UNDelta - UNPAZ

Success for this first Global Dialogue means leaving with a shared diagnosis of the problem and a clear, actionable roadmap. It cannot be measured by a single treaty, but by three concrete outcomes: 1. Shared Understanding That This Is a Human Development Issue We must move beyond framing AI governance as only about ethics or bias. Success means the final statements explicitly recognize that AI reshapes how humans learn and think—and that governance must address this developmental dimension. Libraries and human mediators are named as essential infrastructure, not an afterthought. 2. Concrete Commitments and Funding Diagnosis is not enough. Success means launching a global "HUMAN-AI Literacy" initiative anchored in libraries, building on decades of ALFIN and AMI work. This includes: A formal declaration with multi-stakeholder backing (UNESCO, IFLA, governments) - Specific funding pledges for library-based AI literacy programs - Training materials for librarians and educators, localized globally - A research agenda for "mediating AI"—tools designed to scaffold thinking, not replace it - A lightweight accountability mechanism to track progress annually 3. A Shift in Who Gets Invited The deepest success is cultural. Success means libraries are no longer invisible in tech policy. From now on, every AI governance conversation automatically includes library and information science perspectives. The dominant narrative shifts from "machines replacing humans" to "humans guiding humans through the machine age."

• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models
• Protection and promotion of human rights
• Social, economic, ethical, cultural, linguistic and technical implications of AI

Social, economic, ethical, cultural, linguistic and technical implications of AI This is the foundational priority. AI is not just a technical tool-it is a cultural mediator that reshapes how humans learn, think, and relate to one another. If we govern only the technical dimensions, we miss the deeper story: AI can either amplify human cognition or replace it. Libraries and educators are on the front lines of this cultural shift, and their voices must be central. Protection and promotion of human rights The right to think for oneself is not explicitly in any charter, but it is the foundation of all other rights. An informed, critical citizenry is essential for democracy. If AI teaches people to accept answers rather than question them, we erode the basis of human rights from the inside out. Governance must ensure AI tools protect cognitive autonomy. Transparency, accountability, and human oversight This is where the rubber meets the road. "Human oversight" cannot mean a token checkbox. It must mean meaningful human mediation-librarians, teachers, and community guides who help people navigate AI wisely. Transparency requirements should include clarity about whether an AI tool is designed to scaffold thinking or simply automate it. Open-source software, open data and open AI models Libraries have always been institutions of openness-free access to knowledge for all. Open AI ecosystems align with this mission. They allow public institutions to build local, context-aware AI literacy programs without being locked into commercial products. Openness is the technical foundation for equitable access.

We know from decades of research that human thinking is not innate. It develops through interaction with tools, with language, and with other people. Vygotsky called this the "law of double formation": we learn first socially (with guidance), then individually (on our own). AI breaks this model. It offers the products of thinking without the process. A student can get a finished essay without struggling through arguments. A professional can get a summary without reading the source. This "short-circuit" in development risks creating generations who are consumers of answers rather than producers of thought. This is not an ethics violation. It is not a bias problem. It is a developmental problem-and it cuts across every other theme. Why It Matters for Governance If we ignore this dimension, we risk building beautiful governance frameworks for a world where no one remembers how to think critically. All the transparency and accountability in the world mean little if humans have outsourced their judgment to machines. What Is Needed: A cross-cutting theme on "AI and Human Development" that asks: - How do we ensure AI tools scaffold thinking rather than replace it? - What role do public institutions (libraries, schools) play as human mediators? - How do we measure cognitive autonomy in the AI age? - What does "meaningful human oversight" mean when the human may be losing the capacity to oversee? This theme belongs not as a separate silo, but woven through every other discussion-because if we lose the human capacity to think, we lose everything else we are trying to protect.

Social, Cultural, and Ethical Implications: The Gap in Understanding The Gap: Policymakers treat AI as a technical infrastructure issue. They fund hardware and connectivity but ignore the human mediation layer. The Impact: Libraries are flooded with patrons asking "How do I use this?" and "Can I trust this answer?" without any trained staff or resources to respond. Librarians report a sharp rise in people accepting AI-generated misinformation because they lack the critical skills to question it. The cultural shift from "finding information" to "accepting outputs" is happening in real time, and libraries are scrambling to catch up with no systemic support. Human Rights: The Gap in Cognitive Autonomy The Gap: Human rights frameworks protect freedom of expression and access to information. They do not yet protect the capacity to think critically about that information. The Impact: In my sector, we see students who can no longer write without AI. Professionals who cannot summarize without AI. Citizens who cannot evaluate political claims without AI. This is not a moral failing—it is a developmental consequence of tools designed to replace thinking. The right to develop one's own mind is not encoded anywhere, and we are watching it erode in real time. Transparency, Accountability, and Human Oversight: The Gap in Meaningful Mediation The Gap: "Human oversight" in current frameworks means a technical review at the design stage. It does not mean ongoing human guidance at the point of use. The Impact: Libraries are the front line of human oversight. Every day, a librarian helps a patron question an AI output, reframe a prompt, or seek a second source. But this work is invisible in governance frameworks. There is no funding for it, no training mandate, no recognition that librarians are the human oversight system for millions of ordinary users. The gap leaves libraries under-resourced for the very role they are uniquely positioned to fill. Open-Source and Open Models: The Gap in Public Infrastructure The Gap: Open models exist, but they require technical expertise most librarians and patrons do not have. Meanwhile, commercial products are easy to use but extract data and shape behavior in opaque ways. The Impact: Libraries face an impossible choice: offer patrons polished commercial tools that erode privacy and cognitive autonomy, or offer nothing because open alternatives are too complex. The gap leaves public institutions dependent on private platforms, undermining their mission of equitable, trustworthy access. The Regional Picture: Uneven Development These gaps are global but hit hardest where libraries are already underfunded—rural areas, developing nations, underserved communities. In these contexts, AI is not augmenting human capacity; it is replacing it faster than human guides can respond. The gap between AI-rich and AI-literate widens daily.

1. Legitimizing the Human Development Dimension No single nation can address the cognitive and developmental impacts of AI alone. These are global phenomena driven by global platforms. The Dialogue can elevate this dimension—explicitly naming cognitive autonomy, human mediation, and developmental scaffolding as governance priorities. Once named, they become claimable. Libraries, educators, and civil society gain standing to demand action. 2. Building a Common Framework for HUMAN-AI Literacy Literacy has always been a global project—from ALFIN to media literacy to digital skills. The Dialogue can launch a coordinated effort to define and advance HUMAN-AI literacy as a fundamental competency for the 21st century. This means: Agreeing on core competencies (prompting, evaluating, questioning AI outputs) Developing shared curricula and training materials Creating certification or recognition systems that cross borders 3. Channeling Resources to Public Mediators International cooperation can unlock funding that national budgets alone cannot. The Dialogue can connect governments, foundations, and multilateral institutions around a shared investment agenda: training librarians and educators, equipping public spaces with AI literacy resources, and supporting research into "mediating AI" tools. 4. Ensuring Libraries Are at the Table Too often, library voices are absent from technology governance. The Dialogue can model a different approach by formally integrating library and information science perspectives into all working groups and outcomes. IFLA, national library associations, and frontline librarians must be recognized as essential stakeholders—not special guests. 5. Creating Accountability Without Rigidity The Dialogue can establish lightweight but lasting mechanisms—annual progress reports, regional hubs, communities of practice—that keep momentum between high-level meetings. Not a treaty, but a treaty of attention: a commitment to keep asking "Are we protecting human development?" as AI evolves.

1. UNESCO Recommendation on the Ethics of AI As the first global normative framework on AI ethics, adopted by 194 member states, this Recommendation explicitly recognizes AI's impact on "human thinking, interaction and decision-making" and calls for media and information literacy as essential responses. The accompanying Global AI Ethics and Governance Observatory provides case studies and assessment tools. The Dialogue can operationalize these normative commitments by translating them into actionable HUMAN-AI literacy programs. 2. ITU's AI Skills Coalition Launched in 2025, this coalition has grown to over 70 organizations, assembled 180+ training courses in 13 languages reaching 20,000 learners, and delivers in-person training on AI governance and human rights. The AI for Good Global Summit and Innovate for Impact Report showcase practical AI solutions across education and inclusion. The Dialogue can scale these capacity-building efforts with explicit focus on libraries as delivery platforms. 3. IFLA's Library and AI Framework IFLA's "Entry Point to Libraries and AI" (2025) provides practical tools for librarians to assess AI's benefits and risks, addressing accuracy, diversity, and equitable access. IFLA's Internet Manifesto positions libraries as guardians of knowledge access in the digital age. The Dialogue can legitimize and resource this frontline mediation work. 4. GPAI (Global Partnership on AI) With 44 member countries dedicated to human-centric AI, GPAI bridges theory and practice through multi-stakeholder collaboration. The Dialogue can align with GPAI's practical implementation focus while ensuring library perspectives are integrated. 5. UNICEF's Tinkering with Tech Initiative This program provides hands-on AI learning for children in multiple countries, building foundational digital skills. The Dialogue can connect youth-focused initiatives with lifelong learning through libraries. The Added Value of the AI Dialogue While existing initiatives provide essential building blocks, they remain fragmented. The AI Dialogue can: - Connect the ecosystem - linking UNESCO's norms, ITU's training, IFLA's frontline practice, and UNICEF's youth focus into a coherent global strategy - Elevate the human mediator - explicitly recognizing librarians, educators, and community guides as essential infrastructure, not afterthoughts - Secure sustainable funding - channeling resources to public institutions that provide equitable access to AI literacy - Create accountability - establishing lightweight mechanisms to track progress on HUMAN-AI literacy globally The Dialogue's unique value is integration and legitimization—making visible the invisible work of human mediators and connecting normative frameworks to on-the-ground practice.

1. Design for Integration, Not Isolation Create cross-cutting working groups that weave the "human development" dimension through every thematic area. Do not silo cognitive impacts into a single session—ensure they are discussed wherever transparency, human rights, or social implications are addressed. 2. Include Frontline Voices in Every Panel Every session should include at least one practitioner—a librarian, a teacher, a community educator. Their presence shifts conversation from abstract principle to lived reality. 3. Dedicate a Track to Public Mediators A focused track on "Libraries, Education, and Human Guidance in the AI Age" would develop concrete recommendations for resourcing and supporting frontline institutions. 4. Build a Lightweight Accountability Mechanism Establish an annual "State of HUMAN-AI Literacy" report, co-produced by UNESCO, IFLA, and civil society, tracking progress on protecting cognitive development globally. 5. Create Space for Dialogue, Not Just Declaration Structure significant unscripted time for stakeholders to wrestle with hard questions together. The most valuable outcomes often emerge from genuine conversation, not pre-scripted statements.

1. Frontline Public Mediators: Librarians and Educators The people who actually help ordinary citizens navigate AI are almost never in the room. Librarians, teachers, and community educators work daily with patrons struggling to understand, question, and use AI. Yet governance discussions are dominated by tech executives, policymakers, and academics. How to Include Them: Mandate representation from library associations (IFLA, national bodies) and educator unions in every working group. Fund their participation. Create dedicated advisory roles for frontline practitioners. 2. The "AI-Literate" Public: Everyday Users We hear from experts about what AI does to people. We rarely hear from people themselves—students describing why they stopped thinking, seniors explaining their confusion, workers navigating AI tools they never chose. How to Include Them: Conduct global listening sessions in public libraries (natural gathering spaces). Include qualitative research and ethnographic studies in the evidence base. Create citizen panels that feed directly into governance discussions. 3. The Global South and Non-English Perspectives Most AI governance discourse happens in English, centered in North America and Europe. The developmental impacts hit differently in contexts with weaker educational infrastructure, fewer librarians per capita, and less bargaining power with tech platforms. How to Include Them: Hold regional dialogues co-hosted by local institutions. Translate all materials. Fund participation from civil society in low-income countries. Explicitly ask: "How does this recommendation land in Lagos, not just London?" 4. Developmental Psychologists and Learning Scientists The people who understand how human thinking actually develops are almost entirely absent from AI governance. Their expertise is essential for answering: "What does AI do to a developing mind?" How to Include Them: Create a science advisory panel that includes cognitive scientists and developmental psychologists. Commission white papers on AI's developmental impacts.

Current AI governance dialogues are dominated by technologists and policymakers. Missing are the professionals who actually understand how human thinking develops—librarians, educators, and pedagogues—and who witness daily how AI is reshaping intellectual work.