Analysing Public Consultations
(Originally published on LinkedIn): Like many people in publishing, I’ve been formulating a response to the UK Government’s Consultation on AI and Copyright. With many leading industry and advocacy groups encouraging contributions, it seems likely there will be a big response. So I was interested to read last week that one of the projects developed by the Government’s Incubator for Artificial Intelligence is a tool to use AI to analyse the results of large public consultations.
According to the documentation, analysing consultation responses costs £80m+ a year, so there’s a non-trivial cost-saving. The project uses LLMs for topic modelling and thematic analysis. And there is a lot that I like about the approach to the project and the Incubator generally: open source code, preserving raw responses alongside the analysis, publicly stated governance and assurance principles, and human-in-the-loop. It mentions previous work with DHSC and DWP, and now scoping new areas for evaluation. But that made me wonder: could the responses to the AI and Copyright Consultation end up being analysed with an AI tool? Surely not. But if so, there would be some follow-up questions, not least what LLM is used.
I emailed the Incubator for Artificial Intelligence over a week ago just to ask if this tool would be used for the AI and Copyright Consultation, and I’ve had no response other than an automated reply to say they get a lot of email traffic and would reply as soon as possible. It may well be that there is a very quick and simple answer that the two things are not linked, but I haven’t seen anyone talking about this and it would be great to confirm the position. So I’m highlighting it here and in my newsletter this week (and will gladly update this post if the position is made clear).