Manifesto Project: a word about AI use
Developing a fully comprehensive manifesto would ordinarily be beyond the means of any one person. Having been a political blogger for approaching twenty years, I was in good position to set out a basic framework and add to it, but an effort from a single individual was always going to lack the necessary depth.
It is my view that political parties should set out a basic framework then acquire the necessary expertise to develop it, preferably pulling in as much real world experience as possible. That is certainly within the means on parties such as Restore Britain and Reform UK. You have to as serious questions about why they haven’t. If they’re only just cranking into gear, they’ve already left it too late.
What changes the game, slightly, is the emergence of AI. There are now an array of powerful writing and research tools that can augment this kind of work. It comes with the caveat that none of what it produces is trustworthy. Luckily, there are certain tells. Once you play with the technology for a while, you soon pick up an instinct for what is AI generated slop. The last thing we need is more AI padding.
My original approach was to give AI a hard swerve, which was absolutely the correct approach in setting out the basic framework for this manifesto. If you want to to say what you mean then you have to do the thinking yourself. There is no substitute for having organised your own thoughts through the laborious process of traditional writing.
Where AI comes in useful is to offer critiques where expert consultation is not available – especially in the middle of the night when I do most of my thinking. I’ve found it useful in identifying gaps and suggesting alternative approaches – especially where I lack expertise.
Having set out my own policy objectives and priorities, tools such as Claude AI can bring them up to standard. In my case, I then take the output and feed it through Grok to improve the readability. They each have their own merits, but both require close editorial supervision. In many cases, I end up doing manual final edits and tweaks. That’s going to be a slow process over the next month or so.
Here there is balance to be struck between depth and verbosity. There is a middle ground between the pointless empty slogans of Restore and Reform UK, and more weighty policy papers from specialist think tanks. Political parties should make greater use of think tanks. They provide and entirely free resource, and providing intellectual direction for political parties is a big part of what they are actually for. There is zero point in going into competition with them. They have the funding and the expertise. Why not utilise it? In my case, I’ve turn up some interesting approaches just by training AI tools on outfits such as the Centre for Social Justice.
I’ve been banging the drum for three years now for parties to have detailed manifestos – proving they have thought through what they’re doing. Restore Britain has responded to this by producing bloated and extruded verbal material to give the outward appearance of effort, but it still amounts to the same tropes and assumptions.
The problem with this is that most won’t even bother to read it and those who do will be unconvinced. Document depth goes towards providing reassurance that they’ve properly understood the issues, but having built their knowledge base on their own untested assumptions, the end product is never going to be a decent standard.
In fact, one of the reasons I know populist slop when I see it is because I have in the past been guilty of the same. Like most on the right, I took it as an article of faith that were must leave the ECHR, and that it was a prerequisite for wider reforms. An open minded investigation changed my mind. The same goes for the dogma on purging quangos and reducing the size of the state. It turns out the last thig we need is another top-down reorganisation. What is needed is better intellectual direction, leadership, and accountability mechanisms. there is no substitute for original thinking.
Having now completed an AI augmentation sweep of the Manifesto Project, I think it’s in a lot better shape than it was, with some very necessary improvements to flow and clarity, with a few new ideas, but it is still essentially the same and built along the same principles. There are still gaps in it, and the improvement process never stops. Policy development is a continuous improvement process, and must take into account current developments and the present day circumstances. We do not have the luxury of developing year zero approaches for an ideal world.
With AI becoming more powerful all the time, there is a temptation to lean fully into it, but that then breaks the contract. You can expect the reader to invest intellectual effort into reading something when none has gone into producing it.
Moreover, if you didn’t write something, you won’t have the intimate relationship with that’s written in your name and you won’t be able to defend it or take pride in it. That, ultimately, is what’s missing from the populist right. Everything they do is done with the bare minimum thought and effort, and it shows. They take no pride in what they do, assuming that because slop passes with their fanbase, everyone else will be as satisfied as they are. While AI certainly has its uses, there is no substitute for people who care enough to know what they’re talking about.
On a final note, countless hours have now gone into the Manifesto Project. While nobody will avail themselves of it, not least because of who wrote it, a body of ideas has inherent worth. It will no doubt inform the commentary of this Substack as we creep closer towards an election, where again we will see parties of the right making basic tactical errors simply because they haven’t invested in building an intellectual foundations.
I assume this kind of insight is valuable to you, otherwise you wouldn’t be reading this. On that basis, I humbly ask you to consider a paid subscription (or donate here). There are now perverse incentives to produce rage bait slop over on X, which I am not inclined to do - but that’s proving to be an expensive choice.