Lost in the Literature: AI and the Problem of Mathematical Search
In 2016, Terry Tao published a whimsical blog post in which he derived a mathematical formula for grading. Rather than relying on binary “True/False” answers, he proposed a scheme in which students would report both an answer and a subjective confidence level. Using ideas from information theory, Tao arrived at a logarithmic scoring rule designed to incentivize students to report their true beliefs rather than strategically hedging.
2026-02-17 In 2016, Terry Tao published a whimsical blog post in which he derived a mathematical formula for grading. Rather than relying on binary “True/False” answers, he proposed a scheme in which students would report both an answer and a subjective confidence level. Using ideas from information theory, Tao arrived at a logarithmic scoring rule designed to incentivize students to report their true beliefs rather than strategically hedging. Almost immediately, a commenter going by sflicht pointed out that Tao had independently rediscovered a well-known concept: proper scoring rules, a…
saved by
related reading
- As Rocks May Think | Eric Jangevjang.com
- The fall of the theorem economydavidbessis.substack.com
- Mathematics in the Library of Babel - Daniel Littdaniellitt.com
- Declaration — Math and AImathandai.org
- [2602.05192] First Proofarxiv.org
- What's new | Updates on my research and expository papers, discussion of open problems, and other maths-related topics. By Terence Taoterrytao.wordpress.com
- [2608.16753] Mathematics in the age of AIarxiv.org
- A Severe Misalignment of AI in Mathematicsterrytao.wordpress.com
- A New Consciousness of Mathematicsapoorvapanidapu.substack.com
- A recent experience with ChatGPT 5.5 Pro | Gowers's Webloggowers.wordpress.com
- Terence Tao – Kepler, Newton, and the true nature of mathematical discoverydwarkesh.com
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelsarxiv.org