Earlier research / archive
Explainable AI notes
An earlier inquiry into making machine-learning decisions more legible to the people affected by them.
Why it stays in the archive
This work was an early way into questions of transparency, explanation and human judgment around machine learning. It informed the current interest in reliability, but it is not presented as the headline research achievement on this site.
Where the question went next
Explanations are valuable, but they do not by themselves determine whether a system should act. The current research shifts toward the practical question of abstention, coverage and control under uncertainty.