin.