On the missing training signal behind AI's jagged intelligence.
This preprint proposes that the jagged intelligence landscape of modern AI systems arises from a missing training signal the authors call "cognitive dark matter" (CDM): brain functions that meaningfully shape behavior yet are hard to infer from behavior alone. They identify key CDM domains — metacognition, cognitive flexibility, episodic memory, lifelong learning, abductive reasoning, social and common-sense reasoning, and emotional intelligence — and show that current AI benchmarks and large-scale neuroscience datasets are both heavily skewed toward already-mastered capabilities.
The paper outlines a research program centered on three complementary data types designed to surface CDM for model training: latent variables from large-scale cognitive models; process-tracing data such as eye-tracking and think-aloud protocols; and paired neural–behavioral data. Training on cognitive process rather than behavioral outcome alone could produce models with more general, less jagged intelligence — while advancing our understanding of human intelligence itself.
The preprint (arXiv, March 2026) is by Patrick Mineault and colleagues.
Read the paper: arXiv:2603.03414