What is Broad Listening?

With TV or radio, one voice can reach millions of people. But what about the reverse — how can one person properly listen to millions? The opinions that reach governments and politicians have tended to be either too many to read or too few to be representative. Broad listening is an idea that emerged as an answer to this problem of "listening to many".

In short

Broad listening is technology that lets one person listen to many people. If broadcasting is "one to many", this is "many to one".

In concrete terms, AI reads thousands or tens of thousands of free-form comments, groups similar opinions, and turns the whole into a single "map of issues" [1].

Most of the examples below come from Japan, where the practice spread quickly after a 2024 election campaign.

For example

The 2024 Tokyo governor's election. The campaign of Takahiro Anno, a candidate for Governor of Tokyo, gathered voices through the hashtag "#TOKYOAI" on X (formerly Twitter), an anonymous suggestion box and other channels. Using a tool called Talk to the City (TTTC), it turned them into scatter plots — charts of dots in which similar opinions are placed near each other — producing more than 30 reports during the campaign. Among the 951 opinions extracted from posts to the anonymous suggestion box, several posts called for removing the income cap on child-rearing support. That request was reflected when the manifesto (the campaign's list of promises) was updated from version 1.0 to 2.0 [2]. The article on Takahiro Anno's 2024 Tokyo campaign covers this in more detail.

Tokyo's long-term strategy. From November 22, 2024, the Tokyo Metropolitan Government invited submissions on "the Tokyo of the 2050s as you imagine it". It analyzed the 27,915 submissions with AI and used the results in reviewing how the policies of its "2050 Tokyo Strategy" were grouped [3][4][5]. → Broad listening for Tokyo's 2050 strategy

Japan's parliament. On March 7, 2025, Takae Ito, a member of the House of Councillors for the Democratic Party for the People, questioned the Prime Minister in the Budget Committee using a "map of issues" made with TTTC as supporting material. The book Broad Listening was written by members of the Japanese civic group Digital Democracy 2030 (DD2030), and it is the main source for this article. It describes this as the first time in Japan's parliament that broad listening results were used as the basis for a question to the government [6][7]. In the Diet record, Ito says the party gathered voices through broad listening using AI from social media, phone calls, forms and a YouTube live stream, and presents the material by saying that each dot is one person's opinion (the record does not name the tool) [14]. Earlier, on January 29, 2025, Takanori Kawai of the same party told a plenary session of the House of Councillors that on January 16 the party had used Talk to the City to visualize the opinions it had received, and had published the results eight days later [15].

How it works

Why it is needed

Talk with 10 people for an hour each and you understand them deeply. But you don't know whether those 10 represent everyone. Ask a million people "yes or no" and you learn the overall trend. But you don't learn "why" or "what kind".

As long as people and time are limited, the choice has tended to be either "listen deeply" or "listen widely". Broad listening aims to do both at once [1].

Chapter 5-4 of the book Plurality, "Augmented Deliberation" (deliberation means talking things through carefully), makes the same point: "while methods of broadcast (allowing many to hear a single statement) have dramatically improved, broad listening (allowing one person to thoughtfully digest a range of perspectives) remains extremely costly and time consuming" [8].

How it runs

TTTC, and Kouchou AI (a Japanese-language tool built on it), work in roughly three steps [2].

1. Generative AI extracts "opinions" from the posts.
2. Each text is converted into a list of numbers (a vector), and similar opinions are grouped together (clustering).
3. Generative AI gives each group a heading and a summary.

The result is a scatter plot in which each dot is one opinion. You can follow a group all the way down to the original posts [2].

Doing this at a realistic cost has only recently become possible. It needed three things to come together: technology that turns the meaning of text into numbers (2019), technology that squeezes many dimensions into a two-dimensional chart (2018), and large language models (LLMs) that can summarize large amounts of text (AI like ChatGPT that reads and writes text; it became widely known with ChatGPT's public release in 2022) [1]. The book Broad Listening estimates that analyzing 10,000 free-form comments with Kouchou AI costs about 500 yen [9].

Three meanings

The word means different things in different contexts. The book Broad Listening separates three [1]:

1. As a tool — a "component" that uses AI to summarize large numbers of opinions and make them visible. Examples are TTTC, Kouchou AI and Google Jigsaw's Sensemaker.
2. As part of augmented deliberation (using technology to deepen discussion) — part of an ongoing cycle of gathering opinions, discussing them and deciding policy. Taiwan's vTaiwan and Polis are leading examples.
3. As something politicians and governments use directly — politicians or officials use the tools to gather voices and apply them to policy, without a discussion forum in between. In Japan, this form spread first, triggered by Anno's run for governor.

How it differs from surveys and public comment

Broad listening can handle large numbers of free-form comments. That makes it possible to find issues the people who designed a survey never thought of. The book Broad Listening describes this as opening up to the public the power to decide what gets on the agenda [9].

Meanwhile, the book positions public comment (a procedure in which a government presents a draft and invites opinions on it) as "a final check on regulations that are already largely complete". Broad listening handles the earlier stage of "discovering and defining the problem", and the book says the two complement each other [10].

What is debated

Many of the limits are first spelled out by the book Broad Listening itself. Outside researchers and the press point to some of the same problems, and to others.

Reflections by those involved (the book Broad Listening)

• It is skewed toward the loudest voices. The people who write free-form comments are those with strong opinions. The silent majority doesn't show up. The book cites a 2019 survey by the US research organization Pew Research Center finding that 6% of all users post 73% of tweets about politics. It also points to a gap in which people who are used to writing are more likely to be heard [9].
• Mistaking counts for "public opinion". The axes of the scatter plot have no meaning. The size of a group is the "number of submitted opinions", not the "share of people in society who hold that opinion". The book says it is a classic mistake to conclude that something is supported because supporting comments outnumbered opposing ones three to one, and warns that broad listening alone should not be used to make policy or electoral decisions [9]. What it recommends is a two-step approach: form hypotheses (guesses about what voices exist) with broad listening, then test them with a survey of randomly selected people [9].
• The tool does not gather the opinions. Whether many opinions come in depends on publicity and public interest. The book states plainly that the large numbers gathered by Anno's campaign and by Tokyo were not due to the tool itself. In one case in Hiroshima Prefecture, only 2 opinions came in through the hashtag [9][11]. → Broad listening in local governments
• Overstating stories of "changing policy". Chapter 5 of the book Broad Listening (by Shinta Nakayama) examined the often-told story that in Tokyo tourism became an independent strategy. It found that the expansion of tourism policy was already under way in a revision made in January 2024, before submissions were invited, and concluded that the story that "broad listening changed policy" is at least partly overstated [3].
• Broad listening isn't just "that scatter plot". The scatter plot made such a strong impression that people tended to take the chart itself to be broad listening. The book cautions that some tools, such as Polis, don't fit into a scatter plot of that kind [1][2].

Criticism from outside

• Skewed toward strong opinions; the counts mean nothing. Outside researchers make the same point as the book. On an ABEMA program, Tsukasa Tanihara, an associate professor at Ritsumeikan University, named as the biggest advantage that it can turn up issues beyond what the organizers imagined. But Tanihara said it has a quirk: only people with opinions post, so strong views for and against dominate, and the percentages and volumes in the results have no meaning. Numbers make the results look objective and give them a strange persuasiveness. Katsuhiro Yoneshige of JX Press said the media should report such results with caveats and, when quantitative backing is needed, use ordinary opinion polls as well [16]. When Nippon TV broadcast an analysis of the 2024 general election, it also cautioned that the data were posts on X, not a broad range of voices like a poll, and that because AI did the analysis, mistakes and fakes could be reflected as they were (Nippon TV ran the analysis jointly with Anno's team) [17].
• Concerns raised in Tokyo's public comment. Public comments on the draft 2050 Tokyo Strategy included views that the voices gathered came from people used to expressing opinions and should not be used to justify policy, that the way submissions were solicited should be carefully examined, and questions about whether the city would simply publish a summary and stop there. The city replied that it had done this as a reference for drafting the strategy [12].
• Can the gathered voices lead to deliberation? Political scientist Masaki Taniguchi (University of Tokyo; president of the NIRA think tank) wrote in a NIRA report that Taniguchi is watching broad listening as a new movement, and that one challenge is how to go beyond collecting opinions and connect them to deliberation that can reach agreement [18].
• Public will that cannot be picked up. Yuri Kono, a historian of Japanese political thought at Hosei University, wrote in a comment on an Asahi Shimbun article that the recent stagnation of the Democratic Party for the People is probably one fact that casts doubt on the reliability of broad listening, and that there is public will that AI and the internet can pick up and public will they cannot [19].
• The limits of AI summaries themselves. Researchers Seth Lazar (Australian National University) and Lorenzo Manuali (University of Michigan) argue that LLM summaries handle frequent content well but tend to do much worse at drawing out content that is rare but important (one or two brilliant ideas, or a minority raising serious harm), and that unlike human summarizers they cannot be held accountable for errors [20]. Micah Weinberg of the Carnegie Endowment for International Peace writes that AI summaries tend to sand down disagreements and make people look more aligned than they were, and calls for transparency that lets participants see how their input was categorized and flag misrepresentations [21].
• A cautious government answer. On January 29, 2025, Prime Minister Shigeru Ishiba replied to Kawai that AI and digital technology may make it possible to use more information efficiently, but that there are also points to sort out, such as cost, the handling of data, and whether it is appropriate for the government to develop AI on its own [15].

Common misunderstandings

• The misunderstanding that "AI decides policy". The book says it is a tool that can be used under any political system, and that it is humans (elected politicians and legislatures) who read the results and decide [10].

Where it fits in Plurality

In Plurality, broad listening appears in chapter 5-4, "Augmented Deliberation", as "the 'broad listening' problem", in contrast to broadcasting. The book says that generative foundation models (large AI models usable for many purposes) such as LLMs are beginning to open a way to tackle this problem, and cites the Anno campaign's efforts as an example: a manifesto published on GitHub, visualization with TTTC, and an AI avatar that answered questions around the clock [8].

On the origin of the term, Plurality says in a footnote: "To our knowledge, this concept of 'broad listening' originates with Andrew Trask" [8]. In July 2024, Glen Weyl, one of the book's authors, said the same on X [13].

The book Broad Listening notes that the term appears only three times in Plurality's chapter on augmented deliberation, and says that a hidden aim of its own book is to make a "watered-down Plurality" — diluted enough to be understood through real examples [1].

Neighboring articles: - Talk to the City and Kouchou AI — tools that make maps from free-form comments - Polis — a tool that makes opinion maps from agree/disagree votes - Idobata — a discussion space where AI interviews people one by one - Bridging — opinions that win support across groups

Further reading

1. Plurality, chapter 5-4 "Augmented Deliberation" (English, book) — https://www.plurality.net/v/chapters/5-4/eng/
2. Broad Listening, chapter 1 "What is broad listening?" (Japanese and English, book, free to read) — https://broadlisteningbook.com/
3. Manuscript repository of the book Broad Listening (in Japanese, book; chapters 2 and 3 cover bias and the relationship to public comment) — https://github.com/digitaldemocracy2030/broad-listening-book
4. Hirokazu Nishio, "ブロードリスニング" (Broad listening) (in Japanese, notes) — https://scrapbox.io/nishio/ブロードリスニング
5. Tokyo Metropolitan Government, results of public comment on the draft "2050 Tokyo Strategy" (in Japanese, government document) — https://www.seisakukikaku.metro.tokyo.lg.jp/documents/d/seisakukikaku/2050tokyo_public_comment

Sources

1. Broad Listening, chapter 1 "What Is Broad Listening?" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/01_what_is_broad_listening.md ［CC BY-NC 4.0］
2. Broad Listening, chapter 4, section 1 "Takahiro Anno's Initiatives in the 2024 Tokyo Governor Election" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/04_01_takahiro_anno_initiatives.md ［CC BY-NC 4.0］
3. Broad Listening, chapter 5 "Shin Tokyo 2050—Did Broad Listening Change a City of 14 Million? (2024-2025)" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/05_tokyo_shin_tokyo_2050_policy_shift.md ［CC BY-NC 4.0］
4. Tokyo Metropolitan Government press release (2024-11-22), call for opinions on "Shin Tokyo 2050" (in Japanese) — https://www.metro.tokyo.lg.jp/information/press/2024/11/2024112203
5. Tokyo Metropolitan Government, Policy Planning Bureau, "Approach to broad listening" (2025-01-31, in Japanese) — https://www.seisakukikaku.metro.tokyo.lg.jp/documents/d/seisakukikaku/kangaekata3
6. Broad Listening, chapter 4 "The Spread of Broad Listening in Japan" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/04_spread_of_broad_listening_in_japan.md ［CC BY-NC 4.0］
7. Broad Listening, chapter 4, section 2 "The Democratic Party for the People Uses an Issue-Space Visualization in Questioning the Government in the Japanese Parliament" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/04_02_dpfp_parliamentary_questions.md ［CC BY-NC 4.0］
8. Plurality, 5-4 "Augmented Deliberation" — https://www.plurality.net/v/chapters/5-4/eng/ ［CC0］
9. Broad Listening, chapter 2 "The Difference Between Broad Listening and Surveys, and the Shift from Quantitative to Qualitative Analysis" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/02_broad_listening_vs_surveys.md ［CC BY-NC 4.0］
10. Broad Listening, chapter 3 "Digital Democracy, Broad Listening, and New Ways of Delivering Public Voice" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/03_digital_democracy_and_new_public_voice.md ［CC BY-NC 4.0］
11. Broad Listening, chapter 8, section 2, on Hiroshima Prefecture (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/08_02_hiroshima_case.md ［CC BY-NC 4.0］
12. Tokyo Metropolitan Government, results of public comment on the draft "2050 Tokyo Strategy" (excerpt, in Japanese) — https://www.seisakukikaku.metro.tokyo.lg.jp/documents/d/seisakukikaku/2050tokyo_public_comment
13. Hirokazu Nishio, "The naming of broad listening" (includes Glen Weyl's X post https://x.com/glenweyl/status/1817955522035032270; in Japanese) — https://scrapbox.io/nishio/ブロードリスニングの命名
14. Diet record, House of Councillors Budget Committee (2025-03-07), questioning by Takae Ito (in Japanese) — https://kokkai.ndl.go.jp/txt/121715261X00420250307/304
15. Diet record, House of Councillors plenary session (2025-01-29), question by Takanori Kawai and answer by Prime Minister Ishiba (in Japanese) — https://kokkai.ndl.go.jp/txt/121715254X00320250129/10
16. ABEMA TIMES (2025-07-04), on the "quirks" of broad listening, with Tsukasa Tanihara and Katsuhiro Yoneshige (in Japanese) — https://times.abema.tv/articles/-/10186675
17. Nippon TV NEWS (2024-10-26), broad listening of online voices on the general election (news zero; in Japanese) — https://news.ntv.co.jp/category/politics/95dadf7046b240549506356a7c9a6040
18. Masaki Taniguchi, NIRA Opinion Paper no.84 (2025-07, in Japanese) — https://www.nira.or.jp/paper/opinion84.pdf
19. Asahi Shimbun (2025-06-16), on established parties rushing to use AI, with a comment by Yuri Kono (in Japanese) — https://www.asahi.com/articles/AST6F3V7ST6FUTFK01VM.html
20. Seth Lazar and Lorenzo Manuali, "Can LLMs advance democratic values?" (paper, arXiv, 2024-10; revised 2025-05) — https://arxiv.org/abs/2410.08418
21. Micah Weinberg, "Realizing the Potential Gains of AI-Enabled Deliberative Democracy" (Carnegie Endowment for International Peace, 2026-05) — https://carnegieendowment.org/research/2026/05/realizing-the-potential-gains-of-ai-enabled-deliberative-democracy