Could artificial intelligence benefit democracy?

  • Published
Campaign Lab hack dayImage source, Campaign Lab
Image caption,
Campaign Lab runs regular hack days to work on new campaigning tools

Concern about artificial intelligence (AI) wrecking elections and threatening democracy has reached fever pitch.

Each week sees a new set of warnings about the potential impact of AI-generated deepfakes - realistic video and audio of politicians saying things they never said - spreading confusion and mistrust among the voting public.

And in the UK, regulators, security services and government are battling to protect this year's general election from malign foreign interference.

Less attention has been given to the possible benefits of AI.

But a lot of work is going on, often below the radar, to try to harness its power in ways that might enhance democracy rather than destroy it.

"While this technology does pose some important risks in terms of disinformation, it also offers some significant opportunities for campaigns, which we can't ignore," Hannah O'Rourke, co-founder of Campaign Lab, a left-leaning network of tech volunteers, says.

"Like all technology, what matters is how AI is actually implemented.

"Its impact will be felt in the way campaigners actually use it."

Worrying tendency

Among other things, Campaign Lab runs training courses for Labour and Liberal Democrat campaigners on how to use ChatGPT (Chat Generative Pre-trained Transformer) to create the first draft of election leaflets.

It reminds them to edit the final product carefully, though, as large language models (LLMs) such as ChatGPT have a worrying tendency to "hallucinate" or make things up.

The group is also experimenting with chatbots to help train canvassers to have more engaging conversations on the doorstep.

AI is already embedded in everyday programs, from Microsoft Outlook to Adobe Photoshop, Ms O'Rourke says, so why not use it in a responsible way to free up time for more face-to-face campaigning?

"AI-generated misinformation simply doesn't work when politicians have proper relationships with the voters and the communities they represent, when they actually deliver on what they promise.

"You can't fake that a local park someone passes everyday has been cleaned up, just as you can't fake a conversation on a doorstep."

Image source, Hannah O'Rourke
Image caption,
Hannah O'Rourke says real doorstep conversations cannot be faked

Conservative-supporting AI expert Joe Reeve is another young political campaigner convinced the new technology can transform things for the better.

He runs Future London, a community of "techno optimists" who use AI to seek answers to big questions such as "Why can't I buy a house?" and, crucially, "Where's my robot butler?"

In 2020, Mr Reeve founded Tory Techs, partly as a right-wing response to Campaign Lab.

The group has run programming sessions and explored how to use AI to hone Tory campaign messages but, Mr Reeve says, it now "mostly focuses on speaking with MPs in more private and safe spaces to help coach politicians on what AI means and how it can be a positive force".

"Technology has an opportunity to make the world a lot better for a lot of people and that is regardless of politics," he tells BBC News.

'New fangled'

The UK has already seen its first - and maybe the world's first - AI-powered election candidate.

Andrew Gray was roundly rejected by voters in July's Selby and Ainsty by-election, gaining just 99 votes.

It was a bruising experience for the former lawyer, who stood as an independent candidate "powered by AI", with hopes of causing a political earthquake.

But voters in the rural North Yorkshire constituency were not ready for such a "new fangled" concept, he says, and were often just confused by it.

"When constituents messaged me, they often thought I was a bot."

Mr Gray's campaign was based on Polis, an AI-powered tool that allows groups with widely different opinions to reach a consensus through votes and discussion.

Image source, Andrew Gray
Image caption,
Andrew Gray: The UK's first AI-powered election candidate

Polis first appeared on the political radar when it was used by Taiwan's government to settle controversial questions, such as the legality of e-scooters.

Mr Gray used it to crowdsource a "people's manifesto" for Selby and Ainsty and planned to use it to consult constituents on which way to vote in the House of Commons.

He remains a firm believer in the power of Polis to bridge political divides and bring people together but says it may be better suited to solving local controversies, rather than tackling major national issues at Westminster, at least for now.

Mr Gray's not-for-profit organisation, Crowd Wisdom Project, has helped local councils, such as Wandsworth, in south London, which last year held a consultation on clean air, using Polis, to formulate policies.

Critics will say Polis and other AI-assisted decision-making programs being marketed to local authorities are still open to being manipulated to produce results the results politicians want, like the bogus consultation exercises beloved by an earlier generation of leaders.

But the sheer power of AI technology - and its ability to crunch vast quantities of data - promises to offer far deeper insights into what constituents and voters are thinking and feeling.

'Synthetic polls'

Joe Twyman, co-founder of polling company Deltapoll, says sophisticated "regression" polls - of the kind YouGov produced recently predicting a Labour general election landslide - have been made possible only by AI.

The average regression poll has between 6,000 and 10,000 respondents, according to the British Polling Council.

Deltapoll is working with tech start-up Bombe, which uses AI to analyse polling data collected from smaller samples.

And Mr Twyman is quick to dismiss the idea LLMs could be used to simulate voters.

In 2022, US researchers claimed the results were indistinguishable from real humans responding to a poll.

But Mr Twyman says people are more complex and unpredictable than machines trained on existing data and "synthetic polls" would have "no value".

Although, he admits it is impossible to know exactly which of the growing number of AI tools being offered to political campaigns will catch on.

Wreak havoc

"Some of them will be incredibly powerful, some of them will be complete nonsense promoted by snake-oil salesmen exploiting the next big thing and there will be some things we will never hear about that could potentially make a big difference," Mr Twyman says.

Most tech optimists are worried by the potential of AI-powered disinformation to wreak havoc at elections, particularly as 2024 will see an estimated 49% of the world's population go to the polls.

And the benefits to democracy from this powerful and rapidly evolving technology can seem marginal, when weighed against the damage it could do.

But no-one can predict how it will play out - the only certainty is AI is here to stay and politics will never be the same again.