In discussing the ethics of artificial intelligence (AI) we sometimes get bogged down in the problems and pitfalls of biased algorithms and messy data, overlooking the more hopeful opportunities made possible by the technology. That’s not to say the risks are trivial, merely that they’re worthy of ethical scrutiny precisely so that society may reap the attendant benefits.

One of those benefits is in making government services more efficient. Scanning the headlines, you can find stories about how AI is being used to detect government fraud, inform welfare decisions, and automate mail sorting processes, among other uses.

But for Beth Simone Noveck, core faculty member at the Institute for Experiential AI (EAI) and director of the Burnes Center for Social Change, there’s an aspect to this story that deserves more attention: how AI can be used to strengthen democratic participation. More than merely making existing systems more efficient, AI has the power to elucidate public opinion, connect citizens with institutions, and augment what some researchers refer to as “collective intelligence”—the social processes by which large groups make decisions.

“The intersection between artificial intelligence and collective intelligence, I would argue, is a very important interdisciplinary area that requires more content, more attention, and more focus,” Noveck said in the kick-off to the Fall Seminar Series, a lecture series hosted by EAI that features some of the brightest minds in AI. 

Much of the promise has to do with crowdsourcing citizen inputs to inform legislative procedures: for example, Helsinki’s platform for tracking carbon reduction or Louisville’s citizen science initiative to monitor asthma inhaler usage. In her talk, Noveck sought to broaden the scope even further, asking profound questions about whether AI, with its superhuman ability to analyze data, could be used to re-imagine democracy itself—moving beyond mere voting and polling.

The core challenge here is one of scale. Institutions are outdated and public trust in them has eroded in recent years. But AI has the capacity to institutionalize collective intelligence. Specifically, Noveck sees six ways AI provides tools that can help foster democratic engagement.

1. Deduplication - Crowdsourcing may be democratic in spirit, but at the scale of a democracy it’s not very feasible. One opportunity AI provides is in “deduplicating” the millions of public comments government agencies receive when considering new rules, allowing policymakers to get a distilled view of where public opinion is actually falling.
   
   
2. Summarization - Similar to the deduplication process, text summarization tools allow agencies to condense findings, generate keywords, and synthesize large volumes of content, helping make public comments more digestible for lawmakers.
   
   
3. Sentiment analysis - AI techniques are being deployed to analyze real-time, in-person conversations—the kinds of conversations common at the local government level. These analytic tools allow officials to extract ideas, comments, and sentiments from conversations, which for many citizens are more accessible than written engagements.
   
   
4. Annotation - Text annotation platforms like Rap Genius, Hypothesis, and Factmata add layers of transparency to deliberative processes. A good example is the German government, which in 2018 crowdsourced public annotation for a draft version of its federal strategy on AI.
   
   
5. Sorting and organizing - Soliciting public comment is meaningless without a way to visualize, sort, or organize feedback into digestible information. The Citizen Lab in Brussels, for example, is working with over a hundred cities around the globe to model and cluster citizen feedback. The topic modeling system works both ways, giving government agencies as well as citizens a visual representation of public opinion.
   
   
6. Collaboration and coordination - From an overhead view, AI helps public servants and citizens consider new ways of organizing collective intelligence that move beyond the ballot box. An example here is Unanimous AI, a company that analyzes swarm intelligence to aid organizations in forecasting and predicting public behavior.

With all of these applications, the idea of AI in governance is to help broaden democratic engagement. As Noveck explained at the beginning of her talk, technological innovation often progresses in ways that disregard or even harm social bonds. But it doesn’t have to be that way. The ethical implications of AI behoove us to seek new and diverse inputs for technologies with social impact, underscoring the need for oversight panels like EAI’s new AI Ethics Advisory Board. Efforts like these may end up providing just the kind of landscape needed to question and, ultimately, reimagine the possibilities of democracy.

Catch the full replay of Beth’s talk here and learn more about her research here.