When AI Gets in the Way of a Good Idea...

A small bit of news, I’ve renewed my AHRC Peer Review College membership for another four years. I first joined back in 2016, and I’m very happy to be continuing. It's a lot of work, but I actually really enjoy reviewing. It’s a chance to see what people are working on across the discipline, and I sometimes get genuinely excited with the great ideas people have. It’s also probably one of the best ways of learning how funding applications work, which in turn helps when applying for grants yourself. 

I have noticed that the number of review requests seems to have gone up quite dramatically in the last couple of years. Whether that reflects more applications, fewer people agreeing to review, or something else, I don't know. But there is another change that has become increasingly noticeable, the now ever present AI. For the avoidance of doubt, I am absolutely not against using AI. I use it a lot. It is brilliant for things like reducing an over length paragraph, checking requirements against guidance, spotting repetition, or helping you see where an argument isn't as clear as you thought it was. My most frequent use is in Outlook, having it add meetings to my calendars directly from email conversations where people don't use the meeting invite tool, you would not believe how much time that has saved me. And it isn't all LLMs - with our XCT lab for example, AI is crucial for analysing the highly complex 3D data and to detect features of interest. 

But it is a tool that needs to be used very selectively, and I think many people are falling into the trap of getting it to generate entire sections of text. When you read enough applications, completely AI generated passages start to become quite obvious. There is an oddly distinctive way of structuring sentences and arguments that crops up again and again, and you get a real sense of having read it before. There is a tendency towards padding and waffle, with long introductions to fairly simple points, unnecessary restatements, and big, hyperbolic claims about how something will be “transformative”, “groundbreaking” or “revolutionary”. It feels a little like the new version of those cheesy PowerPoint templates and animations that people used to use. Nothing technically wrong with them, but once you'd seen enough of them, you could spot them instantly and they didn't exactly make the presentation better. Once you had seen enough gradients, animations and WordArt, you noticed the template before you noticed the content.

So I guess the key point is, as ever, don't make things hard for the reviewer, write the first draft yourself, keep your unique voice and think of AI as a slightly more complex spell checker, not a co-author!

Comments