The Transparency Paradox in Explainable AI: A Theory of Autonomy Depletion Through Cognitive Load

More transparency isn't always better: The paper challenges the assumption that explaining AI decisions is universally helpful. The same explanation can improve decision-making in some situations but actually harm it in others, depending on how cognitively taxed a person already is.

Cognitive overload is the hidden culprit: Rather than problems with trust or understanding, the research identifies cognitive load as the key factor. When people are mentally exhausted, additional AI explanations deplete their remaining mental resources, causing them to disengage or make worse decisions—a phenomenon the authors call "autonomy depletion."

Timing and context matter more than content: The optimal approach isn't to provide maximum or minimum explanations, but to adapt explanations based on real-time assessment of a person's cognitive state. Information should be provided when someone is mentally fresh and has capacity, but withheld when their cognitive resources are already depleted.

One-size-fits-all transparency policies fail: Static design choices about how much to explain don't work across different users and situations. Instead, AI systems should dynamically adjust transparency levels based on monitoring whether a person still has the mental capacity to meaningfully engage with explanations.

Protecting human autonomy requires strategic restraint: Counterintuitively, responsible AI design sometimes means not explaining everything. Withholding information at the right moments preserves people's ability to make autonomous decisions, whereas overwhelming them with explanations paradoxically undermines their autonomy.