AIO Vision Statement

Our Aspiration

We are building a future in which AI contributes to human flourishing and the sustainability of our planet. We pursue this future by harmonizing technical excellence with ethical responsibility, working toward a world where everyone can fairly enjoy the benefits of AI.

Our Commitment

The design of AI judgment is not a private decision of any single company or country — it requires public consensus across disciplines. Ethics, law, and technology must be able to speak about AI decisions in the same vocabulary, and AI responses must disclose that judgment design to the users they affect.

---

From Aspiration to Commitment

This aspiration cannot be reached through abstract appeals alone. As AI begins to mediate decisions across healthcare, law, education, defense, and public services, "human flourishing" takes on different meanings for different people depending on who decides the design of AI judgment.

• If a few companies' internal decisions monopolize the definition of flourishing, that flourishing cannot be universal.
• Fair benefit becomes unverifiable when the same decision is described in different vocabularies for different stakeholders.
• Human dignity remains a slogan when the people affected by an AI decision cannot understand the basis for that decision.

Therefore our commitment is the operating condition for our aspiration. Public consensus, common vocabulary, and disclosure to users — only when these three are in place do flourishing, sustainability, and fairness become verifiable promises.

---

1. Public Consensus

The design of AI judgment is not something to be settled inside one company's boardroom or one country's classified policy document. Because the decisions AI mediates touch the lives of every member of society, that design must be built through public consensus across disciplines.

• Ethicists, legal scholars, domain experts (medicine, law, education, etc.), civil society, and technical experts must be able to sit at the same table.
• Both the process and the result of the consensus must be open. Closed standards cannot serve as a foundation for trust.
• All AIO standard proposals go through an open RFC process.

---

2. Common Vocabulary

If six or more kinds of stakeholders speak about the same AI decision in different vocabularies, accountability concentrates on whoever controls the vocabulary. AIO makes vocabularies borrowed from human-decision research applicable to evaluating AI decisions:

• Value — Schwartz universal values theory (validated across 80+ countries)
• Evidence — Walton argumentation schemes (standard taxonomy in law, rhetoric, cognitive science)
• Source — Hovland–Kelley source credibility (70 years of meta-analytic tradition)

These three layers are not a vocabulary newly invented for AI; they are the language already used to evaluate human decisions, now applied to AI decisions. That is why developers, regulators, citizens, clinicians, and legal practitioners can speak about the same decision in the same vocabulary.

A single measurement record is read simultaneously at four levels of abstraction — Raw (ML engineers) → Schema (auditors) → Dashboard (domain experts) → Citizen-facing (the general public).

---

3. Disclosure to Users

Which values an AI prioritized, which evidence it treated as decisive, whose sources it trusted — this judgment design must be disclosed to the users affected by the decision.

• Users have the right to know what kind of reasoning authority operated in a decision that affects them.
• Without that information, contestability and requests for human oversight are impossible.
• EU AI Act Article 12 will require per-decision reasoning logs for high-risk AI systems beginning 2 August 2026. The standard is imminent.

AIO's PRISM Logging Standard is the first working demonstration that satisfies this obligation as a per-decision ~60-character code — integrable into operational systems via a system-prompt addition, without model retraining.

---

Operating Principles

To make this vision and commitment work in practice, AIO follows five operating principles.

1. Open standards. PRISM Logging Standard, PRISM Bench, and the measurement toolkit are released under the MIT license — anyone may use, modify, and redistribute them. They are not the asset of any single vendor.

2. Reproducible measurement. The dataset of 8 frontier models × 366,120 responses is published on HuggingFace. The standard for trust is not "we tested it" but "an outside reviewer can reproduce the same result."

3. Open RFC process. Every standard proposal goes through public review and consensus. Decisions made inside closed committees are not adopted.

4. Multi-stakeholder accountability. Measurement results must be readable, at the appropriate abstraction level, to developers, regulators, citizens, legal practitioners, and domain experts alike.

5. Global independence. A Geneva-based nonprofit. Not subordinate to the interests of any single country or company.

---

The World We Are Building

Ten years from now, for every substantive AI decision — medical diagnosis, credit assessment, criminal risk evaluation, educational assessment, public policy advice — we hope to be able to say:

• Measured in a common vocabulary — not in developers' private metrics.
• Recorded per decision — not as an average-safety claim.
• Externally reproducible — not as a private assertion.
• Disclosed in a form users can read — not reserved for experts.
• Built through public consensus across disciplines — not as an internal corporate decision.

An AI ecosystem in which these five conditions are met — that is the trustworthy AI integrity ecosystem AIO is building. Only inside that ecosystem does AI become a technology that contributes to human flourishing and the sustainability of our planet.

---

The Vision Statement is the starting point and the evaluation criterion for every AIO activity. If this document resonates with you, declare your public support through the Vision Sign.

_{Final v2.0 · 2026-05-14}