What is the broad listening for Tokyo's "Shin Tokyo 2050"?

In November and December 2024, the Tokyo Metropolitan Government invited opinions on "the Tokyo of the 2050s". It used AI to sort and visualize the roughly 28,000 submissions and rearranged how the policies of its long-term plan were grouped. The book *Broad Listening* presents it as the leading example of a government using broad listening.

When a government draws up a long-term plan, it sometimes invites residents' opinions. But once the submissions run into the tens of thousands, reading them all and grasping the overall picture is not easy. In drafting its plan for the 2050s, the Tokyo Metropolitan Government, which governs Japan's capital, used AI on this problem of "how to listen to many voices".

In short

To draw up a long-term plan for the 2050s, the Tokyo Metropolitan Government invited people to share "the Tokyo of the 2050s as you imagine it". The new long-term strategy was provisionally called "Shin Tokyo 2050" (shin means "new") [1]. Channels included social media, a dedicated form, post and street interviews [1][2].

The roughly 28,000 voices it gathered were grouped by AI into clusters of similar opinions and shown in a form that looks like a map. The metropolitan government then used this as material for rearranging how the plan's policies were grouped [2]. This way of listening widely to many voices is called broad listening. The metropolitan government explains that it was the largest number of opinions ever gathered for drafting one of its plans [2].

For example

"Children" split into "children" and "child-rearing". According to the metropolitan government's explanatory material, the grouping of policies was rebuilt on the basis of the analysis. The "children" field was split into "children" and "child-rearing", and a new "young people" field was created. "Tourism" was separated from "culture and entertainment", and "urban functions" was divided into "community development and housing" and "infrastructure and transport". "Safety and security" was split into "urban resilience", "disaster prevention", "public order" and "medical care" [3].

Many voices from young people. According to an answer given in the Tokyo Metropolitan Assembly (Tokyo's elected legislature), about 40 percent of the opinions came from people in their teens to thirties [4]. Metropolitan government staff also heard opinions directly from children at places such as Ueno Zoo [5][6].

How it works

What was gathered, when and how

The metropolitan government opened the call on November 22, 2024. Its press release described broad listening as a method of using AI technology to gather, analyze and visualize many opinions [1].

The channels were a dedicated Google Form, X (with the hashtag "#シン東京2050", Shin Tokyo 2050), comments on YouTube videos of the governor's press conferences, email and post [1]. Later the X hashtag "#百合子に言っといて" (roughly, tell Yuriko for me) was also accepted; the December 16 announcement lists both hashtags [7]. The deadline was originally December 13 but was extended to December 20 [7].

Who did it

The work was carried out jointly by the metropolitan government's Policy Planning Bureau and Digital Service Bureau and GovTech Tokyo, an organization affiliated with the metropolitan government [8]. GovTech Tokyo built the system itself and used Talk to the City, an open-source tool (published in a form anyone can use) [5].

The AI engineer Takahiro Anno became an adviser to GovTech Tokyo (responsible for broad listening) in November 2024. He advised on designing the instructions given to the AI (prompts), adjusting settings and how to use the results [1][5]. At the Tokyo Metropolitan Assembly on December 3, 2024, Governor Yuriko Koike said she wanted to bring broad listening into many areas of the metropolitan government, and positioned Shin Tokyo 2050 as its test case [9].

How much came in, and how it was analyzed

The interim report on December 16 analyzed the 11,622 submissions received by December 13 [7]. In the end, 27,915 opinions were received: 13,755 through the call for opinions and 14,160 through individual hearings and other means [2].

In the final report, AI (Azure OpenAI's gpt-4o) extracted 27,657 "opinions" and divided them into 30 groups (clusters). The largest group was "declining birthrate, aging population and economic issues", with 2,847 opinions [10]. The opinions were then re-sorted by each strategy in the draft plan, producing 25 reports [11].

GovTech Tokyo published all the prompts it used to extract, label and summarize opinions. It explains that this was to prevent arbitrary interpretation and bias [5].

How it was used

On January 31, 2025, the metropolitan government published the results of the analysis together with the draft "2050 Tokyo Strategy". The draft introduces broad listening as a new way of making the plan together with residents, and writes that AI analysis brought residents' concerns and issues to the surface and led to a review of how policies were grouped [2]. Each page of the draft included a reference box titled "AI summary of residents' opinions" [2].

The draft went through public comment (a procedure in which a government presents a draft and invites opinions on it; January 31 to March 3, 2025, 261 submissions containing 582 items) and was adopted in March 2025 [12][13]. The "20+1" strategies of the previous plan, the "Future Tokyo" strategy, became 28 strategies [6].

Afterward

At the Tokyo AI Strategy Council on February 6, 2025, the power to "listen" with AI came up for discussion. Members said that data handling and AI bias should be discussed, and that making the processing steps and prompts public would be a first countermeasure [14]. At the Tokyo Metropolitan Assembly on March 17, the metropolitan government answered that it would share this know-how with cities, wards, towns and villages [15]. Examples from other local governments are collected in the article on broad listening in local governments.

The link with Audrey Tang (an easy point to misread)

None of the metropolitan government materials this site collected shows that Taiwan's Audrey Tang was an adviser to the metropolitan government. Two things can be confirmed. First, the metropolitan government's hackathon (a short software development event) took as a reference Taiwan's "Presidential Hackathon", with which Tang was involved [16]. Second, it was announced that Tang would become a visiting professor at Tokyo Metropolitan University in April 2025, and a participatory lecture was announced for May 8 [17].

What is debated

The claim that AI discovered tourism goes too far. Chapter 5 of the book Broad Listening, written by volunteers in DD2030 (by Shinta Nakayama), points out the following. The 2021 "Future Tokyo" plan that the metropolitan government compares against was written in the middle of the COVID-19 pandemic, so it was natural that tourism was not a separate field. The expansion of tourism policy was also already under way in a January 2024 revision, before broad listening. The book writes that "the real nature of the change in the strategic framework was not 'new discovery' but 'subdivision'", and that the story that "broad listening changed policy" is, "at least in part, overstated" [6].
Still, a view that it meant something. The same chapter reads the gathering of young people's and children's voices through social media and street interviews as an attempt to offset, through the way opinions were collected, a bias in which older people's voices tend to dominate. It cites the fact that, within the tourism cluster, calls for measures against overtourism (too many tourists crowding in) appeared alongside calls to promote tourism. On that basis it concludes that the shape of the result as a map of the issues, rather than the counts, served as grounds for decisions. But it cautions that what can be shown is "the structure within the opinions submitted", not statistical public opinion [6].
Concerns raised in public comment. Submissions said that since the voices came from people used to giving opinions, care should be taken not to use them as grounds for evaluating policy; that the way of soliciting submissions, which might stray from the purpose of the law, should be reviewed; and asked whether the city would simply summarize the results on its website and stop there. The metropolitan government replied that it had done this as a reference for drafting the strategy, and had used it to sort out the direction of the plan and how policies were grouped [12].
People with no room to speak up. One submission suggested that the small number of opinions calling for improvements in schools might be because teachers and parents are too busy even to send in opinions. The metropolitan government replied that it had tried to address this through a variety of ways to submit and an extended period [12].
Who counts as a "resident". The metropolitan government's material notes that "residents" in this analysis does not only mean people who live in Tokyo, but everyone who submitted an opinion [3].

Where it fits in Plurality

Chapter 5-4 of Plurality, "Augmented Deliberation", presents Anno's run for governor as an example of broad listening. In a footnote it writes: "Starting November 22, he has used Talk to the City for a project to gather public input for Tokyo's long-term plan toward 2050. As of December 13, over 10,000 opinions have been collected, far exceeding the typical response to public comment calls" [18]. This footnote is the only place in the book where a case from Japanese government appears. See also the article on augmented deliberation.

In the classification used by the book Broad Listening, this is the "politician- and government-led" type, in which a government uses the tools directly without a forum for citizens' discussion [19]. The experience of Anno's campaign in the governor's election (Takahiro Anno's 2024 Tokyo campaign) was brought into government here. DD2030 describes this experience as the starting point for building Kouchou AI (広聴AI, its open-source tool for broad listening) [20].

Sources

Tokyo Metropolitan Government press release, call for opinions on "Shin Tokyo 2050" (2024-11-22, in Japanese) — https://www.metro.tokyo.lg.jp/information/press/2024/11/2024112203
"2050東京戦略（案）" (Draft 2050 Tokyo Strategy) (in Japanese) — https://www.seisakukikaku.metro.tokyo.lg.jp/documents/d/seisakukikaku/2050tokyo-an
Tokyo Metropolitan Government, "ブロードリスニングによって集まった意見について" (On the opinions gathered through broad listening) (in Japanese) — https://www.seisakukikaku.metro.tokyo.lg.jp/documents/d/seisakukikaku/kangaekata3
Tokyo Metropolitan Assembly, General Affairs Committee (2025-03-19, in Japanese) — https://www.gikai.metro.tokyo.lg.jp/record/general-affairs/2025-05.html
GovTech Tokyo, "ブロードリスニングを技術面から大解剖！" (Dissecting broad listening from the technical side) (in Japanese) — https://note.govtechtokyo.jp/n/ndcf8d6b35939
Broad Listening, chapter 5 "Shin Tokyo 2050—Did Broad Listening Change a City of 14 Million? (2024-2025)" (Shinta Nakayama; 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］
Tokyo Metropolitan Government press release, interim report (2024-12-16, in Japanese) — https://www.metro.tokyo.lg.jp/information/press/2024/12/2024121619
Tokyo Metropolitan Government, structural reform team's note (2024-12-02, in Japanese) — https://note.com/kouzoukaikaku/n/n515f8cd11619
Tokyo Metropolitan Assembly, the governor's policy speech (2024-12-03, in Japanese) — https://www.gikai.metro.tokyo.lg.jp/record/proceedings/2024-4/01.html
Shin Tokyo 2050 broad listening final report (2025-01-31, in Japanese) — https://broadlistening.seisakukikaku.metro.tokyo.lg.jp/20250131/index.html
Tokyo Metropolitan Government, Policy Planning Bureau, "意見募集" (Call for opinions) page (in Japanese) — https://www.seisakukikaku.metro.tokyo.lg.jp/basic-plan/choki-plan/ikenbosyu
Results of public comment on the draft 2050 Tokyo Strategy (in Japanese) — https://www.seisakukikaku.metro.tokyo.lg.jp/documents/d/seisakukikaku/2050tokyo_public_comment
Koho Tokyo-to (the metropolitan government's public-relations publication) (2025-03, in Japanese) — https://www.koho.metro.tokyo.lg.jp/2025/03_02_sp/01.html
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
Tokyo Metropolitan Government, structural reform team's note (2024-02-19, in Japanese) — https://note.com/kouzoukaikaku/n/ndce0b021df74
Tokyo Metropolitan Government press release, lecture by Audrey Tang (2025-03-21, in Japanese) — https://www.metro.tokyo.lg.jp/information/press/2025/03/2025032113
Plurality, 5-4 "Augmented Deliberation", footnote — https://www.plurality.net/v/chapters/5-4/eng/ (Japanese edition: https://www.plurality.net/v/chapters/5-4/jpn/) ［CC0］
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］
DD2030, "広聴AI" (Kouchou AI) page (in Japanese) — https://dd2030.org/kouchou-ai/

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

Report an error in this article ・ Ask a question