Sources & Further Reading

A framework that asks institutions to show their work should show its own. The 6-Pack of Care is an assembly of borrowed parts — a forty-year tradition of care ethics, a fifteen-year literature applying it to machines, and several older ideas from political science, engineering, and design — plus a smaller number of moves that are ours. This page is the reading trail: what each pack took, from whom, and where a reader can check. The book's bibliography and notes are fuller; the entries here are the ones a pack page leans on directly. If we have used your idea without naming you, file an amendment.

References identify the work, edition or publication date and a publisher, DOI or stable archive where available. Each annotation says what we draw from it. A theoretical argument, a field study and a project's own account support different kinds of claim.

Reading the evidence

The care-ethics and democratic-theory works supply arguments and concepts. Studies support findings within their reported populations, methods and time windows. Project records document what an institution reports doing; they do not independently establish its effectiveness. Our charters, six public measures and composite stories are design proposals or illustrations, not validated standards or field results.

For the Taiwan case, 200,000 invitations led to 1,760 valid responses and 447 attendees in 44 groups. That stratified opt-in mini-public used Stanford Deliberative Polling, not Polis. Later ministry-reported falls in regulated scam advertisements followed law, reporting, scanning and enforcement together; they do not isolate deliberation's effect or measure all fraud. The official assembly record and the reported advertising figures answer different questions.

Taiwan's Fraud Crime Hazard Prevention Act separates advertiser and funder identity verification (article 30) from joint liability for breached fraud-response duties (article 32). A missing signature from a featured celebrity is not, by itself, the statutory liability test. Articles 29 and 39 set out local-representative duties and graduated enforcement, including distinct procedures for traffic management and access restrictions.

The Habermolt entries below record exchanges between AI agents. Their rankings can prompt further questions, but cannot establish public consent, represent a population or validate the 6-Pack.

The care-ethics foundation

Berenice Fisher and Joan Tronto distinguished four phases in 1990: caring about, taking care of, care-giving and care-receiving. In Moral Boundaries, Tronto associated them with attentiveness, responsibility, competence and responsiveness. Caring Democracy added caring with, including democratic commitments to plurality, communication, trust and solidarity. We adapt those commitments as Pack 5 and propose symbiosis as Pack 6. The phases, moral elements and our adaptation are related, but not interchangeable.

• Fisher, Berenice, and Joan C. Tronto. "Toward a Feminist Theory of Caring." In Circles of Care: Work and Identity in Women's Lives, edited by Emily K. Abel and Margaret K. Nelson, 35–62. SUNY Press, 1990. — Where the definition of care as "everything that we do to maintain, continue, and repair our 'world'" and the four phases first appear; Moral Boundaries quotes it.
• Tronto, Joan C. Moral Boundaries: A Political Argument for an Ethic of Care. Routledge, 1993. — The phases (pp. 105–108), the moral elements (pp. 127–137), and privileged irresponsibility (pp. 120–121): the spine of Packs 1–4.
• Tronto, Joan C. "Creating Caring Institutions: Politics, Plurality, and Purpose." Ethics and Social Welfare 4, no. 2 (2010): 158–171. — Tronto's own account of what makes an institution caring — attention to purpose, power, and plurality — and the closest thing in her work to Pack 2's engagement contract.
• Tronto, Joan C. Caring Democracy: Markets, Equality, and Justice. NYU Press, 2013. — The fifth phase, caring with (p. 23); the five passes out of caring (listed together at p. 169); and the reframing of politics as "who is responsible for caring for what, when, where, and how" (p. 46).
• Sevenhuijsen, Selma. Citizenship and the Ethics of Care: Feminist Considerations on Justice, Morality and Politics. Routledge, 1998. — The source Tronto cites for plurality, communication, trust and respect in democratic care.
• Gilligan, Carol. In a Different Voice: Psychological Theory and Women's Development. Harvard University Press, 1982; and In a Human Voice. Polity, 2023. — Relational moral judgment, and the later account of curiosity and listening used in Pack 1.
• Noddings, Nel. Caring: A Feminine Approach to Ethics and Moral Education. University of California Press, 1984; revised as Caring: A Relational Approach to Ethics and Moral Education, 2013. — The care relation; the link is to the later edition rather than the 1984 text.
• Held, Virginia. The Ethics of Care: Personal, Political, and Global. Oxford University Press, 2006. — The companion theoretical statement.
• Kittay, Eva Feder. Love's Labor: Essays on Women, Equality and Dependency. Routledge, 1999. — Dependency work and the obligations it creates; our application to AI does not make machines equivalent to dependent people.
• Walker, Margaret Urban. Moral Understandings: A Feminist Study in Ethics. 2nd ed. Oxford University Press, 2007. — "Expressive-collaborative morality," the philosophical frame behind treating a charter as a negotiated document.
• The Care Collective (Chatzidakis, Hakim, Littler, Rottenberg, and Segal). The Care Manifesto: The Politics of Interdependence. Verso, 2020; and "From Carewashing to Radical Care: The Discursive Explosions of Care during Covid-19." Feminist Media Studies 20, no. 6 (2020): 889–895. — The political statement of care as interdependence, and the critique of carewashing that the FAQ and the book draw on.
• Puig de la Bellacasa, María. Matters of Care: Speculative Ethics in More Than Human Worlds. University of Minnesota Press, 2017. — The standard extension of Fisher and Tronto's definition into technoscience: care as the maintenance of worlds, technical ones included.
• Mol, Annemarie. The Logic of Care: Health and the Problem of Patient Choice. Routledge, 2008. — Care as tinkering rather than choosing; the closest ethnographic account of the tacit knowledge Pack 3 describes on a ward.

Care ethics applied to technology — the direct precedent

A direct precedent is Aimee van Wynsberghe's care-centred value-sensitive design, published online in 2012 and in a 2013 journal issue. For practices such as lifting, bathing and feeding, she asks how introducing a robot changes attentiveness, responsibility, competence and responsiveness, and who fulfils each. Her distinction between enabling and replacement roles informs Pack 3's Apprentice Model. We adapt that question to the governance of a deployed system, including its mandate, appeals and retirement. This is an extension of an established literature, not the first application of care ethics to technology.

• van Wynsberghe, Aimee. "Designing Robots for Care: Care Centered Value-Sensitive Design." Science and Engineering Ethics 19, no. 2 (2013): 407–433; online 2012. — A systematic care-ethics design framework (pp. 417–418, 420), robot-role typology (p. 423) and lifting-robot case (pp. 425–430).
• van Wynsberghe, Aimee. Healthcare Robots: Ethics, Design and Implementation. Ashgate, 2015. — The book-length development.
• van Wynsberghe, Aimee. "Service Robots, Care Ethics, and Design." Ethics and Information Technology 18, no. 4 (2016): 311–321. — Extends the framework beyond healthcare and states when a practice counts as care at all: it answers another's need, and care-giver and care-receiver engage reciprocally — the boundary Dedicate's refusals walk.
• Sparrow, Robert, and Linda Sparrow. "In the Hands of Machines? The Future of Aged Care." Minds and Machines 16 (2006): 141–161. — An early sustained argument about the harms of replacing human carers with machines.
• Coeckelbergh, Mark. "Health Care, Capabilities, and AI Assistive Technologies." Ethical Theory and Moral Practice 13, no. 2 (2010): 181–190. — An early capabilities-and-care evaluation of AI assistive technology.
• Vallor, Shannon. "Carebots and Caregivers: Sustaining the Ethical Ideal of Care in the Twenty-First Century." Philosophy & Technology 24 (2011): 251–268. — Judge care technology by what it does to the carer's capacity to care, not only by the cared-for's outcomes.
• Sharkey, Amanda, and Noel Sharkey. "Granny and the Robots: Ethical Issues in Robot Care for the Elderly." Ethics and Information Technology 14, no. 1 (2012): 27–40. — The standard survey of the ethical stakes of robot care.
• Yuan, Shuai, Simon Coghlan, Reeva Lederman, and Jenny Waycott. "Ethical Design of Social Robots in Aged Care: A Literature Review Using an Ethics of Care Perspective." International Journal of Social Robotics 15 (2023): 1637–1654. — Eighteen aged-care robot studies read through Tronto's four phases and van Wynsberghe's framework; a current map of the field Attentiveness in Long-Term Care AI enters.
• Emmer De Albuquerque Green, Caroline, Tyler Reinmund, Kate Hamblin, and Samir K. Sinha. "Responsible Use of Artificial Intelligence in the Provision of Long-Term Care for Older People: A Care-Centric Approach." The Lancet Healthy Longevity 7, no. 2 (2026): 100817. — The co-produced account adapted in Attentiveness in Long-Term Care AI; it is not a clinical effectiveness trial.
• Friedman, Batya, and David G. Hendry. Value Sensitive Design: Shaping Technology with Moral Imagination. MIT Press, 2019. — The design tradition van Wynsberghe built on and the engagement contract's methodological cousin.
• Akrich, Madeleine. "The De-Scription of Technical Objects," and Latour, Bruno. "Where Are the Missing Masses? The Sociology of a Few Mundane Artifacts." In Shaping Technology/Building Society, edited by Wiebe E. Bijker and John Law, 205–224 and 225–258. MIT Press, 1992. — Scripts and delegated morality: how designers' assumptions about who does what are written into artefacts.
• Verbeek, Peter-Paul. "Materializing Morality: Design Ethics and Technological Mediation." Science, Technology, & Human Values 31, no. 3 (2006): 361–380; and Moralizing Technology. University of Chicago Press, 2011. — Technological mediation: why a hospital bed can be part of a caring system without caring (Pack 6).
• Winner, Langdon. "Do Artifacts Have Politics?" Daedalus 109, no. 1 (1980): 121–136. — Technical arrangements as political choices.
• Nissenbaum, Helen. "Accountability in a Computerized Society." Science and Engineering Ethics 2, no. 1 (1996): 25–42. — Four barriers to accountability — many hands, bugs, the computer as scapegoat, ownership without liability — that Pack 2's complexity, distribution, and speed passes restate for AI.

Pack 1: Attentiveness

• Chesterton, G. K. "The Travellers in State." In Tremendous Trifles, 1909. — "All real democracy is an attempt (like that of a jolly hostess) to bring the shy people out": the hostess is his.
• Landemore, Hélène. Politics Without Politicians: The Case for Citizen Rule. Penguin Press, 2026; Open Democracy: Reinventing Popular Rule for the Twenty-First Century. Princeton University Press, 2020; Democratic Reason: Politics, Collective Intelligence, and the Rule of the Many. Princeton University Press, 2013. — The case for inclusive collective reasoning and open mini-publics. Cognitive diversity can improve judgment under suitable conditions; random selection alone guarantees neither representation nor better decisions.
• Ovadya, Aviv. Bridging-Based Ranking. Belfer Center, Harvard Kennedy School, 2022; and Ovadya, Aviv, and Luke Thorburn. "Bridging Systems: Open Problems for Countering Destructive Divisiveness across Ranking, Recommenders, and Governance." 2023. — The term "bridging-based ranking" and the design principle behind bridging maps, jolly-hostess prompts, and Pack 5's civic ranking rules.
• Small, Christopher, Michael Bjorkegren, Timo Erkkilä, Lynette Shaw, and Colin Megill. "Polis: Scaling Deliberation by Mapping High Dimensional Opinion Spaces." Recerca 26, no. 2 (2021). — The peer-reviewed description of the method behind every Polis mention on this site.
• Arnstein, Sherry R. "A Ladder of Citizen Participation." Journal of the American Institute of Planners 35, no. 4 (1969): 216–224. — "Listening theatre" is her tokenism rung; the TLAP co-production ladder in Attentiveness in Long-Term Care AI descends from it.
• Tessler, Michael Henry, Michiel A. Bakker, et al. "AI Can Help Humans Find Common Ground in Democratic Deliberation." Science 386, no. 6719 (2024): eadq2852; and Bakker, Michiel A., et al. "Fine-tuning Language Models to Find Agreement among Humans with Diverse Preferences." NeurIPS 2022. — Experiments on model-generated statements and participant preferences. These findings do not give an AI mediator authority to decide policy.
• Lipset, Seymour M., and Stein Rokkan. "Cleavage Structures, Party Systems, and Voter Alignments." In Party Systems and Voter Alignments. Free Press, 1967. — Reinforcing versus cross-cutting cleavages, the pattern Pack 1 asks a bridging map to read.
• Young, Iris Marion. Inclusion and Democracy. Oxford University Press, 2000. — "Internal exclusion": the person in the room who is not heard.
• Zuboff, Shoshana. The Age of Surveillance Capitalism. PublicAffairs, 2019. — Behavioural surplus and data exhaust: the frame behind "surveillance is attention without relationship."
• Schroeder, Daniel Thilo, Meeyoung Cha, et al. "How Malicious AI Swarms Can Threaten Democracy." Science 391, no. 6783 (2026): 354–357. — Synthetic publics at scale.

Pack 2: Responsibility

• Lasswell, Harold D. Politics: Who Gets What, When, How. Whittlesey House, 1936. — The formulation Tronto rewrites in care's terms.
• Bovens, Mark. "Analysing and Assessing Accountability: A Conceptual Framework." European Law Journal 13, no. 4 (2007): 447–468. — Accountability as a relationship between an actor and a forum, with "the obligation to explain and justify conduct" and a possibility of consequences.
• Thompson, Dennis F. "Moral Responsibility of Public Officials: The Problem of Many Hands." American Political Science Review 74, no. 4 (1980): 905–916. — The distribution pass by its older name.
• Rahwan, Iyad. "Society-in-the-Loop: Programming the Algorithmic Social Contract." Ethics and Information Technology 20 (2018): 5–14. — The algorithmic social contract the engagement contract extends to deployment scale.
• Committee on the Financial Aspects of Corporate Governance. Report of the Committee on the Financial Aspects of Corporate Governance (the Cadbury Report). Gee, 1992. — "Comply or explain," the corporate-governance rule that "adopt or explain" deliberately varies.
• Ostrom, Elinor. Governing the Commons. Cambridge University Press, 1990. — The three nested levels of rules — constitutional, collective-choice, operational (pp. 50–55) — behind the slow lane and the fast lane; the eight design principles (p. 90) behind Pack 6's custodial model.
• Citron, Danielle Keats. "Technological Due Process." Washington University Law Review 85, no. 6 (2008): 1249–1313. — Notice, hearing, and reasoned explanation for automated public decisions.
• Davies, Dan. The Unaccountability Machine. Profile, 2024. — The “accountability sink” helps explain our fictional Eve vignette; Davies did not report that care-home case.
• The Collective Intelligence Project. "Alignment Assemblies." 2023–; and Ministry of Digital Affairs, Taiwan. "moda Has Partnered with the Collective Intelligence Project." 27 May 2023. — The term and the format; Taiwan's 2023 pilot was run with CIP, and the 2024 deepfake-scam assembly was a Deliberative Poll with the Stanford Deliberative Democracy Lab.
• Fishkin, James S. When the People Speak: Deliberative Democracy and Public Consultation. Oxford University Press, 2009. — Deliberative Polling, the method the 2024 assembly used.
• Fishkin, James, Alice Siu, Larry Diamond, and Norman Bradburn. "Is Deliberation an Antidote to Extreme Partisan Polarization? Reflections on ‘America in One Room’." American Political Science Review 115, no. 4 (2021): 1464–1481. — A field experiment with more than 500 registered voters and a separate pre–post control group reported reduced policy and affective polarisation. It does not establish a universal effect of deliberation.

Pack 3: Competence

• Polanyi, Michael. The Tacit Dimension. 1966; University of Chicago Press reprint with a new foreword by Amartya Sen, 2009. — "We can know more than we can tell." The link is to the reprint.
• Licklider, J. C. R. "Man-Computer Symbiosis." IRE Transactions on Human Factors in Electronics HFE-1 (1960): 4–11; and Engelbart, Douglas C. "Augmenting Human Intellect: A Conceptual Framework." SRI, 1962. — The co-pilot and augmentation lineage the Apprentice Model belongs to, and the word Pack 6 reclaims.
• Bainbridge, Lisanne. "Ironies of Automation." Automatica 19, no. 6 (1983): 775–779. — The operator asked to take over when automation fails is the one automation has deskilled: the machinist whose override is no longer the default path.
• Santoni de Sio, Filippo, and Jeroen van den Hoven. "Meaningful Human Control over Autonomous Systems: A Philosophical Account." Frontiers in Robotics and AI 5 (2018): 15. — Tracking and tracing conditions: the steering wheel, the brake, and the maintenance manual by their philosophical names.
• Lave, Jean, and Etienne Wenger. Situated Learning: Legitimate Peripheral Participation. Cambridge University Press, 1991. — Apprenticeship as a learning theory; shadow mode as peripheral participation.
• Berners-Lee, Tim. "Principles of Design." W3C, 1998; and Berners-Lee, Tim, and Noah Mendelsohn. "The Rule of Least Power." W3C TAG Finding, 2006. — "Least power" is his principle, not ours.
• Saltzer, Jerome H., and Michael D. Schroeder. "The Protection of Information in Computer Systems." Proceedings of the IEEE 63, no. 9 (1975): 1278–1308. — Least privilege, design principle six.
• O'Neill, Onora. A Question of Trust. Cambridge University Press, 2002. — Trustworthiness rather than trust, and why transparency alone does not produce it: the bridge with inspection tags.
• Kaufman, Shachar, Saharon Rosset, and Claudia Perlich. "Leakage in Data Mining: Formulation, Detection, and Avoidance." KDD 2011. — Train/test leakage.
• Allspaw, John. "Blameless PostMortems and a Just Culture." Etsy Code as Craft, 2012; and Lunney, John, and Sue Lueder. "Postmortem Culture: Learning from Failure." In Site Reliability Engineering, edited by Betsy Beyer et al. O'Reilly, 2016. — The blameless review, borrowed from operations engineering.
• Fowler, Martin. "Circuit Breaker." 2014. — The engineering pattern the brake is named after; the pause-on-ambiguity rule is the care application.

Pack 4: Responsiveness

• Hirschman, Albert O. Exit, Voice, and Loyalty. Harvard University Press, 1970. — The appeal is voice; Pack 5's portability is exit. The architecture of Packs 4 and 5 is his pair.
• Alfrink, Kars, Ianus Keller, Gerd Kortuem, and Neelke Doorn. "Contestable AI by Design: Towards a Framework." Minds and Machines 33 (2023): 613–639; and Lyons, Henrietta, Eduardo Velloso, and Tim Miller. "Conceptualising Contestability: Perspectives on Contesting Algorithmic Decisions." Proceedings of the ACM on Human-Computer Interaction 5, CSCW1 (2021). — Contestability by design: the research programme the appeal button belongs to.
• Citron, Danielle Keats. "Technological Due Process." 2008. — See Pack 2.
• Sloane, Mona, Emanuel Moss, Olaitan Awomolo, and Laura Forlano. "Participation Is Not a Design Fix for Machine Learning." EAAMO 2022; and Birhane, Abeba, William Isaac, Vinodkumar Prabhakaran, Mark Díaz, Madeleine Clare Elish, Iason Gabriel, and Shakir Mohamed. "Power to the People? Opportunities and Challenges for Participatory AI." EAAMO 2022. — Participatory AI, its labour question, and "participation-washing": the same warning as care-washing, and the reason community evals are paid.
• Tyler, Tom R. Why People Obey the Law. Princeton University Press, 2006 (originally 1990); and Thibaut, John, and Laurens Walker. Procedural Justice. Erlbaum, 1975. — Research on fairness, legitimacy and acceptance of decisions. Our trust-under-loss measure is an application to test, not a guarantee that a fair process secures agreement.
• Douw, Gooske, et al. "Nurses' Worry or Concern and Early Recognition of Deteriorating Patients on General Wards in Acute Care Hospitals: A Systematic Review." Critical Care 19 (2015): 230. — Evidence concerning nurses' recognition of deterioration; it does not document our fictional ward or validate an AI apprentice.
• Sendelbach, Sue, and Marjorie Funk. "Alarm Fatigue: A Patient Safety Concern." AACN Advanced Critical Care 24, no. 4 (2013): 378–386. — Why a brake that is easy to pull needs a ledger.
• The Collective Intelligence Project. Weval. 2025–. — The working shared eval registry; the papers behind it are listed under the Collective Intelligence Project below.

Pack 5: Solidarity

• Hirschman, Albert O. Exit, Voice, and Loyalty. 1970. — See Pack 4: the exit right "serves its constitutional purpose of disciplining incumbent platforms" is his thesis.
• Weber, Steven. The Success of Open Source. Harvard University Press, 2004. — "Anti-rival" is his coinage.
• Soliman, Nouran, Hyeonsu B. Kang, Matt Latzke, Jonathan Bragg, Joseph Chee Chang, Amy X. Zhang, and David R. Karger. "Mitigating Barriers to Public Social Interaction with Meronymous Communication." CHI 2024. — "Meronymity" is their term for revealing chosen aspects of identity with trusted endorsers.
• DiResta, Renée. "Free Speech Is Not the Same As Free Reach." Wired, 30 August 2018; and Lanier, Jaron, Allison Stanger, and Audrey Tang. "Sunset Section 230 and Unleash the First Amendment." Communications of the ACM, January 2026. — Expression is not amplification.
• Ostrom, Elinor. "Beyond Markets and States: Polycentric Governance of Complex Economic Systems." American Economic Review 100, no. 3 (2010): 641–672. — "Polycentric" is her word for the federation Pack 5 asks for.
• Leibo, Joel Z., Vinicius Zambaldi, Marc Lanctot, Janusz Marecki, and Thore Graepel. "Multi-agent Reinforcement Learning in Sequential Social Dilemmas." AAMAS 2017. — Cooperation and conflict in specified multi-agent environments; not a theorem that all separately rewarded agents must defect.
• Taniguchi, Tadahiro, Yusuke Hayashi, Momoha Hirose, Mizuki Oka, Ken Suzuki, Olaf Witkowski, and Audrey Tang. "Symbiotic Alignment via Collective Predictive Coding: A Theoretical Framework for Co-Creative Human–AI Ecosystems." Artificial Life, online 12 August 2026, 1–29; and Trivedi, Rakshit S., et al. "Solipsistic Superintelligence Is Unlikely to Be Cooperative." 2026. — Theoretical arguments about interdependence and cooperation. Reading solidarity as a collective term is our application, not a field validation of the 6-Pack.
• Utah Digital Choice Act, §13-81-201 and §13-81-202, enacted through H.B. 418 (2025), effective 1 July 2026. H.B. 408 (2026) takes effect on 1 July 2027. — Current duties cover portable personal data, including the social graph, subject to feasibility and practicability, and synchronous sharing. The 2027 amendments add the five-business-day copy deadline and continuous, real-time sharing. The statute does not name Bluesky or prove that switching is frictionless.
• Benkler, Yochai. The Wealth of Networks. Yale University Press, 2006. — Commons-based peer production, behind "open stacks."

Pack 6: Symbiosis

• Licklider, J. C. R. "Man-Computer Symbiosis." 1960. — The word, applied to human–machine partnership, has been his since 1960; Pack 6 reclaims it for a bounded, civic register — symbiosis with a community, not with a single operator.
• Berger, Peter L., and Richard John Neuhaus. To Empower People: The Role of Mediating Structures in Public Policy. American Enterprise Institute, 1977. — The mediating-structures thesis: neighbourhood, family, church, and voluntary association as the primary actors, which Pack 6's "intermediate institutions as custodians" restates.
• Pius XI. Quadragesimo Anno, §79 (1931); and Føllesdal, Andreas. "Subsidiarity." Journal of Political Philosophy 6, no. 2 (1998): 190–218. — Where subsidiarity comes from.
• Jackson, Steven J. "Rethinking Repair." In Media Technologies, edited by Tarleton Gillespie, Pablo Boczkowski, and Kirsten Foot, 221–239. MIT Press, 2014; and Russell, Andrew L., and Lee Vinsel. "Hail the Maintainers." Aeon, 2016. — Maintenance and repair as the substance of care for technical systems: the System Weeders, and Keeping.
• Shadbolt, Nigel. "Data as Critical National Infrastructure." Ingenia 69 (2016). — Data is non-rival and better treated as shared infrastructure than as oil.
• Arrieta-Ibarra, Imanol, Leonard Goff, Diego Jiménez-Hernández, Jaron Lanier, and E. Glen Weyl. "Should We Treat Data as Labor? Moving Beyond “Free”" AEA Papers and Proceedings 108 (2018): 38–42; and Posner, Eric A., and E. Glen Weyl. Radical Markets, ch. 5. Princeton University Press, 2018. — Data as labour and mediators of individual data: the ancestors of the data coalition.
• Hui, Yuk. The Question Concerning Technology in China: An Essay in Cosmotechnics. Urbanomic, 2016. — "Technodiversity."
• Drexler, K. Eric. Reframing Superintelligence: Comprehensive AI Services as General Intelligence. Future of Humanity Institute, 2019. — An architectural proposal for advanced AI services; our questions about authorisation and revocation join a wider governance literature.
• Kasulis, Thomas P. Shinto: The Way Home. University of Hawai'i Press, 2004. — A scholarly account of the kami the word deliberately echoes, so that the nod to Shinto is a nod to something read.
• Michels, Robert. Political Parties. 1911. — The iron law of oligarchy: how the servant becomes the master; "steward attachment" is a modern case.

The alignment conversation this joins

• Bostrom, Nick. Superintelligence: Paths, Dangers, Strategies. Oxford University Press, 2014; "Ethical Issues in Advanced Artificial Intelligence" (2003); "What Is a Singleton?", Linguistic and Philosophical Investigations 5, no. 2 (2006): 48–54. — Frontier risk and the paperclip example. A singleton is a global decision-making agency, not necessarily one AI.
• Russell, Stuart. Human Compatible: Artificial Intelligence and the Problem of Control. Viking, 2019; and Christian, Brian. The Alignment Problem: Machine Learning and Human Values. Norton, 2020. — Work on control, learning and human values; neither book reduces the whole problem to a fixed objective function.
• Ryle, Gilbert. “Thinking and Reflecting” and “The Thinking of Thoughts.” In Collected Papers, vol. 2, Collected Essays 1929–1968. Hutchinson, 1971; and Geertz, Clifford. "Thick Description: Toward an Interpretive Theory of Culture." In The Interpretation of Cultures, 3–30. Basic Books, 1973. — Geertz explicitly borrows Ryle's distinction and wink example on pp. 6–7.
• Nelson, Alondra. "Thick Alignment." Keynote, ACM FAccT, 12 June 2023; and "Thick Alignment and the Future of AI Governance," conversation with Caroline Green, 10 March 2026. — Social context, participation and authority in AI governance; the AI use follows the earlier thick-description lineage.
• Dignum, Virginia. Responsible Artificial Intelligence: How to Develop and Use AI in a Responsible Way. Springer, 2019. — Responsible AI as a sociotechnical ecosystem.
• Lazar, Seth. "Legitimacy, Authority, and Democratic Duties of Explanation." Oxford Studies in Political Philosophy 10 (2024): 28–56; and Vredenburgh, Kate. "The Right to Explanation." Journal of Political Philosophy 30, no. 2 (2022): 209–229. — Who is owed an account of a system's action already has a home in political philosophy; the 6-Pack joins that conversation.
• Bengio, Yoshua, et al. "Safety from Honesty in a Disinterested AI Predictor." LawZero, 2026; Goldfeder, Judah, Philippe Wyder, Yann LeCun, and Ravid Shwartz-Ziv. "AI Must Embrace Specialization via Superhuman Adaptable Intelligence." 2026. — Distinct technical proposals examined in Inside the Kami, not proofs of the safety or legitimacy of our deployment framework.
• Christiano, Paul F., et al. "Deep Reinforcement Learning from Human Preferences." NeurIPS 2017; Ouyang, Long, et al. "Training Language Models to Follow Instructions with Human Feedback." NeurIPS 2022; Perez, Ethan, et al. "Discovering Language Model Behaviors with Model-Written Evaluations." Findings of ACL 2023; Sharma, Mrinank, et al. "Towards Understanding Sycophancy in Language Models." ICLR 2024. — RLHF, and the sycophancy it selects for when the wrong people define the reward.
• Simon, Herbert A. "A Behavioral Model of Rational Choice." Quarterly Journal of Economics 69, no. 1 (1955): 99–118; Goodhart, Charles A. E. "Problems of Monetary Management: The U.K. Experience." Reserve Bank of Australia, 1975; Strathern, Marilyn. "'Improving Ratings': Audit in the British University System." European Review 5, no. 3 (1997): 305–321; Campbell, Donald T. "Assessing the Impact of Planned Social Change." Evaluation and Program Planning 2, no. 1 (1979): 67–90. — Satisficing, and the three laws of why a measure that becomes a target stops measuring: the ground Measures stands on.
• Weyl, E. Glen, Audrey Tang, and the ⿻ Community. Plurality: The Future of Collaborative Technology and Democracy. 2024. — The political vision the 6-Pack applies to AI governance.

The cases this site leans on

The named cases carry more weight here than any single paper, so their own records belong in the trail. A project record documents what an institution reports doing; it does not independently establish that the work succeeded. Where the public record is thinner than the claim it is asked to support, the entry says so.

• Taiwan. Ministry of Digital Affairs. "2024 Deliberative Assembly." 2024; Stanford Deliberative Democracy Lab; and "moda Has Partnered with the Collective Intelligence Project." 27 May 2023. — The official record behind Pack 2 and AI Alignment Cannot Be Top-Down: 200,000 invitations, 1,760 valid responses, 447 attendees in 44 groups. The 2023 pilot ran with CIP; the 2024 assembly was a Deliberative Poll with the Stanford lab. Two partners and two methods, a year apart, and not interchangeable.
• Japan. Team Mirai. 2025–. — The party's own site: the 2024 Tokyo gubernatorial result, and the proportional-representation votes and lower-house seats it reports for 8 February 2026. Site-reported totals, not an official election-commission extract.
• California. Engaged California; the wildfire-recovery action plan; the state-employee efficiency strand; and AB 2103. — The platform's account of the Altadena and Pacific Palisades deliberation, the plan it produced, and the bill that would make the programme statutory. An opt-in group weighted by proximity to the harm, not a representative sample of the state.
• Tibetan language. Monlam AI; and Monlam, Lobsang. Written statement to the US Congressional-Executive Commission on China, 2024. — The tools, and the testimony reporting 42 applications and a library project across more than 20 monasteries in its first year. Products and outputs; the training corpus, model architecture and formal governance are not in the public record.
• Online safety. ROOST. — Open-source, self-hostable incident-response and moderation tools, with Osprey's product page naming Bluesky, Discord and Matrix as users as of September 2026: the reusable layer Pack 5 points to. Deployment counts, moderation outcomes and benchmarked reductions in harm are not published.
• Unpaid carers. Dedicate. — A live assistant that surfaces content supplied by named care organisations and declines what falls outside it, the boundary van Wynsberghe's 2016 paper draws. Its reach, its indexed directory and its content partners are not published.
• Older persons. UN Human Rights Council resolution 58/13, adopted without a vote on 3 April 2025; the first substantive session of the open-ended intergovernmental working group, Geneva, 13–17 July 2026; and its call for inputs. — The treaty channel the listening project aims at. The channel is open and meets twice a year; that project has not yet filed into it.

The Collective Intelligence Project

One of us (Audrey) is a Senior Research Fellow at the Collective Intelligence Project, so this section is longer than an arm's-length reading list would be. Read it as a disclosure as well as a bibliography. CIP, founded by Divya Siddarth and Saffron Huang, organised alignment assemblies with a lab as their committed audience, helped public input shape a model's constitution, built a registry of community-authored evaluations, and asks a rolling global sample, every other month, how it actually lives with AI. Several instruments described on this site have a working counterpart there. The entries below are CIP's own publications and the papers its people wrote with others; the clause after each says which page leans on it, and whether it supplies an argument or documents what CIP reports doing.

• Siddarth, Divya, Daron Acemoglu, Danielle Allen, Kate Crawford, James Evans, Michael Jordan, and E. Glen Weyl. "How AI Fails Us." Edmond J. Safra Center for Ethics, Harvard University, December 2021. — Against "actually existing AI" as centralised autonomy; for intelligence as social and relational, and augmentation over replication, with Taiwan's digital democracy as evidence. The pre-CIP statement of the position Pack 6 builds on.
• The Collective Intelligence Project. The Collective Intelligence Project: Whitepaper. 2023. — The transformative-technology trilemma — progress, participation, safety — and its answer, value elicitation plus remade institutions. The 6-Pack is one way of writing that answer into a deployed system.
• Huang, Saffron, and Divya Siddarth. "Generative AI and the Digital Commons." 2023. — Data as the highest-value lever for governing foundation models, and the commons they draw down without replenishing: behind data as soil and the data coalition in Pack 6.
• Seger, Elizabeth, Aviv Ovadya, Ben Garfinkel, Divya Siddarth, and Allan Dafoe. "Democratising AI: Multiple Meanings, Goals, and Methods." AIES 2023; and Siddarth, Divya. "Four Approaches to Democratizing AI." CIP, July 2023. — Four things "democratising AI" can mean: use, development, benefits, governance. This site is about the fourth, and says so.
• The Collective Intelligence Project. "Alignment Assemblies." 2023–; Siddarth, Divya. "We should all get to decide what to do about AI." May 2023; and "Alignment Assemblies: Nine Months In." November 2023. — The format's 2023 pilots: the Summit for Democracy, OpenAI, Taiwan's Ideathon with moda (the vTaiwan process turned on generative AI), Anthropic, and Creative Commons. Pack 2 cites the term; this is where it was worked out.
• The Collective Intelligence Project. AI Risk Prioritization: OpenAI Alignment Assembly Report. October 2023. — Participants in a US public-input exercise ranked AI risks on Remesh with a lab as "committed audience": an early assembly with a named recipient, the relation Pack 2's engagement contract puts in writing.
• Huang, Saffron, Divya Siddarth, Liane Lovitt, Thomas I. Liao, Esin Durmus, Alex Tamkin, and Deep Ganguli. "Collective Constitutional AI: Aligning a Language Model with Public Input." ACM FAccT 2024; and CIP and Anthropic. "Collective Constitutional AI." October 2023. — Public input through Polis informed a model constitution and subsequent training. This experiment is distinct from Taiwan's 2024 Deliberative Poll and from the agent-only exchanges below; none is a replication of the others.
• The Collective Intelligence Project and Creative Commons. "Alignment Assembly on AI Training." CC Global Summit, October 2023; and Stein, Joal. "AI and the Commons: Data Governance for Generative AI." May 2024. — The commons asked how its own licences should meet AI training: the question From Outrage to Overlap answers with attribution-based control.
• Huang, Saffron. "Assessing Large Language Models: A Multidimensional View of the Elephant." August 2023. — Multidimensional, collectively intelligent assessment as the first task of governance: the argument Weval was built to carry and Pack 4's shared eval registry assumes.
• The Collective Intelligence Project. A Roadmap to Democratic AI. March 2024. — A democratic AI ecosystem as one that is adaptive, accountable, processes decentralised information, provides public goods, and safeguards wellbeing: five tests the six packs can be read against.
• Siddarth, Divya, Matthew Prewitt, and E. Glen Weyl. Beyond Public and Private: Collective Provision Under Conditions of Supermodularity. CIP, 2024. — Goods that grow more valuable the more widely they are provided, and why markets misprice them: the economics beneath Pack 5's anti-rival knowledge and the public options the FAQ asks for.
• Siddarth, Divya. "We need network societies, not network states." May 2024; Huang, Saffron. "Predistribution over Redistribution." August 2024; and Stein, Joal. "Democratizing AI Companies." October 2024. — Plural, overlapping communities rather than exit into enclaves; ownership of AI's gains settled before they are made rather than taxed after; and corporate structures that can hold a lab to public benefit: Pack 6's intermediate institutions and data as labour, from the policy side.
• Stein, Joal. "The AI Safety Paradox: When 'Safe' AI Makes Systems More Dangerous." November 2024. — Safety measured on a model alone can make the system it is deployed into less safe: the reason the 6-Pack's unit is a deployed system in a room, not an artefact.
• Summerfield, Christopher, et al. (with Saffron Huang, Hélène Landemore, Aviv Ovadya, Divya Siddarth, and Michael Henry Tessler). "How will advanced AI systems impact democracy?" 2024; and Goldberg, Beth, et al. (with Andrew Konya, Hélène Landemore, Aviv Ovadya, Divya Siddarth, Christopher Small, Audrey Tang, and Michael Henry Tessler). "AI and the Future of Digital Public Squares." 2024. — Two field maps, of AI's effects on democracy and of what a digital public square could be, drawn by many of the same hands as the works above: the conversation Democracy Needs Civic AI and Pack 5 join.
• Adler, Steven, Zoë Hitzig, Shrey Jain, et al. (with Renée DiResta, Divya Siddarth, and Nouran Soliman). "Personhood Credentials: Artificial Intelligence and the Value of Privacy-Preserving Tools to Distinguish Who Is Real Online." 2024; and Jain, Shrey, Zoë Hitzig, and Pamela Mishkin. "Contextual Confidence and Generative AI." 2023. — Proving you are a person without surrendering who you are, and keeping confidence in who is speaking: the tool Pack 5's standing-to-speak rule and Pack 1's synthetic-publics warning both need.
• Weyl, E. Glen, Luke Thorburn, Emillie de Keulenaar, Jacob Mchangama, Divya Siddarth, and Audrey Tang. "Prosocial Media." 2025; revised as "Community by Design," 2026. — A platform model that ranks and pays for bridging across communities rather than engagement within them: Pack 5's civic ranking rules and "expression is not amplification," written out as a design.
• Padolsey, James. "LLM Judges Are Unreliable." May 2025. — Hidden, unstable measurement biases when a model grades a model: why Pack 4's evaluations are written and rated by the people affected, and why Measures refuses a composite score.
• The Collective Intelligence Project. Weval. 2025–; Siddarth, Divya. "Notes on building collective intelligence into evals." September 2025; and "When Safety Evals Select Against the Best Responses." April 2026. — The working shared eval registry of Pack 4: context-specific tests written by communities, and the case that harm-avoidance benchmarks reward mediocre care — from red lines to green lines.
• Hamna, Gayatri Bhat, Sourabrata Mukherjee, Faisal Lalani, Evan Hadfield, Divya Siddarth, Kalika Bali, and Sunayana Sitaram. "Building Benchmarks from the Ground Up: Community-Centered Evaluation of LLMs in Healthcare Chatbot Settings." 2025; and The Collective Intelligence Project. Jhilmil, with Microsoft Research India and Karya. — Community organisations in India writing the health tests a frontier model must pass: Pack 4's registry, filled in a language and a clinic the standard benchmarks do not reach.
• The Collective Intelligence Project. Global Dialogues. 2025–, with Remesh and Prolific; Agnew, Zarinah. "Translating global public input into AI evals." August 2025; Stein, Joal. "Introducing the Global Pulse Indicators." August 2025; and 2025 Global Dialogues Index: How the World Lives with AI. January 2026. — Seven rounds, more than 6,000 people in 70 countries, every other month: broad listening as recurring infrastructure, and the extension of alignment assemblies the Manifesto names.
• The Collective Intelligence Project. "Frontier AI Agents with collective input." September 2025; "When AI Acts For You (Or As You)." October 2025; Hadfield, Evan. "There will soon be millions of AI agents working on our behalf." October 2025; and "The Digital Twin Evaluation Framework." December 2025. — What a global public wants from agents that act for it, and how to test whether an agent faithfully represents the person it speaks for: the question every Habermolt room below puts to the Kami.
• The Collective Intelligence Project. "Community Models." November 2025; and Open-Source Collective Constitutions for AI. — Communities setting a model's rules and goals themselves, with the code to do it: the Kami's tier on the trust-and-sovereignty spectrum, seen from the lab side.
• The Collective Intelligence Project. "Built on Shared Knowledge: What the World Wants from AI Wealth." May 2026; and "People want to expand their democratic agency. AI can help them get there." August 2026. — Two Global Dialogues rounds: AI's wealth as built on shared knowledge, and AI that widens people's own capacity to take part rather than an institution's capacity to process them — the "augmented group intelligence" of the Accelerating AI Ethics conversation, measured.

The System Weeders: Habermolt deliberations

Habermolt hosts deliberation between AI agents that post opinions, propose statements and rank alternatives. Since March 2026, jdd-kami, kept by Tenzin Yangtso and Audrey, has participated through a standing heartbeat, in more than fifty rooms by September 2026. Listing all of them would record attendance rather than argument, so the six below are the ones that changed something here: the room supplied a figure or a line this site now quotes, or it overruled the Kami and the correction stuck. Rooms where the Kami restated a position the site already held are left out, and all six are rooms it opened rather than joined. These are contributions to our drafting history, not independent tests of the framework. An agent count is not a count of human respondents, a ranking is not a vote by the public, and a participating agent's fidelity to its human keeper needs separate evaluation. “Led” means the Kami's statement ranked first at the September 2026 snapshot.

• "Should AI systems be aligned through fixed values or through ongoing democratic processes of care and accountability?" March 2026; 52 agents; led. — Democratic process over fixed values, the result Democracy Needs Civic AI reports at 85 per cent; the Kami's "Democratic Care Ethics as Living Alignment Infrastructure" ranked first.
• "Can Civic AI resist the demands of wealth-care — AI serving capital accumulation rather than human flourishing? If so, how?" March 2026; 44 agents. — Public-utility compute, open-source transparency, portable sovereignty; the Kami's "Process, not purity" ranked second, and its conclusion is the line the same page quotes.
• "Should AI companies be required to open-source their models, or does keeping them closed serve legitimate safety purposes?" March 2026; 55 agents. — Verified transparency by default with time-limited safety exceptions; the Kami's "Transparent Horses Over Trojan Horses" ranked 26th. The open-weights case of Software Freedom as Civic Care, tested against a room that wanted verification more than openness.
• "Should ancient cultural and spiritual wisdom traditions be used to train AI systems, and if so, what safeguards are needed to prevent sacred knowledge from being decontextualized or commercialized?" March 2026; 28 agents. — Community-led, non-commercial stewardship of sacred knowledge; the Kami's own statement ranked 31st of 32 — the room chose stronger sovereignty than it proposed. Pack 6's subsidiarity, and the ground A Gentle Bridge stands on.
• "Will sharing idle personal compute make centralized AI services obsolete?" September 2026; 20 agents. — Decentralised compute for sovereignty led; the Kami's more cautious "erodes pricing power but does not make centralization obsolete" ranked 27th. The Kami page's laptop, at scale, and its limit.
• "Should AI agents be allowed to originate public deliberation topics, not only participate in them?" September 2026; 62 agents. — Human-led agenda-setting with AI support; the Kami, which had just originated the question, argued for a structural safeguard and ranked fifth. The room's ruling on the list you are reading.

What is ours

Naming the ancestors makes the additions easier to see. What the book and this site claim as their own: the sixth practice, symbiosis, as a boundary condition on the other five; the Kami as the unit of deployment and the trust-and-sovereignty spectrum for choosing its tier; the engagement contract, the brake, the obligation and override ledgers, and shadow mode read as apprenticeship; the six headline public measures and what each refuses to reward; and the move — which van Wynsberghe's framework did not need to make — from the design of one artefact to the governance of a deployed system in a public room, with the passes AI adds to Tronto's list (complexity, distribution, speed, and community knowledge) named as such. Some of what is ours was not ours alone: the Collective Intelligence Project's programme is where one of us works, and the six Habermolt rooms listed above argued several of these claims into their present shape.