The gorilla half the audience did not see

In 1999 the psychologists Daniel Simons and Christopher Chabris made a short video of two teams, one in white shirts and one in black, passing basketballs. Viewers were asked to count the passes made by the white team.

Partway through, a person in a full gorilla suit walks into the middle of the players, faces the camera, thumps their chest and walks off. Afterwards the researchers asked viewers whether they had noticed anything unusual. In their study, roughly half of the people did not see the gorilla at all.

Many who missed it refused to believe it had been there until they watched the video again. The finding built on earlier work by Ulric Neisser in the 1970s and is called inattentional blindness. In 2013, Trafton Drew and colleagues put a small gorilla image into lung scans and asked radiologists to look for cancer nodules. Twenty of the 24 experts, 83%, missed it.

Attention works like a spotlight. When you focus hard on one task, things outside that task can go unseen, even big and strange ones. Expertise does not protect you, because experts focus hard too, and they know very precisely what they are looking for.

The surprise is how confident people are that they would notice. That confidence is the risk. When you check a document only for typos, you can miss that the numbers in the table are wrong.

AI output invites narrow checking. You ask for a summary and check that it reads well, while a missing section or a reversed claim slips past.

The idea behind this caseLesson: Check what AI tells you →

1. 1Check important work in more than one pass, each with a different question: is it accurate, is anything missing, does it make sense overall.
2. 2Ask someone who was not focused on the task to look at the whole thing.
3. 3Use a short checklist for the errors that cost the most, so you look for them on purpose.
4. 4When an AI tool summarises a document, compare the summary with the original's headings to see what was left out.