What is the AI Objectives Institute (AOI)?

Even if thousands of people write down their opinions freely, reading and organizing them all is hard work. The AI Objectives Institute (AOI) is the organization that built a tool that uses AI to read such large numbers of opinions and turn them into a map of the issues. That tool traveled from Taiwan to Japan and became the starting point for Japan's experiments in digital democracy.

In short

The AI Objectives Institute (AOI) is a US nonprofit. The Japanese edition of Plurality translates its name as "AI目的研究所" [1].

AOI does research and practical work so that AI develops in a direction that protects and strengthens human agency — that is, people's ability to think, decide and act for themselves [2]. It is known for building Talk to the City (TTTC), a tool that uses AI to read large numbers of free-form opinions and turn them into a "map of issues", and for publishing it for anyone to use [1][3].

For example

Talk to the City (2023 onward). Colleen McKenzie of AOI explains TTTC as a discourse analysis tool that can summarize large volumes of conversation, suited to gathering opinions from large groups [4].

Chapter 5-4 of Plurality cites TTTC as an example that "illustrates how GFMs can be used to replace a list of statements characterizing a group's views with an interactive agent one can talk to and get a sense of the perspective" [5]. (GFMs are generative foundation models, general-purpose AI trained on large amounts of data.)

Taiwan's Alignment Assemblies (2023). In 2023, Taiwan's Ministry of Digital Affairs held workshops for discussing generative AI. According to its page, AOI used the public data from the activity to produce interactive visual reports [6]. In May 2024, a plan for three-way collaboration between AOI, vTaiwan and the Ministry of Digital Affairs was discussed at a meeting with Audrey Tang (唐鳳) [4]. This is covered in more detail in the article on Alignment Assemblies.

On to broad listening in Japan (2024 onward). According to the book Broad Listening, on May 23, 2024, Haruyuki Seki of Code for Japan introduced TTTC at a meeting with Takahiro Anno and Hirokazu Nishio. This led to its use in Anno's run for Governor of Tokyo [7].

The README (explanatory document) of Kouchou AI, made by the citizens' group DD2030, states that it was built with reference to TTTC, developed by AOI, and uses part of TTTC's source code under its license [8].

In more detail

Who founded it

At a May 2024 meeting with Tang, AOI co-founder Brittney Gallagher said: "Audrey, your work really inspired Peter Eckersley and I originally" [4].

Shinta Nakayama's slides explaining the technology of Kouchou AI introduce AOI as a nonprofit founded with the aim of steering AI to protect and strengthen human agency. They say its founder Eckersley was one of the founders of Let's Encrypt, a system for securing internet communications, and died of illness [2].

Hirokazu Nishio's notes from a study session describe AOI as a nonprofit founded by Eckersley in 2021. They say Eckersley died in 2022, and that Deger Turan and Gallagher have taken over its work at the center [9].

How TTTC was born

Anders Sandberg has posted, in effect, that TTTC was originally conceived by Eckersley during discussions among them about using AI to elicit the preferences of large numbers of people [10].

Research pillars

Nakayama's slides list three strands of AOI's work [2].

• TTTC — a tool that supports policymaking by analyzing the content of opinions, which counting alone cannot reveal.
• Moral Learning — research on having AI acquire moral values.
• AI Supply Chain Observatory — an effort to use AI to monitor supply chains and markets.

In 2023, Turan, Eckersley, Divya Siddarth, Gallagher and others published a white paper setting out a research agenda for building a flourishing civilization [11]. Turan is also one of the co-authors of "Gradual Disempowerment" (2025), a paper discussing the risk that AI gradually takes power away from humanity [12].

Back and forth with Taiwan

At a meeting with Tang on October 25, 2023, Turan said AOI wanted to handle not only short posts but also richer data, such as hour-long recordings of interviews with 13 African American residents of Detroit in the United States [13].

Tang replied that the top priority was to make importing data self-service, "like Google Forms", so that communities could start their own discussions without technical expertise [13]. Turan also spoke of a policy of leaving experimental features out of the first version anyone could use, and releasing features in order of ease of use [13].

At a meeting on November 3, 2023, the Ministry of Digital Affairs, AI Academy and vTaiwan all said they wanted to use TTTC in discussions, and it was discussed that proposals for improvements would be accepted on GitHub [14].

What is debated

Criticism of AOI itself does not appear in the materials presented here. There are, however, points to keep in mind when reading about it.

• Much of the information about its origins comes from secondary sources. The founding year and how the work was handed over rely mainly on Nakayama's slides and Nishio's notes (neither of which is a document by AOI itself) [2][9].
• The editions of the book describe it differently. The English and Japanese editions of Plurality give AOI's name and a link to its website, but the Traditional Chinese edition translates TTTC as "我城對談" and does not name AOI [5][1][15].

Where it fits in Plurality

Chapter 5-4 of Plurality, "Augmented Deliberation", names AOI's TTTC as an example of "one of the most obvious directions" of active development: extending systems like Polis and Community Notes "with modern graph theory and GFMs" (graph theory is the mathematics of connections; GFMs are generative foundation models) [5][1].

A white paper by the Collective Intelligence Project (CIP) also lists TTTC among closely related efforts it supports [16].

The book Broad Listening describes TTTC as "an open-source tool released in 2023 by the US-based AI Objectives Institute", and says "it was the starting point for broad listening practice in Japan" [17]. The spread of broad listening (listening widely to many voices and grasping the whole picture) is covered in that article.

Further reading

1. Plurality, chapter 5-4 "Augmented Deliberation" (English, book) — https://www.plurality.net/v/chapters/5-4/eng/
2. AOI, "Talk to the City" (English, introduction page) — https://ai.objectives.institute/talk-to-the-city
3. Audrey Tang's conversation with AOI (2023-10-25, English, transcript) — https://github.com/audreyt/transcript/blob/main/2023-10-25-Conversation-with-AI-Objectives-Institute.md
4. Broad Listening, preface and glossary (Japanese and English, book) — https://github.com/digitaldemocracy2030/broad-listening-book
5. Shinta Nakayama, "広聴AI技術解説" (Technical explanation of Kouchou AI) (in Japanese, slides) — https://www.docswell.com/s/tokoroten/ZL1M88-2025-06-14-014546
6. Kouchou AI README (in Japanese, document) — https://github.com/digitaldemocracy2030/kouchou-ai

Sources

1. PLURALITY, Japanese edition, 5-4 "拡張熟議" (Augmented Deliberation) (in Japanese) — https://www.plurality.net/v/chapters/5-4/jpn/ ［CC0］
2. Shinta Nakayama, "広聴AI技術解説" (Technical explanation of Kouchou AI) (2025-06-14, in Japanese) — https://www.docswell.com/s/tokoroten/ZL1M88-2025-06-14-014546
3. Broad Listening, "Preface: The Current State and Future of Broad Listening" (in Japanese; English translation in the same repository) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/00_preface.md ［CC BY-NC 4.0］
4. Audrey Tang, Conversation with AI Objectives Institute (2024-05-06) — https://github.com/audreyt/transcript/blob/main/2024-05-06-Conversation-with-AI-Objectives-Institute.md ［CC0］
5. Plurality, 5-4 "Augmented Deliberation" — https://www.plurality.net/v/chapters/5-4/eng/ ［CC0］
6. Ministry of Digital Affairs, "AI 未來民主化審議式工作坊" (Deliberative workshop on democratizing the future of AI) (in Chinese) — https://moda.gov.tw/major-policies/alignment-assemblies/2023-ideathon/1459
7. 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］
8. Kouchou AI (kouchou-ai) README (in Japanese) — https://github.com/digitaldemocracy2030/kouchou-ai
9. Hirokazu Nishio, "Talk to the City勉強会" (Talk to the City study session) (in Japanese) — https://scrapbox.io/nishio/Talk_to_the_City勉強会
10. Anders Sandberg, post on X — https://x.com/anderssandberg/status/1849570772869656662
11. AI Objectives Institute white paper (2023) — https://ai.objectives.institute/whitepaper
12. Gradual Disempowerment (2025) — https://gradual-disempowerment.ai/
13. Audrey Tang, Conversation with AI Objectives Institute (2023-10-25) — https://github.com/audreyt/transcript/blob/main/2023-10-25-Conversation-with-AI-Objectives-Institute.md ［CC0］
14. Conversation with Vehbi Deger Turan, Isabel Hou, Jia-Wei Cui (2023-11-03) — https://github.com/audreyt/transcript/blob/main/2023-11-03-Conversation-with-Vehbi-Deger-Turan,-Isabel-Hou,-Jia-Wei-Cui.md ［CC0］
15. Plurality, Traditional Chinese edition, 5-4 "擴增審議" (in Chinese) — https://github.com/pluralitybook/plurality/tree/main/contents/traditional-mandarin ［CC0］
16. CIP white paper (2023) — https://www.cip.org/whitepaper
17. Broad Listening, "Appendix: Glossary" (English translation) — https://github.com/digitaldemocracy2030/broad-listening-book/blob/f69c60c838a91b91668abf109d7093b240610139/en/99_appendix_glossary.md ［CC BY-NC 4.0］