AI Exceptionalism
AI is often wrong, but never in doubt. The tendency to accept AI output as correct, as if it is somehow magical, is known as "AI exceptionalism."
Larissa Hamilton
Director
AI
/
Trusting the output because of how realistic and persuasive it sounds, and how instantly it arrives, is one of the more underrated risks of AI.
How “AI exceptionalism” plays out depends on how far along the professional pathway you sit:
For graduates, the issue can be learned behaviour.
Apps and platforms are intentionally designed to reward clicking, scrolling, and skimming, not questioning. AI invites the same kind of interaction. The smooth, confident-sounding “answer” can make us less inclined to scrutinise it. When pressed for time, it's tempting to take the output at face value.
For experienced professionals, the risk is more selective.
Obvious mistakes are filtered out quickly, but the danger lies where the AI output aligns with an existing viewpoint, follows a familiar structure and reaches a plausible conclusion. In those situations, the output can be accepted too quickly, because it sounds right, even where the underlying reasoning is far from sound.
To overcome AI exceptionalism, the discipline is the same in both cases. Treat AI output as a starting point to interrogate and verify, not as the final answer to be relied on.
