The Crowd Wisdom Project uses Polis, the AI-driven consensus-finding software, to find communities' complex views

Harrogate is the site for Pol.is’s latest consensus-taking. From Wikipedia

We have a strong democratic innovation strand here - and a big part of that has been the community polling software Pol.is. This is how we explained it in a 2020 blog:

What’s unique about Pol.is, compared to other opinion-polling exercises, is that 1) the participants themselves largely shape the questions around the big topics they’re exploring. And 2) a set of algorithms and AI is able to sift through hundreds of thousands of responses, and identifies groups of likes and dislikes about these same questions.

The results are then displayed in ways that help everyone - both the commissioners and the participants - see unusual and unexpected consensus on issues. Bottom-up phrasing of concerns often articulate issues beyond the standard opinion poll questions.

An open-source project (indeed created under copyleft conditions), Pol.is is most known for powering vTaiwan’s cultures of civic digital participation in the policy making of the country (again, well covered here).

But Pol.is is being deployed at a much more local level in UK also, to grapple directly with issues about resources and infrastructure. Something that contains both the creative self-expression of social media, and the rigour of polling technology, in order to show the richness of opinions around an issue.

That’s what seems to have panned out with Pol.is advocates The Crowd Wisdom Project, as they took their method to the question of a projected new “gateway” in the area around Harrogate station. Very local politics… but a thorny question for locals for years.

In its findings article, the community newspaper the Harrogate Informer noted that:

The Gateway poll was a success, in that residents actively engaged in a productive tone, far less hostile than on social media, despite the poll being anonymous. The poll results show that there are two strongly opinionated groups.

[There were] those for the scheme and those against it, though there are areas of consensus. Those for the scheme believe that the new proposals will make the town more attractive and increase active travel.

Those against it worry about increasingly congested roads with little utilisation of the space sacrificed.

Both sides agree that better bus provision will reduce traffic and that the current scheme does not have enough flora which is not in keeping with the rest of the town.

The scheme was approved by the council, but is still being reported as a divisive issue by the local media. But the use of Pol.is in the process at least shows a map of opinions, with positions and groups overlapping, and a rich archive of statements generated by participants themselves, which they organically responded to and voted on.

The following images are taken from the full Crowd Wisdom report on Pol.is. Below is a gif of using the statements map from the software, where they are grouped along a graph of greater consensus <-> greater division. Taken a little slower than the gif, you can see concerns defined by locals around this road scheme expressed in often pointed and salty language. But if you point the cursor throughout the bulges at the consensus end, you see a range of practical and aspirational pointer—very useful for both community members and developers in showing the culture of the debate around the issue.

Pol.is also delivers a table on “Majority” opinions (below) - where 60% of participants either agree or disagree on a statement that others have generated. What’s interesting is that many of these consensus statements express a wider vision than just this traffic improvement - questions about the wider economic policy for the town, or the mismatch between the building of houses but insufficient infrastructure to support this expansion.

Looking into the Crowd Wisdom Project’s FAQ is also of great interest. It shows their idealism and integrity around community democracy, but it also shows how the AI/Machine Learning that reads, collates and groups these statements itself needs some assistance from diligent humans:

Once a Polis survey is live, during the duration of its existence, voters can suggest statements for others to vote on. Our job as moderators is to only allow appropriate statements to be voted on. Often, voters ask questions, rather than make statements. This is a frequent occurrence, even when we make it clear that additional questions ought not to be asked as statements.

Moderation also ensures that bad language and confidential information etc cannot be voted upon. There is skill and time needed to effectively moderate. The more voters, the more statements are usually submitted for moderation. Therefore, the largest surveys are the most time intensive for us.

This is important - Pol.is isn’t some computational “black box” into which data is fed and consensus is ticker-taped out. It’s a democratic tool, which speeds up some processes but still requires the opposite of “Garbage in, garbage out” in terms of the data it’s working on.

There’s a list of case studies here from Crowd Wisdom. A/UK hopes to be working with the Pol.is software on a CAN project later this year.