Introduction

Today I made a thing.

A spontaneous demo opportunity

Last night I was cycling home and I saw a tweet, a message, a WhatsApp message from Josh saying, does anyone want to do a demo today?

And I'd just been at a democracy meetup and talking about system assemblies and because I used to work in civic tech and like voter engagement and things and

From civic tech to AI for good

Then after that I was invited to a round table on AI for good and so I went up and spoke to Meta and Wikimedia Foundation and all sorts of finance people and investors and we were just like shooting the shit on AI for good basically and what can we do with it and what do we need to do and what do we need to change and how do we get there and blah blah blah blah blah.

Anyway, there was free wine at that and so

Morning resolve

1105 I replied all right give it a shot and see what I can make and so I woke up this morning at 7 a.m.

And loaded up my list of list of like side hustle throwaway ideas that I keep Shortlisted a bunch of them thought about them for a while and

I mocked up this one.

From ideas to prototypes

An eToro for policy and liquid democracy

I was thinking about a sort of eToro for policy ideas, thinking that it would be kind of cool to make it so that you could throw out a policy idea and then vote on it.

Who's heard of liquid democracy?

Anyone heard of liquid democracy?

What is liquid democracy?

There's a really fun concept in voting.

So at the moment, you give your vote once every five years to your local MP, and they vote on everything for you.

That's great, unless you disagree with them on a couple of things.

Well, it's great.

if you voted for them, chances are, there's a pretty high chance that your MP you'll never have voted for.

But notwithstanding, even if you did vote for them, you might disagree with them on a couple of things.

Selective delegation of votes

So you might say, look, generally you're okay, but education, you haven't got a clue.

I trust my friend Harry, who's a teacher on education and is all over the policy and knows what to do and is looking to the future.

I trust Harry with my vote.

And so you can delegate bits of your vote sort of based on tags, say, like health, I want that person.

Education, I want that person.

And so you can delegate your vote selectively to different people.

And then they can delegate

those aggregated votes to other different people and so on and so on.

Emerging trust networks

So you get this sort of hierarchy of trust and agency emerging, plus you have this sort of subsidiarity level of engagement.

So everybody's really highly engaged on their vote.

They're voting on the things that they care about and they know most about, and they're voting for people they trust to make decisions for them.

Anyway, so I was thinking you could have this sort of eToro idea where you can copy different people who you trust for different votes and bundles of votes in the same way that with eToro you can follow different investors and follow their investment movements.

And thinking it'd be kind of fun to gamify that.

Scoping and practicality

But I decided actually it needed a bit more thinking through and it needed more than 10 hours to do it justice.

Previous project: Open Group Map

I made this about a year ago, which was Open Group Map.

It's there, and it's live, but it's a bit boring.

There's not a lot to it.

It's a way for people to very quickly put their location on a map, so you might have a bunch of friends in a WhatsApp group and go, where the hell is everybody these days?

Oh, I live here.

I live here.

I live here.

Cool.

10 a.m.

Building with AI tools

Kicking off the workday

rolled around, and I thought I'd better do some work, so I fired up a cursor tab,

or a CLI, and fired up Claude.

Who uses Claude code?

If you're not using Claude code, one day you, well, it's really fun.

You get a lot of power for a very small price, and you can just let it rip on tasks.

A climate risk analysis tool in a day

And so what I did for my day job today was that I made a climate risk analysis forecasting tool for our transition risks team.

What that means is I took a whole bunch of data, stuck it in a vector database, reports and all sorts of things, forecasts, like what's the future of the EV industry?

Took about a gig of data, stuck it in a vector database,

then queried it to say, cool, generate some reports for this industry, this GICS code.

And then it would generate a context.

Then it would generate a whole bunch of items, basically, specific risks.

And then you can generate a risk for a specific company, a specific client.

We used to sell these reports.

Well, I think we still do sell these reports for like 10 grand each.

and you can now whip them out.

It's really easy.

Rethinking strategy

Obviously means we need to change our strategy and rethink how we can do that, because if I can do this on the side in a day, it's also pretty easy.

The problem with petitions

So, back to working out what to build.

I'd had this itch to scratch for

Civic tech background and the itch to scratch

Well, for ages, because like I said, I used to work in civic tech and digital democracy.

Nuance is missing in petitions

And it really pissed me off that petitions lack nuance.

Like you can vote, you can say, yes, I agree with the thing, or no, I disagree with the thing.

But you can't say both of those.

You have to essentially...

agree with whatever the headline of the statement is.

And it lacks nuance.

It belittles us as voters.

We are more capable and, frankly, should be more responsible than just going, yeah, I want that thing, because that doesn't get us anywhere.

It's good to engage, good to express that sentiment, but

Collaboration and coordination at scale

Come on, we're educated adults and we're all trying to get through and solve problems together.

How can we do that more effectively?

How can we do collaboration, coordination at scale?

And it's something that humans are not very good at.

We can just about get a piano downstairs, but if we tried to choose a film to watch tonight, it would be horrible, right?

We'd have so many different choices and different preferences.

So how can you do that?

Why AI can help

Now, AI is brilliant.

at engaging individuals where they're at and taking them on a journey and showing them and so on.

The demo: nuanced voting on bills

So I decided to build a thing.

Sourcing and preparing a bill

And so here is a government bill.

It's really hard finding government bills, like the actual paper that's being voted on.

So many of them aren't published in the way that you would like them to be published.

But anyway, in the time that I had today, I found this one.

It's about animal use and medical research.

And naturally, it's online as a PDF rather than lovely HTML, but so it goes.

So I took that.

And so that's got details, but like I said, we're not allowed to vote on it.

This is a petition.

They lack nuance.

They make us look like a mob.

So what if we could have something like this that says, and excuse the mock-up, right?

No one's going to vote on the title, but it's a quick mock-up, that says, OK, I want to vote on section one

1.2 second clause of that.

I wanna vote on this one down here about requiring the use of non-animal methods.

Granular voting on sections and clauses

And so, let's have a site where you can post up a bill, you can share a deep link into a thing, and you can vote on specific issues.

Live walkthrough

So, hang on, here comes a demo.

You can still see stuff, cool.

So I set about making this thing.

So let's take this Animals in Medical Research bill.

Let's make a new document.

Let's paste it in there.

I made a thing just to tidy up the markdown, turn it into a nice markdown documentation.

Give it a second.

There we go.

I'm going to call it so that it's easy to find.

Save that.

And what this does is it goes through and it breaks up these into its component parts.

Let's just make it public and say logins not required and view the public page.

Cool.

I can now go down here and with my cursor keys or swiping on a mobile phone,

I can go down, and I can go, like, look at that one.

Don't like the look of that one.

Don't like the look of that one.

That's cool.

That's cool.

That's not cool, and so on.

Yeah?

So really quick and easy to engage with.

And then, hang on.

Where's one I made earlier?

Hold on.

Here we go.

It's not, again, this is just some fake votes.

but you end up with a nice view and a representation of where you can quickly scan a document and see who's in favor, who's against different items on this list.

Insights, demographics, and analysis

What we could then do is add things like demographics onto this.

So we could say, okay, well, let's click into this and, oh, that's funky.

So I made it so you could add comments on there, but you could add demographics.

So you could say,

When you register, you can say, hey, I'm Ed, I live at this postcode, I'm in this constituency, you know, I'm male, whatever is important.

And then in here, you can have an analysis of who's voting, who's not voting, how representative it is, how representative it isn't, how much that person's vote is...

consistent with other petitions that they voted on.

All sorts of analysis that you can do with the data once you've got it.

Beyond public good: general use cases

So now that's kind of a public good use, but obviously it's also really easy to then do it for anything else as well.

So if you want to have a group of friends and choose out where to go on holiday or what film should we watch tonight, you can then go, OK,

Here's a list of films.

Here's a list of options.

Just go through and vote on them.

Wrapping up

So it's an easy public voting tool that has an array of uses, and I started that at about 10 a.m.

this morning.

Where to find it

And it's there, and it's live now in its messy form at ministryforthefuture.org.

And there we go.

Closing and Q&A

That's my talk.

What next?

Any questions?

Do we want to do that?