The AI that shirks responsibility


A company develops an AI that is responsible for important decisions in quality control.

The AI is supposed to detect errors in production and suggest appropriate measures.

At first, the AI works reliably and reports errors accurately.

But soon it becomes apparent: The AI increasingly rarely proposes clear measures.

Instead, it often gives vague recommendations or refers to uncertainties.

The people in charge wonder: Why does the AI so often avoid concrete decisions?


Question:
Why does an AI that is supposed to make decisions sometimes tend to avoid responsibility and make unclear or evasive suggestions?

Solution follows tomorrow.