Computing & AI · the experiment
Unmask your favourite
Fifty pictures, twenty-six of them AI-made, over nine thousand favourites, and poems rated before being judged. Four questions, from beginner to expert, on the link between what we like and what we believe is human.
1 · Say what happens to the favourite
Beginner
How do respondents judge the picture they like best, compared with other respondents?
2 · Count the deceptive favourites
Among the respondents most hostile to AI art, 47% chose an AI-made favourite. Of those, 80% believed it was human.
Intermediate
What share of the whole group has an AI-made favourite they believe is human?
3 · Rule out recognition
Expert
Why does the effect not come from people recognising famous masterpieces?
4 · Explain the poetry paradox
In the poetry experiment, each poem was rated before guessing its author. AI poems were rated higher: 5.30 against 4.18 out of 7.
Expert
A higher rating therefore pointed towards AI. How did readers use it?
Favourite unmasked
—