Achieving political consensus with AI

How could we accelerate the reform of democracy and especially achieve a wider political consensus using AI? The answer might be in the approach developed by an American company Polis (pol.is) and applied in several countries such as Canada, New Zealand or Taiwan. The concept is explained by the inventors on YouTube video. Broadly it seems to have solved the following problem in democracy. Imagine that MPs discuss a complex new legislation or that you have signed an on-line petition. Many countries now apply this form of political engagement. For example, in Britain, there must be 100,000 signatories for a parliamentary petition to be discussed in the Parliament. Polis has a similar starting point. But here the similarities end. In the UK, when you sign a Parliamentary petition it is understood you fully agree with it. Since you cannot modify the petition’s wording in any way, the only other option is to abstain. But it does not have to be like that.

This is where the innovative approach proposed by Polis can help. It starts with an assumption that people need time to understand the implications of the proposed legislation. To enable that understanding, Polis uses a customised version of Facebook. It enables potential signatories to see the initial wording of the petition (max. 280 characters), its all modified versions, comments left by users, and how many people have signed up for each of these versions. A potential signatory can sign one of the existing versions of the proposed legislation or propose his own. He can also leave comments or suggestions for all others to read.

It enables potential signatories to see the initial wording of the petition (max. 280 characters), its all modified versions, comments left by users, and how many people have signed up for each of these versions. A potential signatory can sign one of the existing versions of the proposed legislation or propose his own. He can also leave comments or suggestions for all others to read.

It enables potential signatories to see the initial wording of the petition (max. 280 characters), its all modified versions, comments left by users, and how many people have signed up for each of these versions. A potential signatory can sign one of the existing versions of the proposed legislation or propose his own. He can also leave comments or suggestions for all others to read.

Leaving comments is a crucial part of the Polis system, which is an AI-powered conversation platform. Comments left by signatories on a petition create an indirect “conversation”. The AI machine learning methods uncover patterns in real-time, mapping out the entire conversation by visualizing correlations between opinions and participants, sorting participants into opinion groups, and surfacing areas of consensus and divisiveness. Therefore, a signatory, can after some time, assess the changes in an on-line visual representation of various groups’ support for each of the variants of the legislation. He can then switch his support for another version of the petition. In this way, the most preferred version of the petition will be chosen through a consensus and compromise.

One of the countries that has experimented with the Polis system for over 4 years is Taiwan, which has built a dedicated on-line platform. In one of the platform’s early successes, the topic at issue was how to regulate the Uber company, which had, as in many places around the world, run into fierce opposition from local taxi drivers. As new people joined the online debate, they were shown and asked to vote on comments that ranged from calls to ban Uber, subject it to strict regulation, or calls to let the market decide. Within a few days, the voting had coalesced to define two groups, one pro-Uber and one, about twice as large, anti-Uber. But then the magic happened. As the groups sought to attract more supporters, their members started posting comments on matters that everyone could agree were important, such as rider safety and liability insurance. Gradually, they refined them and gained more votes. The end result was a set of seven comments that enjoyed almost universal approval, containing such recommendations as “The government should set up a fair regulatory regime,” or “Private passenger vehicles should be registered”. The divide between pro- and anti-Uber camps had been replaced by consensus on how to create a level playing field for Uber and the taxi firms, protect consumers, and create more competition.

Overall, a system like Polis could solve several problems in modern democracies, not only related to petitions, such as:

  1. It raises significantly the participation in politics giving people real influence on the outcome of a proposed legislation
  2. It achieves consensus on a proposed legislation by constantly redefining the initial wording of the proposed new law. This allows people to adapt gradually their views to the views of the largest group and therefore, achieving much broader political consent through a better understanding of the issue, since the initial wording of the proposed legislation changes to reflect the views of a growing majority.
  3. It is the best antidote against fake news. People learn from each other, and if they find themselves in a really small minority, they can then gather more information to understand the issues better and perhaps change their mind.
  4. It can replace referenda and make the decisions such as on Brexit achieved with a far wider consensus.

An approach like this may play an important role in Consensual Presidential Democracy, since it allows for blending of a representative and a direct democracy. However, to have a real impact on a country’s politics, it needs to be accompanied by a legally binding procedure, such as in Taiwan or Canada, so that a parliament discusses and eventually implements the legislation proposed in a petition.