What is broad listening in local governments?

Efforts by towns and prefectures in Japan to summarize residents' voices, such as free-form survey answers, automatically with tools like Kouchou AI, and so make visible what residents are saying as a whole. In 2025 this was done in the town of Utazu in Kagawa, Iwate Prefecture, Shibuya Ward in Tokyo, Hiroshima Prefecture and elsewhere.

City halls and prefectural offices often include a box in residents' surveys that says "please write freely". But for staff to read and sort thousands of free-form answers one by one takes enormous effort. As a result, voices that were carefully gathered are sometimes not put to full use. Broad listening in local governments is an attempt to cut this effort with AI tools and grasp the overall picture of residents' voices.

(In Japan, local government has two tiers: prefectures, and below them cities, wards, towns and villages. The examples here come from both.)

In short

It is an attempt by towns and prefectures to have AI tools read free-form survey answers and similar material they have collected, and produce a "map" in which similar opinions are grouped together, along with summaries. The aim is to grasp, in a short time, "what residents are saying as a whole". The tool used most is Kouchou AI, a free tool made by the citizens' group DD2030 [1][2].

Broad listening means listening widely to many voices and grasping the whole picture. From 2025, following the large-scale effort by the Tokyo Metropolitan Government, it began spreading to governments from small towns to prefectures [1][3][4].

For example

The town of Utazu, Kagawa Prefecture (published April 30, 2025). The town used Kouchou AI to organize 396 free-form answers from a residents' survey taken when drawing up its long-term plan. The town reports that the AI processed in about 10 minutes a summary that used to take a long time. The mayor commented, in effect, that another attraction is that the AI's classification is based on data rather than being swayed by emotions or preconceptions [1].

Shibuya Ward, Tokyo (published July 2, 2025). The ward analyzed 6,037 free-form answers from its residents' opinion survey and produced an overall summary, a classification by topic with a chart of how the opinions were spread out, and trends by generation and by area [5][6]. Cluster names included calls for cracking down on smoking and drinking on the street. The ward also tried uses such as picking out only the 615 answers from people in their thirties to look at issues of child-rearing and education [6].

Hiroshima Prefecture (August–September 2025). The prefecture used Kouchou AI to analyze 2,953 submissions to a public call for opinions titled "What kind of future Hiroshima can we create through digitalization?", and published interim results while the call was still open. According to staff, "we were able to classify 3,000 free-response entries in under 30 minutes" [7][19]. It is estimated that doing the same work by hand might have taken about 500 hours [7]. According to the prefecture's call page, submissions were accepted from August 12 to September 15, 2025, through a web survey, a hashtag on X, email and post, and partway through also by an automated phone line; the prefecture writes that the broad listening was carried out with the cooperation of DD2030 [17]. The final report the prefecture published extracted 3,179 opinions from the 2,953 submissions and divided them into 20 opinion groups (with 140 below them). The largest was "高齢者と地域のデジタル共生" (digital coexistence of older people and the community; 347 opinions) [18]. The caption of a figure in the book Broad Listening says "15 clusters", which does not match the prefecture's 20 [7][18].

How it works

What is being done

The three cases of Utazu, Iwate and Shibuya all analyzed free-form answers from surveys or discussions that had already been collected — after the fact. Unlike Tokyo, they did not issue a new call for opinions [1][3][5]. Iwate Prefecture tried it on impressions from a "workshop on thinking about happiness" and on opinions gathered from elementary to high school students [3][8].

All three used Kouchou AI. Kouchou AI was released on March 16, 2025 as open source that anyone can use [2]. Boots Inc., a company that advises on digitalization, helped with the introduction [1][3][5].

The benefits local governments cite

The three local governments cite almost the same benefits. One is that summaries that used to be done by hand can be done in a short time. The other is that the overall picture becomes visible, making it easier to grasp residents' voices and issues [3][1][9].

Iwate Prefecture also writes that free-form answers, which it used to publish simply as a list of opinions, can now be made visible with headings and published [3]. Shibuya Ward says, in effect, that analyses it had rarely been able to do because of the effort of tallying can now be done more often [9].

Hiroshima Prefecture — staff training the instructions given to the AI

The book that explains broad listening with Kouchou AI at its center (below, the book Broad Listening) says that the biggest gain for Hiroshima Prefecture was the experience of staff adjusting the instructions to the AI (prompts) themselves [7].

The first problem was that the AI added plausible-sounding meaning to short posts. When staff reworked the prompt, posts like "I don't know" could be excluded from the opinions [7].

The second was the names of the clusters. Staff steered them away from technical words such as "Utilization of Digital Technology" and toward the words of government work, such as "transportation" and "education". As a result, the number of the 20 names containing "digital" dropped from 16 to 6 [7][19]. The book says this made them easier to discuss within the prefectural government [7]. The results were shared at a meeting attended by the governor, and each department was asked how it was responding [7].

These figures can also be checked in the original annual report by the prefecture's Hiroshima AI Lab (Shota Emori). The report also says Kouchou AI analyzed 4,924 posts on X about the November 2025 Hiroshima gubernatorial election, extracting 3,570 opinions in about 15 minutes [19].

Forms other than Kouchou AI — AI that helps discussion

In fiscal 2025, Ota City in Gunma Prefecture brought a tool called "Baisoku Kaigi" by Plural Reality LLC into its "Jibungotokaigi", a citizen meeting in which residents chosen at random discuss issues. When participants answer the AI's questions on their smartphones, a report summarizing points of agreement, points of difference and points of hesitation appears right away and becomes the starting point for the meeting [10]. According to the city's minutes, Plural Reality introduced the AI chat tool "Idobata" at the first meeting and Baisoku Kaigi (answer 15 questions on a five-point scale and a report is produced) at the second, and at the fourth collected comments on the draft proposal with "Idobata Manifesto" [20]. Of the 23 residents chosen at random, 16, 13, 9 and 10 attended the first to fourth meetings. This does not match the book, whose section was written by Plural Reality (the tool's maker) and says "about 20 randomly selected residents participated each time" [10][20]. The book Broad Listening presents this as an example of raising the quality of meetings, distinct from tools that summarize free-form answers [11]. Efforts like this are one form of augmented deliberation (using technology to deepen discussion).

Spreading further

An appendix to the book Broad Listening also lists cases from Maizuru City in Kyoto Prefecture, Nara City, the town of Yosano in Kyoto Prefecture, Higashiosaka City in Osaka Prefecture and others [4]. The Tokyo Metropolitan Government, too, raised at its AI Strategy Council in February 2025 the question of how broad listening could be used widely in local governments' calls for opinions, and answered in the Tokyo Metropolitan Assembly that it would share its know-how with cities, wards, towns and villages [12][13]. The book Broad Listening writes that in January 2026, Hiroshima Prefecture, the Tokyo Metropolitan Government and GovTech Tokyo signed an agreement to cooperate on the use of AI [7].

Outside the book, too, others are reusing Kouchou AI's code. In August 2026 the association Maniken released "Kiku AI" ("listening AI"), built on Kouchou AI with added features, and called especially on local assemblies to use it (Maniken's own announcement) [21]. On October 1, 2026, Tagawa City in Fukuoka Prefecture began using "broad listening" to collect residents' requests online, with the cooperation of a Waseda University institute on democracy (デモクラシー創造研究所) and Maniken (Yomiuri Shimbun) [22]. According to Maniken, the tool is Kiku AI [23].

What is debated

Opinions do not come in to begin with. The book Broad Listening writes that DD2030 approached local governments itself, but that even after trying the tool, many found it hard to move on to full use. The reasons were weak motivation to take it on, a gap in enthusiasm between the heads of local governments and front-line staff, personnel transfers, and difficulty securing engineers. The book also writes that in many municipalities "the challenge lay not in analysis but in collecting enough opinions in the first place" [2].
Results depend on whom you reach. In Hiroshima Prefecture, the call was publicized the same way as the usual surveys, so respondents skewed toward people close to government, such as organizations with ties to the prefecture. Only two opinions came in through the social media hashtag. The book Broad Listening writes that "broad listening is not a tool for 'collecting opinions' but for 'structuring the opinions that have been collected'" [7].
Is the AI's classification really objective? The mayor of Utazu praised the AI's classification as objective [1]. In the Hiroshima case, on the other hand, changing the prompt changed both the number of opinions extracted and the names of the clusters. In other words, the intentions of those using the tool can find their way into the results [7]. The prefecture's annual report recommends publishing the prompts and the AI used, on the grounds that residents' voices must not be selected or altered arbitrarily [7][19]. The head of Boots, which helped with the introductions, also writes that the approach is effective for grasping broad trends but "not necessarily a universal solution" for matters such as picking up opinions that are few in number but important [14].
Limits named in the report of the prefecture that used it. The Hiroshima AI Lab's report says Kouchou AI is effective but not all-powerful. It notes that AI processing can change the original form of residents' opinions (an opinion that was just "applying for bulky-waste collection" was extracted as a request to digitize the application), and that dissimilar topics can be placed in one cluster, making the issues harder to grasp. It asks users to check the original texts and verify results rather than take them at face value. In its acknowledgments, the report says it received support from DD2030 for introducing and running the tool and designing the call [19].
Listening and leaving it there. Maniken, which released Kiku AI, says that conventional resident surveys and public comment reached only a limited set of residents and that the opinions collected could not be fully analyzed, and that if people cannot see what happened to their input afterward, the process becomes mere listening that leads nowhere; so it added a feature to report progress back to residents (Maniken is the maker of Kiku AI) [23].
Separate from use in elections. There are also cases of candidates in the Tokyo Metropolitan Assembly election using Kouchou AI, but that is use in election campaigns, not an effort by local governments [15].

Where it fits in Plurality

The book Plurality does not cover cases from cities, towns or prefectures. The only case from Japanese government is Tokyo's long-term plan, which appears in a footnote to chapter 5-4 (see broad listening for Tokyo's long-term plan).

These local efforts can be seen as a form in which government uses the broad listening of chapter 5-4, "Augmented Deliberation", as a tool for its everyday "public listening" (the work of hearing residents' voices). The book Broad Listening also cites use in local government planning as an example of the spread of broad listening led by politicians and governments [16].

The lineage of the tools runs from Talk to the City to Kouchou AI.

Sources

Town of Utazu, "『ブロードリスニング』の先進的なトライアルを実施しました" (We carried out an advanced trial of broad listening) (2025-04-30, in Japanese) — https://www.town.utazu.lg.jp/page/4114.html
Broad Listening, "Development Work on 広聴AI (Kouchou AI, Broad Listening AI) by DD2030" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/10_00_dd2030_kocho_ai_development.md ［CC BY-NC 4.0］
Iwate Prefecture, "『ブロードリスニング』の検証結果" (Results of testing broad listening) (in Japanese) — https://www.pref.iwate.jp/kensei/kouchoukouhou/1086730.html
Broad Listening, "Appendix: List of Publicly Available Case Studies" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/99_appendix_public_cases.md ［CC BY-NC 4.0］
Shibuya Ward, press release (2025-07-02, in Japanese) — https://www.city.shibuya.tokyo.jp/kusei/hodo/hodo-2025/hodo_20250702.html
Shibuya Ward, detailed material on the AI analysis of its residents' opinion survey (in Japanese) — https://files.city.shibuya.tokyo.jp/assets/12995aba8b194961be709ba879857f70/38d509b421d049c4ad53536fc88ae4ca/kuminishikichosa_ai.pdf
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］
Iwate Prefecture, detailed material on the test (in Japanese) — https://www.pref.iwate.jp/res/projects/default_project/_page/001/086/730/genkou.pdf
Shibuya Ward, results of the AI analysis of its residents' opinion survey (in Japanese) — https://www.city.shibuya.tokyo.jp/kusei/kocho/questionnaire/kuminishikichosa_ai.html
Broad Listening, chapter 8, section 1 "Ota City, Gunma Prefecture: Introducing AI into 'Jibungotokaigi' Citizen Deliberation Meetings" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/08_01_ota_city_jibungotoka_meeting.md ［CC BY-NC 4.0］
Broad Listening, chapter 8 "Use in Local Governments" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/08_use_in_local_governments.md ［CC BY-NC 4.0］
Materials for the 2nd Tokyo AI Strategy Council (2025-02-06, in Japanese) — https://www.digitalservice.metro.tokyo.lg.jp/documents/d/digitalservice/ai_strategy_council02_shiryo
Tokyo Metropolitan Assembly, General Affairs Committee (2025-03-17, in Japanese) — https://www.gikai.metro.tokyo.lg.jp/record/general-affairs/2025-03.html
Broad Listening, chapter 10, section 1 "Boots Inc." (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/10_01_boots_inc.md ［CC BY-NC 4.0］
Broad Listening, chapter 7 "Using Broad Listening in Local Elections" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/07_use_in_local_elections.md ［CC BY-NC 4.0］
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］
Hiroshima Prefecture, "デジタル化で描く未来の広島は？" (call for opinions page; in Japanese) — https://www.pref.hiroshima.lg.jp/site/hiroshima-dx-torikumi/hiroshimanomirai.html
Hiroshima Prefecture, final results report of "デジタル化で描く未来の広島" (in Japanese) — https://www.pref.hiroshima.lg.jp/uploaded/life/1149282_9822914_misc.pdf
Shota Emori (Hiroshima AI Lab), chapter on broad listening with Kouchou AI, in the FY2025 AI Lab annual report (2026-03, in Japanese) — https://www.pref.hiroshima.lg.jp/uploaded/attachment/658524.pdf
Ota City, detailed minutes of "Ota City Jibungotokaigi 2025" (in Japanese) — meeting 1 https://www.city.ota.gunma.jp/uploaded/attachment/37187.pdf , meeting 2 https://www.city.ota.gunma.jp/uploaded/attachment/37188.pdf , meeting 3 https://www.city.ota.gunma.jp/uploaded/attachment/38085.pdf , meeting 4 https://www.city.ota.gunma.jp/uploaded/attachment/38086.pdf
Maniken, announcement of "Kiku AI" for local governments and assemblies (2026-08-01, in Japanese) — https://maniken.jp/blog/3658/
Yomiuri Shimbun (2026-10-01), Tagawa City starts "broad listening" (in Japanese) — https://www.yomiuri.co.jp/local/kyushu/news/20261001-GYS1T00051/
Maniken, "Kiku AI" pilot study starts in Tagawa City (2026-10-01, in Japanese) — https://maniken.jp/blog/3876/

We welcome error reports and questions ("What is…?"). Submissions become public issues.

Report an error in this article ・ Ask a question