Growing discontent and drop-outs who do not feel heard: artificial intelligence can offer solace. It can help make decision-making much more attentive and complete, as professor of Hybrid Intelligence Catholijn Jonker and innovation consultant Ilyaz Nasrullah show.
Let’s first set the record straight about democracy
Computer scientists Catholijn Jonker and Ilyaz Nasrullah are completely at ease with algorithms and data systems, but also look with keen eyes at how we live together and what democracy actually is. Nasrullah: ‘That the majority decides what happens doesn’t constitute the whole idea of a democracy. Besides the rule of law that protects interests, it’s also essential that you’re willing to listen to each other.'
Jonker sees it exactly the same way. ‘You have to understand each other before you come to a decision. Having an eye for the norms and values of others, wanting to care for others: what are they struggling with, what’s blocking them from saying yes to a proposal? As long as you don’t understand each other, the other is an angry person who doesn’t want to see your interests.'
As long as you don’t understand each other, the other is just an angry person."
She cites corona as an example. 'With the strict corona measures, different interests were at play. These had a big impact on companies, so they preferred to abandon them quickly. On the other hand, there was the fear people had about their health. Who was at fault isn’t interesting here, one way or another we had to decide something together. It really is important, though, to feel that your interest has been heard. I want to help improve that with AI. That’s why I’m participating in the Centre for BOLD Cities.'
An algorithm can hear all voices
Algorithms offer unprecedented opportunities to improve decision-making. Jonker: 'Imagine this case: inhabitants of a residential area can vote on the allocation of three hundred thousand euros to renovate the neighbourhood. People have submitted proposals and now everyone can rank those proposals and also record why some proposal is preferred or not.'
You cannot do anything with that, you may say. 'When 20,000 people vote, it's impossible for an official to plough through all those votes. You can only count them and let the majority have their way. Then the money goes to the best little story.' An algorithm can, however, plough through all the votes and explanations. 'Suppose a project narrowly fails to get a majority. Because a few things weren't clear, or there are some concerns, or it didn't fit the budget. An algorithm can bring that to the surface, and so a conversation can start about adjusting the proposal.' In a second round of voting, a small effort with such an algorithm, the proposals that people truly want can then win.
Such algorithms exist and are already in use
Most learning algorithms are aimed at bringing to the surface ‘more or less the average opinion’, says Jonker. 'But the feeling people get after such a consultation is: you say you listen to me, but you don't act on it! You can also train AI to measure diversity instead, and to cluster values, concerns and motivations.' A silent majority can then add nuance. This is already happening with PWE: the Participatory Value Evaluation, developed by scientists at TU Delft, among others.
Three universities and the National Institute of Public Health (RIVM) thus held the corona consultation in spring 2020, during the first corona lockdown. The 30,000 participants were shown eight options to relax measures, with their pros and cons, on their phones or tablets. As an additional condition, the pressure on the healthcare system could not be allowed to increase by more than 50%. Participants were also given information on the effects of each option on the death toll and, for instance, the decrease in the number of households suffering a long-term loss of income. Participants recommended which measures they thought should be relaxed, after which they could justify their choices. They could also contribute ideas for possible other relaxations. An algorithm analysed the results to understand people's values, motivations and concerns. In March 2023, parliament launched the National Climate Consultation, which also works this way. At the end of 2020, the PWE researchers at TU Delft founded a startup around the method: Populytics.
Always consider: what are we doing it for?
Nasrullah finds the main question in technology issues so obvious that he is almost ashamed to pose it. Yet, he often finds that this very question is not asked. 'What is the purpose of your technology? You can use AI to strengthen the strongest voice or to bring out the weaker voice. The technology choice you make comes only after you’ve answered your purpose question.'
Jonker: 'Exactly, that is super important! You want to do justice to the population, be representative and inclusive and balance all interests as well as possible.' Nasrullah: 'Technologists are too quickly given the task of just going and building what they come up with. For instance, there was the useful and successful system that could measure coronavirus in sewage, I wrote about it in one of my columns. This is where you know what is being measured and what it is good for. But then plans arose to extract information about obesity, diabetes and certain cancers from sewage water as well. Now what is that good for? Will there be a fat-people alert? That's useless data hunger, exactly the smart city we don't want.'
If you collect rebuttals, you have to act on them, too."
Jonker: ‘And if you collect rebuttals, you have to act on them. So that people feel you’ve heard them. How you organise that is a complex socio-technical problem with legal, technical, practical and social aspects.'
We can take our cue from Taiwan
The very young democracy of Taiwan is leading the way in using technology to help govern, both experts say. The country has a minister for digital affairs: Audrey Tang, one of the best software programmers in the world, according to Nasrullah. 'While in our country, citizens' councils are on the rise, a technological variant of this has been developed in Taiwan.
The online-offline consultation process vTaiwan involved people's representatives, scientists and other experts, businesses, civil organisations and citizens. The aim was to reach legitimate - that is, genuinely consensus-based - decisions. The process went through the stages of proposal, opinion, reflection and legislation.' At each stage, algorithms ensured that diversity or rather consensus was brought to the fore. Now vTaiwan is part of the platform Join, about which MIT Technology Review writes: 'A simple but ingenious system that allows Taiwan to crowdsource on legislation.'
'There’s a huge willingness to experiment in Taiwan,' says Nasrullah. A minister for digital affairs seems great for the Netherlands, too; he sees opportunities for innovation in our democracy. ‘It’s a great pity that there are very few people in Dutch parliament who really know what can and can’t be done with technology. I’d really like to see all political parties put an expert on IT high on the electoral list, because technology substantially affects the shape of our society.'
Catholijn Jonker is professor of Interactive Intelligence at TU Delft and is affiliated with the Leyden Institute of Advanced Computer Science (LIACS). As a computer scientist, she is working on the embedding of artificial intelligence in the social environment.
Ilyaz Nasrullah is an independent innovation consultant. As a computer scientist, he worked on digital technology for startups, large companies and governments. To provide a humane perspective on digital technology, he writes columns for daily newspaper Trouw.
Text: Rianne Lindhout
