Time for AI to pull up a chair to the negotiating table

Science fiction has sparked decades of debate about whether future wars will be made more deadly by weaponised artificial intelligence. Yet, in the real world, AI is already being harnessed to broker peace.

This idea first emerged four years ago when members of a UN team asked how technology could be used to survey residents of conflict zones about what they wanted from peace agreements. Asking people’s views is important because peace deals and ceasefires are more likely to last if they reflect the views of the whole population, rather than just a few negotiators. But polling people face to face is difficult in a country at war: security requirements make it complex, expensive and time-consuming.

A New York start-up, Remesh, may have found a solution — an AI-assisted poll that can reach more than 1,000 people, engaging them in real-time conversations about the changes they want to see.

The polls were trialled in Libya in October ahead of a permanent ceasefire that ended nearly a decade of instability. The dates of the discussions were publicised on social media and interested participants were invited to access Remesh’s platform anonymously from their mobile or computer. During each dialogue, respondents were asked their priorities for reunifying the country, stabilising the economy and improving living conditions.

To the organisers’ surprise, the initiative attracted such attention that one of the mass polls was broadcast live on a Libyan news channel, with commentators in the studio discussing the responses. “The most important part of it was that people felt engaged,” says Jean El Alam, an official at the UN mission in Libya. “We brought these findings to the [peacemaking] dialogue table and said, this is what your people truly want — not what you’re saying or you think they want . . . This added to the transparency of the process. It gives it more credibility.” The resulting ceasefire deal, though fragile, has held so far. Further dialogues are planned as the peace talks progress.

We brought these findings to the dialogue table and said, this is what your people truly want — not what you’re saying or you think they want

Beyond the UN negotiating table, AI-powered polling platforms are already helping to build consensus on more everyday disagreements. Taiwan has used Pol.is, an open-source technology developed in Seattle, to reach popular agreement on how to regulate Uber, the taxi-hailing app. And local government associations in the US and Canada have used Ethelo, another survey tool, to poll communities on topics such as town planning and transport pricing.

Andrew Konya, co-founder of Remesh, says running the dialogues has been described as being like “talking to a collective superintelligence”. To a journalist who asks questions for a living, this was an enticing prospect.

Remesh put together a mini-poll of 80 Americans, drawn from a variety of backgrounds. Over the course of 45 minutes, I asked them about the US’s place in the world, and whether its international reputation was rising or falling (55 per cent said the latter). The system grouped replies with similar meanings and asked respondents to agree or disagree with selected statements on a spectrum of views. At any point, I could ask the system to analyse responses according to a breakdown of age, gender or political affiliation.

The breakthrough was ditching my planned script and adapting my questions to the answers I was receiving. When several participants said they thought the US should use its power and influence responsibly, I asked what this meant. “Accepting equal footing in the international scene instead of acting like an overgrown bully,” came one reply that was highly rated by fellow respondents.

Online polling tools have obvious pitfalls. They rely on participants having access both to good internet and functioning devices. There is also the risk that without the highest levels of cyber security, these survey platforms are vulnerable to hacking.

Still, Katharina Höne, an expert in the use of technology for diplomacy at the non-profit DiploFoundation, believes these systems have promise. As they develop, she suggests they could be helpful in combating the political and ideological divisions that flourish during online interactions.

“We have social media tools which seem to amplify social echo chambers . . . and focus on scandal and outrage, and on the other hand we have the potential to build tools that allow us to have conversations in a different way,” she says. “Digital technology can come down on both sides, fuelling conflict potentially, but contributing to overcoming it too.”

Helen Warrell is the FT’s defence and security editor

Follow @FTMag on Twitter to find out about our latest stories first.