Active Listening in the AI Age: Using LLMs to Help People Be Heard

Jigsaw
Jigsaw
Published in
5 min readJun 25, 2024

Since 2017, Jigsaw’s Perspective API has used machine learning to enable more voices to participate in online conversation, by identifying and mitigating toxicity that frequently reduces participation in online debates. This year we are expanding the scope of our work in this area to not only protect online spaces from hate and toxicity, but to enable individuals to more meaningfully participate in conversations online.

Supporting more voices to be part of the conversation

Since its inception, Perspective has been used by more than 10,000 organizations, ranging from small publishers to large social media platforms. Moderators can use Perspective to quickly prioritize and review comments that have been reported and give feedback to commenters who post toxic comments. Perspective now receives over 2 billion requests per day across its 18 supported languages.

Recently, Perspective expanded in scope to add bridging attributes, a new suite of tools to recognize qualities like reasoning, curiosity, and personal stories that correlate with more constructive contributions and help keep conversations going among disparate groups.

The bridging attributes are a first step in a broader shift for Jigsaw in this area of our work. In addition to supporting more voices to be part of the conversation, we’re now expanding our work to enable every voice to be meaningfully heard. We aim to build on the existing foundation of deliberative technology, a set of tools that, in the words of Lisa Schirch, “allows a large-scale exchange of views in an iterative discussion, allowing participants to evolve in their understanding.” Jigsaw will explore how the latest AI technology might enhance and scale these technologies, supporting both conversation participants and community leaders to make sense of and act on the opinions shared.

Together with a growing and diverse set of collaborators, we are starting with three anchor initiatives: a proposed agenda for research and development in these technologies, a new joint initiative with Google DeepMind to explore how sense can be made of large scale conversations, and supporting an open call for proposals with Google.org to help scale social impact initiatives promoting democratic resilience in Europe.

“AI and the Future of the Public Square” Position Paper

In April 2024, Jigsaw brought together experts from across civil society, private sector, governments and academia to explore how AI may be harnessed to improve the future of the public square. The gathering broke into small groups and began by defining the greatest challenges today for the online public square. We dug deeply into the AI-enabled technologies that are driving deliberative tech forward today before spending the afternoon reimagining their future. For example, participants discussed applications of Polis, an open source software application that has already been used in places like Taiwan to gather, analyze, and understand what large groups of people think in their own words. They also discussed Remesh, another tool that looks to generate insights quickly and at scale from large groups, which has been used by the UN to help broker a peace deal in Libya. Participants explored how deliberative technologies like these might scale and allow more voices to strengthen democratic discourse.

Capturing the key insights from the gathering, Jigsaw and 20 co-authors from among the attendees, will soon publish what we collectively see as an agenda for the field of deliberative technology. We are hopeful that the paper will serve as a roadmap, helping to prioritize research, technical development and future investment. Specifically, the paper will identify how advances in large language models (LLMs) pose new opportunities (and challenges) for deliberative tech, bridging systems and community moderation.

Sensemaking for Deliberative Technologies

Deliberative technologies empower the public to collectively explore and express their opinions on topics of interest, giving them an opportunity to be heard and contribute to shaping a deeper, more coherent understanding of public opinion. These explorations can be journalistic, reflecting directly back to the public this deeper understanding of itself. They can also be used as part of formal decision making processes, allowing experts and decision makers to consider a more nuanced understanding of public opinion on the issues before them. These processes tend to avoid zero-sum dynamics by providing space for novel perspectives to emerge and transcend ingrained narratives.

With these platforms becoming more widely used, more ideas can be shared, giving more people access to participate. As conversations scale, however, new challenges emerge, from identifying central themes and areas of commonality to developing concrete plans of action that reflect the will of the broader community. To begin to address these challenges, Jigsaw is partnering with Google DeepMind and Polis, an open source deliberative platform, to explore how LLMs might help make sense of large scale conversations. This might include summarization, visualizations, topic or theme analysis, or new ways to identify areas of emerging consensus and divergence. This sensemaking during and after dialogue is not only helpful for conversation participants to see the full opinion landscape of their community, but also valuable for decision-makers looking for input on how best to serve the public.

Building on our past work with Perspective API, we’ll also be investing in moderation tools and bridging technology for deliberative processes and systems. Based on our experience and research, we believe that when high quality moderation, bridging techniques and effective sensemaking work together, more individuals will participate, and participate more deeply, allowing a more complete picture of the community to emerge, alongside insights that can be acted upon. But we can’t do this alone. To make the greatest impact, we want to partner with the many dedicated organizations already working toward this vision.

Google.org Impact Challenge: Strengthening Democracy in Europe

Jigsaw is partnering with Google.org on the Google.org Impact Challenge: Strengthening Democracy in Europe, a €15M open call to fund European nonprofits, civic entities, academic institutions, and social enterprises to help scale initiatives promoting democratic resilience in the region. Google is looking to support organizations that leverage advanced technology and AI in innovative ways, including those dedicated to strengthening civic discourse, such as encouraging scaled dialogue and community building or advancing deliberative tech that fosters, supports, and leads to informed decision-making.

The present moment offers an unprecedented opportunity to harness technology and AI’s potential, from fostering the essential exchange of ideas to broadening civic engagement. Fully capturing that opportunity will require representation from the broadest possible range of backgrounds, skill sets, and knowledge, and effective partnership between public, private and civic organizations. Across Google and at Jigsaw, we believe democracy and civic dialogue have always sustained cooperation and innovation. That’s why we look forward to supporting a diverse array of individuals and organizations across our initiatives as we strive to ensure voices can be heard and understood at scale in our most important conversations. We’re also excited to see how organizations take the concepts of prebunking, deliberative tech and bridging and apply them in new impactful ways. We encourage interested organizations to apply to the Google.org Impact Challenge at g.co/ImpactChallenge/DemocracyEurope. Applications are now open until August 9.

--

--

Jigsaw
Jigsaw
Editor for

Jigsaw is a unit within Google that explores threats to open societies, and builds technology that inspires scalable solutions.