Wednesday, 5 June 2019

Sarcasm, nuance, research and AI


“Studies frequently rate Scotland as one of the happiest places in the world … largely due to researchers’ inability to detect sarcasm” – Frankie Boyle

There’s a big slice of truth in that gag. That’s exactly the kind of subtlety any researcher always needs to be on their guard for.

Exemplum: Today, I gave a focus group a choice of three names for a new project. While I got an answer, I wasn’t convinced by it. So, I pressed, and they admitted that they didn’t really “get” any of them. That’s the answer I needed, that’s the real insight.


I also think that’s why humans have an advantage over tech and AI. It’s the same reason that astronomers trust humans to spot anomalies better than computers. Humans see more clearly what doesn’t feel right. They get nuance. They can spot a lacklustre answer. And they (should) feel sarcasm.

A human researcher won’t accept the first, or second, answer if it doesn’t convince, explain, make sense, or if it isn’t delivered credibly. That enables them to understand more, to dig deeper, to get to the reason behind the answer.

AI can do some of the grunt work now, especially for the large data sets. And I long for the day when they understand us as well as we do – but I’m not yet holding my breath…

No comments:

Post a Comment