How does Pol.is support online public engagements?

*Note this post was co-written with my policy counterpart on the project Deepika Grover

This post is the second in a series of posts focused on the topic of leveraging technology and more specifically Artificial Intelligence (AI) in the space of online public engagements. Feel free to jump to any other post as they become available.

• How does AI and public engagement fit together?
• The silent majority is louder than usual
• Supporting multiple languages in participant-driven conversations
• Lessons learned while harnessing advanced digital capabilities within the government of Canada

One of the biggest hurdles of interacting with the public online is ‘scale’. How might we tap into millions of minds without breaking the bank? Multiple points of view are welcome and needed when it comes to policies designed to affect millions of Canadians. Even when — especially when — those points of view come into conflict with one another. Different outlooks often bring forward fresh, innovative perspectives. It’s increasingly important to tease out outliers to round out our understanding of a domain or ecosystem, as well as pave the way for empathy and understanding between all players in a particular space.

Engaging in a nuanced conversation without having it devolve into an endless debate has been a key challenge on the road to open policymaking. Certain viewpoints tend to dominate public dialogue, while many others perspectives remain virtually unheard. So how might the Government of Canada hold meaningful conversations with its diverse stakeholders?

Canadian Heritage (PCH) has been piloting Pol.is, an open-source software, using an algorithm called ‘principle component analysis’ designed to conduct interactive online public engagements. The Department’s first initiative with this software sought to better understand the issues emerging in Canada’s visual arts marketplace.

Screen capture of the Visual Arts Marketplace Engagement Initiative that leveraged Pol.is technology. Here you can see an example of the machine translation of a user-contributed statement.

Pol.is is a platform that combines the ease of Twitter with a capacity to “vote” on tweet-style statements. The accumulation of 140-character statements — submitted by participants who identify key positions, contribute their own experiences, express sentiments, etc. — helps scale up online conversations and provides data on patterns, convergences and major points of contention. Rather than reducing complex issues to simplistic views, Pol.is creates a map for government and its stakeholders that starts to untangle complex policy challenges.

Here you can see a real-time visualization of all submitted statements arranged in a bee swarm plot which demonstrates the group dynamics on a spectrum where consensus pools to the left and divisiveness pools to the right.

This allows Canadians to engage with each other through statements that are restricted to 140 characters in length. Participants can add statements and “vote” (i.e. agree, disagree, pass) on statements submitted by others. In this way, a conversation is built statement by statement. A conversation is deepened, vote by vote.

This leads to a conversation that is participant-driven and not overpowered by any single voice or perspective. It also allows conversations to explore many directions — a way to include everyone’s point of view. Canadians are anonymous and can come back to the conversation when they want and see new statements that have since appeared since their last visit. The conversation is monitored only to avoid off-topic statements, including
derogatory or for commercial purposes.

This is an image of some of the seed statements that were metadata driven, more focused on who was participating in the conversation rather than what the conversation was about. You can also see the algorithm at work within Pol.is here as it pools like-minded participants into opinion/interest groups allowing you to further visualize the group dynamics at play in the ecosystem.

As statements are added to the platform, the list of viewpoints grows, creating a measurable conversation. Pol.is collects both qualitative data (i.e. self-contained statements that can indicate the breadth of different experiences and sentiments) and quantitative data (i.e. individual and aggregated votes against all statements). By fostering an environment where people and positions are detached from the problem itself and removing direct confrontation from participant interaction, Pol.is brings forward data from which new insights can emerge. This, combined with design processes might, ultimately, empower open policy-making communities to visualize data in ways that might move dialogue beyond known impasses.