PMQs

First published: ; Last modified:

In classic Sixth Form work experience fashion, I spent a week last summer shadowing an MP. It was not an especially exciting experience – there was very little work for us to do, and the staffer I “reported to” seemed to have no intention to actually use any of the research I produced – so I mostly just sat in Portcullis House doing my own projects and switching into anthropologist mode.1

We were sent off to watch some debates in the chamber, which I found pretty depressing. The main thing I remember is spending ten minutes listening to peers talking about cats. This is a cherry-picked example: they discussed teacher vacancies, decarbonisation, and rape prosecutions in the same session. Even on the weightier topics,2 though, it felt like there was a debilitating lack of detail. Theatrical, adversarial arguments are not conducive to good truth-seeking or policy-making.

There was a time when I loved hearing about the ups and downs of Westminster; the ingenious procedural power plays and dramatic late-night government defeats. Prime Minister’s Questions is the best-known part of that culture, and also the most obvious demonstration of what is wrong with it.

.

Fact-checking

The puerility of PMQs particularly irked me mid-March (scrolling through the Guardian live feed and sighing in exasperation). I thought it might be an interesting idea to try to catalogue all the factual claims made in one session, and see how accurate they were.

This post is partly inspired by the idea of epistemic spot-checking. Some general notes about the practice:

  • It’s incredibly time-consuming
    • I don’t know many examples of people doing it. Stefan Schubert annotated a US debate but he didn’t say how long it took.
    • Even just to get a list of the claims took me a while, because GPT-4 on a first attempt was not very effective at extracting them from all the fluff and bluster.
  • There doesn’t seem to be any routine fact-checking of House of Commons
    • Full Fact apparently used to live-tweet fact checks of every PMQs,3 but it looks like they haven’t done that since at least the start of the year.
  • I think it’s something you could automate successfully with GPT-4 and internet access

PMQs has fewer empirical assertions than you might expect. Isabel Hardman has written about the abundance of “motherhood-and-apple-pie” questions backbenchers like to ask – things along the lines of whether the Prime Minister can “reassure [an MP’s] constituents that he will continue working hard to get inflation as low as possible”. They’re a distinct problem, but equally indicative of the inefficacy of much parliamentary scrutiny. I reckon LLMs could classify them fairly accurately too, but unless there is some kind of social or political penalty for MPs who ask these time-wasting questions, they will presumably just continue.

I have not yet got around to actually fact-checking any of these claims, but it’s something I do intend to make time for eventually.

Economy

Crime

Immigration

Health

Fertility Treatment and Employment Rights

Local Issues and Policies

Foreign affairs

  1. I’ll write a post about this at some point. In lieu of that, the rough idea is that when an environment feels boring to participate in, you can observe it instead (or be a participant-observer, I suppose). This is especially fun when you’re in a social setting whose rules you haven’t quite learned, and you try to infer them from what you see, a bit like inverse reinforcement learning. See Aella’s post on “Learning the Elite Class”. More practically, different workplaces have very different cultures (from my admittedly limited experience). GPI was very different to Edge Health, and I substantially preferred both to Parliament. Potentially apart from credentials and “climbing the ladder of social proof”, as someone else put it, learning stuff like this has been by far the most valuable thing to come out of any work experience / internships I’ve done to date. ↩︎

  2. I don’t object to raising the profile of animal welfare – far from it! But it is obviously absurd when a speaker prefaces their contribution by “declar[ing] an interest as the property of a very sophisticated cat called Loki” [sic]. Mildly entertaining, perhaps, but not that funny when you remember that amateur comedy is being performed by a ↩︎

  3. https://twitter.com/FullFact/status/1567460294250594304, https://fullfact.org/blog/2017/nov/how-we-live-factcheck-prime-ministers-questions/ ↩︎