Same same but different

I've been working with AI… who hasn't right? Well anyway I have been using Dovetail to do a lot of classification. At first I ran the classification on interviews and I was like eh, this is bang average. There's a bunch of random hallucinations and stuff in here, I'm having to spend 10-20 minutes QA'ing it. Dovetail uses the Claude AI agent. So the next project I sat and tagged it by hand. Hours later I had completed my tagging. Bored out of my mind I finally tagged the last segment and thought I had done a great job… I had not.

A week later a colleague asked about the research and I was convinced there was a clip talking to their point, several in fact. So I jumped into a quick search to pull up the quotes. A quick search became a 10-minute exercise, turned into a 20-minute search for a unicorn. In the end I gave up. I couldn't find it anywhere, and some clips I thought were talking to the point were about something similar but different. I had experienced my own personal hallucinations.

I looked back at my own tagging, and I found mistakes and missed tags, but there was still no evidence to support my point. What I had realised instead was that I'm making just as many mistakes as the AI did, except it took me hours instead of minutes to get there. Yes, the mistakes are more human, but they are mistakes nonetheless. Same same but different.