- Author
- Matthew Somerville
Posted by
At SocietyWorks we run a service for local authorities called WasteWorks, one part of which includes, alongside bulky collections, garden subscriptions and the like, the important service of finding out when your bin days are (speaking as someone who hasn’t had a recycling collection in 18 months due to the Birmingham bin strike).
I was recently reading Tom Loosemore’s blog post about AI agents (meaning LLM-powered services operating browsers directly or requesting things from websites, notionally on someone’s behalf), and his previous post about being “beaten” by council websites when trying to make a national bin day site (due to the variations in council websites being too much for his attempt), and got to thinking – AI agents can be useful to individuals, but using them in this way is collectively causing problems, and we need to keep the focus on improving and fixing the plumbing that helps us deliver sustainable digital services.
Tom’s Bluesky post led to a number of follow-on discussions about bins:
- Tom Steel mentioned localgovdrupal’s bin module, which acts in a similar way to our own sites, and would let councils that use localgovdrupal more easily add a bin day section to their site – a win for collaborative and shared working;
- Phil Rumens and Paul Connell had a discussion about data publishing and a previously proposed waste data standard that wasn’t adopted;
- And Rachel 🦁 via Owen Boswarva linked to the UKBinCollectionData project.
The last of those is especially interesting, as it is an existing way of getting the data Tom was trying to, with collaborative (169 contributors so far) open source software, and it even handles the one (Westminster Council) that stumped Tom’s AI fine. It integrates with Home Assistant, if you use that, and could presumably be used as a base for other things. It may be boring, compared to the thought of a magic-seeming “agent”, but scrapers/agents that accept instructions from what they read, where there is no privileged set of instructions or separation of data, are fundamentally risky (e.g. “Word worm crawls into Copilot”) compared to a script using and reading an open standard.
Meeting user needs in a simple, sustainable way
I wonder whether people actually would want a single national service to look up their bin days. They just want to find out when their bin day is, and I wouldn’t search for a special app/site for that. People might go on their council website (it’s printed on the bin), they might search, they might ask AI (though Google’s AI overview being asked when is your bin day currently says “use the council tool” but asks for your postcode(!), then says “use the council tool” but still asks for your precise address(!!), and then says… “use the council tool”), they might (shock) ask their neighbour – but what’s needed by all of those is for the data to be available in some manner for all these routes to use, which doesn’t need consolidation.
All WasteWorks bin day pages publish iCal feeds (.ics files) of collections, an old (RFC 2445 was in 1998) data standard for publishing calendar events. They can all utilise UPRNs, the UK data standard for uniquely referencing properties (though in some places for performance reasons we use other property identifiers). They all have nice, bookmarkable URLs, so people can save their bin day page in their bookmarks, or as a progressive web app, if that’s how they want to get the information.
In terms of standards, I think a good role of the centre is to provide platforms and ways for things to operate, and inter-operate, as we talk about in our paper Unlocking the value of fragmented public data. In this case, for example, they could recommend that councils provide such iCal feeds for bin day pages, so people, scripts, or agents don’t have to parse HTML web pages or PDFs; they could propose a URL standard for the location of a bin day page via UPRN, so automation could locate them more easily (https://www.gov.uk/rubbish-collection-day already exists, at a less automated level, for example). Backend systems can’t publish all the data, and I’m not sure we need them to, but locating that information is what people want to do, and we then surface what we can.
The tragedy of the commons
However, we also, like many others, are under increased constant barrage by badly written AI bots (perhaps running off your cheap streaming stick). Be that for scraping for training data, bots trying to find security holes, or people making badly written apps to look up information on a large scale, such activity can potentially cause knock-on issues for actual users of the site. And if everyone is doing the same thing, wasting resources over and over, with no thought to the collective benefit or disbenefit, then we find ourselves in a tricky situation.
My colleague Alex talks about 3 types of state interaction in the context of AI-driven demand – private benefit (for the requestor, e.g. a grant, or a benefit application), collective benefit (benefits the requester and wider society, e.g. a public FOI request, or reporting a pothole), or a zero-sum interaction (e.g. planning). Getting a script, or AI agent, or whatever today’s buzzword is to fetch bin day information is a private benefit, whilst having some shared code on GitHub helps turn that into a collective benefit. And having easily usable bin day pages with iCalendar feeds helps increase the collective benefit.
In that vein, I have submitted a Pull Request to the UKBinCollectionData project to switch to using our iCal feeds for all the council bin day pages we run, rather than the various more brittle web scraping that they currently use, which should help both users and us, providing more data, in a more performant manner. This I feel is how we best deliver what people want – when is bins?
Bin photo CC-BY, by David Merrett