---
id: "workshop:towards-deep-listening-rethinking-public-engagement-through-generative-ai"
type: "InnovateUS workshop"
title: "Towards Deep Listening: Rethinking Public Engagement Through Generative AI"
url: "https://innovate-us.org/towards-deep-listening-rethinking-public-engagement-through-generative-ai"
date: "2024-05-14"
duration: "60 minutes"
language: "English"
instructors: ["Eric Gordon"]
recording: "https://www.youtube.com/watch?v=1Wsu6asVKJE"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Towards Deep Listening: Rethinking Public Engagement Through Generative AI

InnovateUS workshop · 14 May 2024 · 60 minutes

Join us for this 60-minute virtual workshop with Eric Gordon, professor of Civic Media at Emerson College and director of the Engagement Lab. This training will provide a practical framework for government practitioners to rethink why and how public engagement happens at a time when most public-serving institutions are facing significant trust deficits. The availability of generative AI tools creates new opportunities for “deep listening” practices, in which public sector workers can make sense of large data sets, provide feedback in multiple languages, and for government agencies to process input with transparency and accountability. The workshop will cover these practices and will offer use cases to imagine what’s possible, as well as address potential harms that may arise with these technologies.

## Learning goals

- Understand the importance of rethinking public engagement strategies in the context of trust deficits within public-serving institutions
- Gain practical knowledge of utilizing generative AI tools for "deep listening" practices, including processing large data sets, providing multilingual feedback, and enhancing transparency and accountability in government processes
- Explore real-world use cases to envision the potential applications of AI-enabled public engagement practices while also addressing and mitigating potential harms associated with these technologies

## Speakers

- **Eric Gordon**, Professor of Civic Media at Emerson College and director of the Engagement Lab

Recording: https://www.youtube.com/watch?v=1Wsu6asVKJE

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video.

**Unknown speaker**

[0:01](https://youtu.be/1Wsu6asVKJE?t=1) I'll hand it over to you.

**Eric Gordon**

[0:08](https://youtu.be/1Wsu6asVKJE?t=8) Great, thank you so much for that introduction. If you could just give me a second to find to get into presenter mode here. Okay, great. Can you see my slide okay? Yes. Okay, wonderful. All right. What a pleasure to be here. Thanks for the invitation and thanks everyone for being here today. I'm really looking forward to sharing some thoughts with you on this topic and I'm really looking forward to the conversation afterwards. So I appreciate your questions and and inquiries. So as Hannah said, I'm a professor. I study public engagement. That has been my focus. I'm actually a media studies professor. That's my background is in media studies and communication. So I look at this from the kind of communication lens and I've been working in the space of urbanism and governance for my entire career. And I'm also a practitioner and I think that's an important part of this. So as well as being a scholar, I also do the work. And so I'm really excited to share with you today some of my thoughts about where this space is going and specifically how Gen AI might be changing some of these directions. So let me get started. As I said, I'm a scholar, a practitioner, and a designer. I've spent the last decade building tools for public participation. And you know, I've worked with governments and NGOs around the world to create engaging experiences, often for marginalized or minoritized populations, so that new and different voices can be heard. I've explored games or playful systems and a range of technical tools to make for user-friendly institutional interfaces. And I've spent the majority of my career trying to enhance the front end of engagement. So really like how people

[2:09](https://youtu.be/1Wsu6asVKJE?t=129) experience the interaction. But I realized at some point that the front end doesn't matter if the back end is broken. We can take in all the data in the world, but if our institutions can't make sense of it, then it does nothing. Or worse, it does harm by trafficking in false promises. So I want to start with a story about a current project of mine that aims to address the disconnect between the front end and the back end. So in January 2024, a high school student whom I'll call Anca, in Cluj-Napoca, Romania, is sitting on a beanbag at Casa Transit, an art and culture space in a converted synagogue, where she is interacting with a chatbot on her phone that claims to be a teacher. The teacher invites Anca to share stories or memories of interacting with teachers, and then responds with generous curiosity, inviting Anca to share her ideas about how they can be better in the future. So Anca attentively texts the teacher for about 10 minutes as part of an organized class visit to try out this new AI-enabled process called School of Possibilities. And while she and her classmates interact with the chatbots, which include school- related people, such as a teacher and principal, and school-related objects, such as tests and textbook and whiteboard and mobile phone, etc., insights from the conversations are processed in real time and displayed on a projection screen in the front of the room. Based on Anca's interaction, the teacher says, I may come across as disrespectful to students, which may make them feel undervalued. So Anca glances at the projection screen and continues to interact with the bot.

[3:51](https://youtu.be/1Wsu6asVKJE?t=231) But when the organized experience is completed, Anca raises her hand and shares with the group that the experience was freeing. And she thanked me as one of the designers of the project for making her feel heard. And I responded with, you know it's a chatbot, right? And her response was, yes, and it doesn't matter. I've never been asked what I think about my teachers, and I've never been invited to think creatively about how they could be better. So Anca's feelings of being heard represent a novel experience for her. In extremely low-trust environments like Romania, the AI-enabled interactions motivated her to share some personal stories and to garner some faith that her stories might influence better, more student-focused policy. So this pilot project that was supported by Foundation Botnar, it's a Swiss foundation, was tested with over 300 students in January of this year. It's an experiment in novel governance. It asks the questions, what if honest stories from students in any language could be collected with a trusted platform, effectively summarized in real time, and then shared with decision makers in such a way that makes participants feel heard and that holds institutions accountable? Easy, right? So not so easy. So when government is in crisis, we tend to talk about listening better. But the problem is we mostly just invite more speech.

[5:19](https://youtu.be/1Wsu6asVKJE?t=319) The vast majority of theories of deliberative democracy emphasize the speech moment. So Jürgen Habermas, who's perhaps the foremost philosopher of democracy in the public sphere, talks about what he calls the ideal speech act, but he makes no mention of the ideal listening act. He talks about refining speech so that it rationally can communicate, but again doesn't talk about what it means to listen. Even our contemporary social justice frames are mostly attentive to giving voices to the voiceless. But according to Gideon Calder, the disenfranchised are not just denied a voice, they're denied an audience. Indeed, our current social media culture, speech is relatively easy. It's making sense of it that's hard. So let's zoom into the interactions between people and institutions, and I'll say that it starts with trust. So I just want to say a few words about that and what I mean by that. So this is a bit of a mouthful, but I'm going to go through this. Trust is a relationship between two actors, the trustor and the trustee, in which the trustor is relied upon by the trustee to deliver a forecasted future. So let me unpack that a little bit. Trust involves the decision-making process where different alternatives are considered. There's always a choice involved in the trust relationship. Trust involves the prioritization and selection of risks. Whether it's a good friend or an institution, we are consistently assessing how tolerant we are of certain risks. And trust involves non-cognitive, non-verifiable factors. This includes emotional associations, ideological alliances, etc. So all told, the experience of trusting someone or something always includes some element of faith. Trust is a leap into the unknown, of course mitigated by the knowns, but the subjective experience of trust is premised on one's tolerance of uncertainty.

[7:25](https://youtu.be/1Wsu6asVKJE?t=445) So there's a relationship between interpersonal trust and institutional trust. Each involves a rational and affective assessment of situations, but when we put our trust in institutions, we're dealing with more uncertainty and more abstractions. This is true even at the level, or perhaps mostly at the level, of a nation-state. So the Edelman Trust Index is an annual global survey that looks at trust in institutions. These numbers from 2024 paint a picture where many functioning quote-unquote democracies around the world are experiencing significant trust deficits, and compare this to the most trusted nations, which include Saudi Arabia, China, and the UAE. So institutional trust involves choice. This is why we see that the lack of choice in some national contexts represents even greater perceived trust. So the crisis of democracies is value and expectation misalignment. So while countries like Argentina and Romania continue to score near the bottom, and we might write them off as suffering from corruption and weak institutions, the U.S. and Western European democracies are not far removed, so we really have to ask ourselves what's going on here.

[8:43](https://youtu.be/1Wsu6asVKJE?t=523) So we need to take a hard look at how our democratic institutions govern to understand this. Public institutions intent on legitimizing institutional decision-making through dialogue are being forced to reckon with a rapidly changing communication landscape, much of which came into focus during COVID. So this is this example from the Congressional Management Foundation surveyed members across the country in 2020. 77% of the people surveyed 77% of respondents believed that they would be engaging with constituents remotely in the future, and 60% imagined video town halls, and most predicted that there would be fewer and fewer in-person interactions. Well, all these predictions have turned out to be true, and it comes with consequences. Remote interactions produce digital data traces. While before, analog interactions produced little, if any, digital data, now governments of all shapes and sizes are sitting on mountains of data, and as might be expected, there is increased pressure on government to do something with this data, or for institutional actors to provide meaningful feedback based on this data.

[9:53](https://youtu.be/1Wsu6asVKJE?t=593) So the Edelman report gives us some clue to where the solution rests. According to global respondents, listening is the top trust-building action that government and other institutions can take. Elevated expectations of listening are clear, but our institutions are not prepared to listen any differently than they've been, and this is the problem that I want to really unpack with you today. We're stuck in our analog ways. We're stuck with the perception that data has a short shelf life, and that we need fresh data to protect against community blowback, and most prominently, our data and decision-making infrastructures have not changed for decades. We have more data than ever before, and we are terrible at coordinating our institutions to make sense of it, and it's not uncommon for multiple agencies and local government to reach out to the same community and ask the same questions. One day, a community might be consulted by the parks department. The next day, the schools department. The following day, the streets department. We squander trust with communities when we don't take listening seriously. We're spending so much time on the front end soliciting more and more speech that we're ignoring the processing and analyzing data. Prioritizing the data layer can allow for the creative reimagination of our institutional interfaces. So the data layer involves the collection, analysis, and use of data, and this is not just a database. It involves all sorts of social protocols to make it possible, so with collection, the question is what, how, and why data is collected. Analysis, what questions are we asking of the data, and who gets to ask them, and use, how is the data used, how are insights communicated, and who maintains ownership.

[11:46](https://youtu.be/1Wsu6asVKJE?t=706) Focusing on the data layer requires a rethinking of the underlying logics of governance we have grown so comfortable with. Public engagement has been the dominant paradigm. We seek public voices to influence institutional decision-making. In this paradigm, the institution sets the conditions, and decision-making is primarily centralized. But in a context of data abundance and heightened expectations of personalized responsiveness, the same old approach to interfacing with publics is just not viable. Generative listening is an institutional commitment to collaborate with constituents on shaping and making decisions to the critical collection, analysis, and use of data. So here's the main point. I think engagement has run its course, and we need to listen. I'm going to introduce five qualities of governance and show how generative listening is distinct from public engagement. So those five qualities, I'm going to talk about intake, so how data comes in to an institution. I'm going to talk about speech, the nature of the speech act that happens that we listen to and consult with. Transparency, how we think about transparency within our institutions and why it's important. Feedback, what are the cycles of feedback that we are providing, and then decision-making, ultimately, or where the data flows, where the data is owned, and how we make decisions based on it. So I want to go through each of those now for just a few minutes. So the first one will be intake, and I'm going to distinguish between what I'm calling the public engagement paradigm with what I'm calling the generative listening paradigm. So in the public engagement paradigm, intake is largely closed ended or closed system. So listening takes place within fixed categories and typically supports

[13:48](https://youtu.be/1Wsu6asVKJE?t=828) organizational priorities. Values of the listening agents are not necessarily explicit. So that would be things like surveys or traditional town hall meetings or other ways in which we define the questions, and then we gather the data because that's the only way we're able to analyze that data is within silos that we predetermined so that we can make sense of that consultative data. That's a closed-ended structure. Open-ended is a little bit different, and this is a generative listening paradigm. Listening is open-ended with the capacity to understand new perspectives and inform questions. It provides a clear articulation of values on the part of the listening agent. And one of the things I've already mentioned is that this misalignment of values is a big part of it. It has to do with both the expectations of constituents for their institutions to be aligned with their own values, and it has to do with previous experiences with the digital culture of the way in which feedback and responsiveness happens. So this is the context that we're working within. So one example of this open-ended listening is the project Real Talk for Change.

[14:59](https://youtu.be/1Wsu6asVKJE?t=899) And I just I want to share with you, I will share it in a second, but let me just first tell you a story. So on a hot summer night in 2021, a small group of neighbors gathered in the Roxbury neighborhood of Boston to have a conversation about violence. This community is one of the most impacted by street violence in the city and is regularly asked by city departments to share their experiences. But this conversation was different. It wasn't part of a city-led consultation. It was facilitated by a community organizer and a graduate student and was recorded with the intention of it becoming a public resource. The intimate and wide-reaching conversation was part of a project called Real Talk for Change, and this is a collaboration between MIT, the city of Boston, and an NGO. And this Real Talk for Change sought to enable Boston's, this is what they say, Boston's historically underheard to have a voice in the mayoral election of that year. So the recording of this conversation and dozens of others were analyzed by an AI system to discover themes, manually coded by a group of community members, and then made available on a public website in interactive searchable visualizations. The conversations themselves, audio and all, became part of a public resource of community voice intended to inform the mayoral candidates on violence prevention, along with a range of other related issues such as housing, food access, school reform, etc. Years after the election, the conversations remain accessible and usable by community groups and government to inform local decisions. Now, this is a radical departure from traditional practices of public engagement, where a local government consults with the community on predefined questions and then makes decisions that may or may not be informed by the consultation. In the case of Real Talk, the data precedes the questions. It is analyzed to make the data itself legible and usable, and it is trusted and verified by the people who participated, and it is stored outside of institutional walls. So I just want to show you a little bit from the Real Talk website. So as we see here, this is part of the some of the insights that came from

[17:03](https://youtu.be/1Wsu6asVKJE?t=1023) that were categorized as safety. This is some of the sentiment analysis, which is not that surprising to look at. But then what's interesting is that we have access to the actual audio, so I'll just play a little. So that's just one of thousands of these clips that are shared in the audio. So what's interesting about this is that this database now becomes a resource that stays with the community over time, and in its ideal form, there's a sense of ownership from the community. So when that community is consulted by a government institution, this database is that kind of data layer that I was talking about before, that provides access so that the only way to understand what's going on is not to start a brand new conversation with a brand new agency. So the next element of generative listening that I want to unpack is speech, speech itself. And I talked a bit about why, you know, the sort of overabundance of speech is as part of the problem here, is that that's what we're focusing on

[19:08](https://youtu.be/1Wsu6asVKJE?t=1148) too heavily. And speech in the public engagement paradigm, we tend to think about the kind of conversational nature of speech. So emphasis is placed on motivating speech from as many stakeholders as possible, and often our incentives are aligned such that we want to get more voices, more people speaking, and that's how we identify and justify the work that we're doing, you know, that is directly facing communities. The dialogic paradigm, the dialogic modality that can be, that is part of the generative listening paradigm, emphasizes dialogue between stakeholders and the interaction between publics and organizations. So the difference is subtle, but it's important. The difference here with conversation, it's about more speech, the quantifiable speech. With dialogue, it's about speech that builds, right? And one of the problems with dialogic speech is that it's really hard to make sense of. So when people have conversations and they build off of each other, it's hard to get at the kernel of what's happening within dialogue.

[20:13](https://youtu.be/1Wsu6asVKJE?t=1213) So we often downplay it. It's not the priority in the way that we take in data from communities. I already showed you the School of Possibilities example that used the dialogue between a human and a chatbot and then turned that dialogue into insights that could then be shared and become legible to a community. But there are other, lots of other examples as well. One that I like quite a lot is a project called Talk to the City that I want to share with you. So in collaboration with Silent Cry, which is an NGO serving formerly incarcerated people in Michigan, Talk to the City launched a project called Heal Michigan. So in this case, it uses video interviews to create dialogue and the AI creates claims and subtopics from video interviews and people are invited to interact with them in an ongoing way. So the claims are the kernels of dialogue where ideas build on one another as opposed to threaded discussion or more traditional speech generating tactics like surveys. Some interesting insights came out of this experiment and some interesting challenges with this kind of work. So for example, so originally the system generated language like inmate or ex-convict in the summaries, even though they weren't used by interviewees.

[21:33](https://youtu.be/1Wsu6asVKJE?t=1293) But within the LLM, those words were the ones associated with the sentiment that the interviewees were saying and it generated these terms, which the people involved were were not happy with. And so they created a process where these mistakes, these hallucinations as it were within the LLM were captured and manually corrected. So what's interesting here is that there is room in this, now I should say that this was not an experiment at scale, it was relatively small, but what's useful about this is that it helps us understand what are some of the safeguards that are going to be necessary as we try to extract meaning from dialogue. And some of that is that in the summaries themselves we might see some inaccuracies that we're going to have to address or build structures or scaffolding around it to make sure that they're captured. Transparency is the next topic. So in the public engagement paradigm, attention is paid to transparency of specific transactions. And this includes rules of engagement in town meetings or participation in surveys. And part of this is that, you know, as public engagement professionals are very good at this, right? Very good at holding or creating engagements and being very, very clear about what the rules of engagement are and how people's time and data will be treated within that transaction. A lot of this has to do with that, you know, some of this, a lot of this work is happening by consultants or is being executed by consultants. And so that even that creates another layer of complexity that, you know, having that like consultants only have so much control over the sort of larger structure of the system. In the generative listening paradigm, on the other hand, attention is paid to transparency and how the organization itself collects, analyzes, and uses data and how values

[23:35](https://youtu.be/1Wsu6asVKJE?t=1415) are used to justify organizational decision-making. And this leads to what's often called procedural justice, which is a sense that the procedure itself is just, even if the outcomes are not necessarily what one had hoped for. And so an example of this, I want to use the example of data walks. And I'm sure many of you are familiar with this approach, but data walks are used in all sorts of different ways. And I think they're quite interesting as a means of kind of creating this transparency that I'm talking about. So in Evanston, Illinois, which is a suburb about 15 miles north of Chicago, the city is using the city health dashboard, which is a national resource that helps municipalities access community health and equity data. And it's currently being used by hundreds of cities in the United States. But instead of just using novel data sources in the dashboard and visualizations as a decision support tool, the city has partnered with NGOs and activists in its fifth ward, a predominantly black neighborhood where the life expectancy is 10 years shorter than the city's wealthiest and whitest neighborhoods. Together, they have arranged regular data walks where community members walk the neighborhood and check and verify data sources, including asthma rates, eviction rates, access to healthy foods, et cetera. The walks are led by community members with the representatives from the city present and listening. And the result of these walks has been a bounty of goodwill on the part of the community, as well as state of trust in the data and the city's interpretation of it.

[25:07](https://youtu.be/1Wsu6asVKJE?t=1507) So another example here that's on this slide is a project that I was involved with actually in Boston. This is part of a smart city planning project where youth identified relevant open data about their communities and led walks with community members to discuss and validate the data. So data walks provide transparency in the data layer, which is especially important if that data is fed into LLMs and produce more data. So if we want the data that's coming out of the LLMs to be trusted and trustworthy, this level of legitimation that happens from these data walks is remarkably important. And again, data walks is just one way of doing this, but it essentially provides that human check on the data source that is going to lead to other generative data. The fourth one is feedback. So the public engagement paradigm, traditionally, feedback on specific contributions is given after a period of analysis. And this is just, you know, we'll hold the meeting, we'll do some data collection, and then we'll come back to the community with, here's what we learned, and we'll share that. And there's a time lapse there by necessity. Now, generative listening implies the kind of real-time feedback. This immediate feedback is given to acknowledge a listening relationship, both in terms of feeling heard and being represented. Now, this is both important and incredibly complex and problematic.

[26:35](https://youtu.be/1Wsu6asVKJE?t=1595) So let me give you this example. So we're seeing this most prominently now with government's experimentation with Gen AI chatbots. So the city of New York launched an AI-driven chatbot in late 23 to help small business owners access city services. Users could prompt the system with questions about getting a business license, reporting on tip income, safety and insurance procedures, or a range of other content already available on the city's website. It was solving a number of known problems, non-intuitive interface of the website, language limitations, and inaccessibility of legal jargon. The chatbot uses a simple prompt line similar to chatGBT. It can be prompted in most major languages, and it is able to translate legal and regulatory documents into conversational language. It presents itself as a proxy to the city, encouraging people to interact with it instead of searching the website. So New York is definitely not alone in launching these chatbots. So the city of Buenos Aires, Argentina, launched Botty in 2020. Stockholm, Sweden released its digital concierge around the same time. And these have been not necessarily Gen AI chatbots, but they were in that kind of concierge model would help people with basic city services. But what's interesting is that not five months after the release of the New York chatbot, controversies around its content started circulating. So the bot was producing inaccurate bias and sometimes illegal suggestions. For example, when prompted about whether or not landlords need to accept Section 8 vouchers, the chatbot said no. This is in fact not the case. It is illegal to discriminate based on income in New York City. And so while these mistakes have been addressed and disclaimers have been added to the interface, the criticism is growing in intensity, with many calling for the city to develop more ethical standards for the development of generative AI. So this example was built with limited consultation from end users, and it resulted in some suspicion in its use and continued development. And I think it's an

[28:38](https://youtu.be/1Wsu6asVKJE?t=1718) interesting case for us to consider as we think about chatbots and what they can do, and some of the controversy that surrounds chatbots and the fear really of city-sanctioned misinformation, right? Like that's, I'm sure it's on everybody's mind. So the space is growing more crowded, you know, by the day, it seems, increasingly with for-profit companies like CityBot getting in on the action. And I want to show you, this is a promo video from CityBot that I think sort of captures what the marketing angle is here and how we, you know, how, what need or what itch the market is scratching. So this is a 90-second video. CityBot makes it easy for citizens to connect with their local government and immediately get answers to their questions or make citizen requests. It's as simple as sending a text message. Hmm, I wonder when trash will be picked up this week with the holiday. I should text CityBot.

[29:45](https://youtu.be/1Wsu6asVKJE?t=1785) Oh great, trash pickup is tomorrow. Thanks, CityBot. With CityBot, it's just that simple. You have questions, we find the answers for you instantly. Find something broken in your community that needs a fix? Just report it with CityBot. This pothole report is automatically placed into the local government's workflow system, and the Public Works team receives the order for fixing. The team fixes the pothole, and CityBot sends a text message to the citizen letting them know the issue has been resolved. Wow, that was fast! The pothole on our street has already been fixed. I'm proud of my local government for resolving this. Let's go on another bike ride soon! CityBot connects citizens to their local government, and helps build better communities for all. It's now easier than ever before to communicate with local government, anytime, anywhere. CityBot is citizen engagement fueled by technology.

[31:00](https://youtu.be/1Wsu6asVKJE?t=1860) Okay, so I share that with you, you know, just as it sort of touches on all the themes, right? It touches on trust, you know, now I really believe in my local government. It touches on sort of increased, you know, increased interaction. And then also it touches on, you know, some of the challenges around data integration with like 311 systems and the like. So it's an interesting example that I think is using this idea of listening, of immediate feedback, as a way of kind of expanding what's possible. But I'd love to sort of unpack that with you later, if possible, and just because I see some real challenges with tools like this, especially that are kind of developed and then procured by local governments. I have some questions. The last part that I want to talk about is decision-making. So the public engagement paradigm, data storage and decision-making is centralized within existing institutional structures, and institutions are held accountable by data traces and conversations from specific transactions.

[32:06](https://youtu.be/1Wsu6asVKJE?t=1926) In the generative listening paradigm, data storage and decision-making is distributed across multiple agencies or departments. Institutions are held accountable by publicly accessible data traces and conversations that transcend specific transactions. So the example that I want to use here to illustrate this is the Data Collaborative for Justice, also in New York City. This is an organization devoted to criminal legal system reform. It adopts a collaborative governance approach, meaning that one or more public agencies directly engage non-state stakeholders in a formal decision-making process that is consensus-oriented and deliberative. So the DCJ is committed to shared ownership of data and is operating in collaboration with public institutions, not in opposition. As a result, there are sufficient opportunities for a range of stakeholders to verify and authenticate the data so that it is not questioned when it is put to work. This is one model of a data collaborative that is instructive.

[33:07](https://youtu.be/1Wsu6asVKJE?t=1987) The bottom line is that generative listening requires a fundamental restructuring of where data lives, who has access, and under what conditions. We need to move beyond rhetoric of data-driven decision-making. More data to justify institutional decision-making will not build trust. We need to build a data layer that is verified and authenticated by as many stakeholders as possible because that is the only way decisions will be seen as legitimate. So to review, here are the characteristics of generative listening. It is open-ended, deliberative, systemically transparent, with real-time feedback, and supportive of distributed decision-making. So I want to end with this. Generative listening is a fundamental shift in our thinking about democratic governance. Generative AI is not the solution, but it's a call to action. We need to stop engaging and start listening.

[34:08](https://youtu.be/1Wsu6asVKJE?t=2048) Thank you very much.

**Unknown speaker**

[34:10](https://youtu.be/1Wsu6asVKJE?t=2050) Thank you so much, Eric. Let me just take you off recording, as I promised.

* * *

Original page: https://innovate-us.org/towards-deep-listening-rethinking-public-engagement-through-generative-ai · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org