Democracy AI

We’ve tracked civic, democratic, and ethical AI projects for you since 2016. Explore our wide-ranging collection of examples of people using AI for public interest goals (including work to prevent harms from AI itself).

AI tools

That you can use in your work

AI for Participatory Democracy

Including AI features within digital participation platforms:

“Several participation platforms already use Natural Language Processing (NLP) to summarize large volumes of text submitted by participants and identify topical clusters in conversations. As more people participate on digital platforms, these solutions will become more necessary, since few institutions have the capacity to read millions of submissions. IBM Watson’s Project Debater is an example of how multiple AI models can be integrated to summarize political arguments.
NLP can also be used to categorize the emotional tone of feedback through sentiment analysis, although this technology still isn’t reliably accurate.
A hybrid model that combines AI, crowdsourced filtering, and expert human intelligence could help ensure the success of large-scale engagements.
AI also can flag efforts to subvert a vote with fake levels of support by identifying identically phrased language.
Speech-to-text technology is rapidly evolving to produce more accurate transcripts of phone conversations and meetings.

It is important to note, however, that AI could also have negative effects on participation:

People are already able to use AI to generate significant volumes of text with a simple prompt. For example, today’s AI software can instantly generate a large volume of content to support a simple, initial statement like, “We shouldn’t build a park here.” Such bot-generated text can be used to pollute the participatory process and, in turn, diminish the value of contributions from actual humans.
“Deepfake” technology allows people to generate realistic-seeming images, video, and audio of people who don’t really exist. As this technology becomes more effective and readily available, it will be easier to generate authentic-seeming audio and video clips from fictional personas.”

– Matt Stempeck, Guide to Digital Participation Platforms