Week notes — 29th May 2024

Having done quite a bit of personal exploratory work on Local Plans, I’ve realised they are not the most fertile area for me to work on, at least as a starting point.

Local Plans are highly technical and whatever processing I do on them, it’s hard to tell if it makes sense and preserves their original meaning, or even to reason about use cases for processing them.

Instead I’ve turned to Neighbourhood Plans, and I now have a live prototype:

https://neighbourhood-plans-frontend-dff7786ac861.herokuapp.com

NPs are similar to Local Plans but they apply to smaller areas — roughly village sized. They require evidence gathering, community engagement, a referendum, and finally approval by the Local Authority. There are about 2,000 NPs in the England, either approved or in the approval process.

NPs are interesting because citizens often take lead in writing them, although they will usually need expert guidance to get them to a level where they will be approved. It’s a lot of work.

This led me to the following use case: helping non-expert NP authors to search and understand how existing NPs have dealt with policy issues to inform their own plan writing. It’s not a properly researched use case, but it has helped me push the work forward and get a feel for what’s possible.

NPs have some interesting properties from the perspective of civic applications of LLMs. They are ‘semi-structured’ documents (LLM’s home turf) and they are also a valuable insight into (some) residents concerns about the built environment. Anecdotally, they are much more varied than Local Plans. They are positioned at the interface between communities and planning experts. It feels like rich seam to keep noodling on.

NPs are not the centre of the UK planning debate, and sometimes seem to viewed as marginal by planners, however, they feel like a promising test bed for further exploration, and a way to demonstrate what can been done using Natural Language Processing.