Cognitive Dark Matter: Measuring What AI Misses
Patrick J. Mineault, Thomas L. Griffiths, Sean Escola · 2026
Read Cognitive Dark Matter: Measuring What AI Misses by Patrick J. Mineault, Thomas L. Griffiths, Sean Escola (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-03414.txt.
Description
Patrick J Mineault, Thomas L Griffiths, and Sean Escola argue that artificial intelligence has a blind spot shaped like the human mind itself. Modern systems are oddly uneven: brilliant at some tasks, helpless at nearby ones. Their 2026 paper blames a missing training signal they call cognitive dark matter, brain functions that shape behavior but are hard to infer from behavior alone.
The missing domains include metacognition, mental flexibility, episodic memory, lifelong learning, abductive reasoning, social and common-sense reasoning, and emotional intelligence. The authors show that benchmarks and large neuroscience datasets both over-measure what systems already do well. Their research program proposes three new data types: latent variables from cognitive models, process traces like eye movements and think-aloud records, and paired neural and behavioral data. Training on process, not just outcomes, could smooth the jagged profile.
Catalog record
| CREATOR | Patrick J. Mineault, Thomas L. Griffiths, Sean Escola |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31019 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.03414