Indiana University

Success for the first Global Dialogue on AI Governance cannot be measured by the number of frameworks agreed upon or declarations signed. It must be measured by whose voices shaped those frameworks — and whose were missing. For this Dialogue to be genuinely historic, it must achieve one foundational outcome above all others: the full, substantive inclusion of every category of stakeholder in the development and governance of AI — not as a procedural courtesy, but as an epistemic necessity. AI governance that is designed without the people most affected by AI is not governance; it is administration by another name. This means including not only governments, corporations, and technical experts, but educators who teach AI to the next generation, students who will inherit the systems being built today, community organizers who see AI's social consequences before policymakers do, and individuals in the Global South whose digital and social transformation is being shaped by tools they had no role in designing. It means including the voices of those who study AI's impact — on labor, on identity, on culture, on civic life — alongside those who build it. From my work developing the Enlightened Intelligence framework and the HAILEI educational AI platform across diverse student populations, I have learned that the most consequential governance failures in AI are not technical. They are failures of inclusion: systems designed without the communities they serve, deployed without the educators who must explain them, and evaluated without the individuals whose lives they change. Success means this Dialogue produces a shared commitment to multi-stakeholder governance architecture — one that embeds participation from education, civil society, affected communities, and individual citizens at every stage: in training data governance, in design and deployment decisions, and in the ongoing assessment of AI's transformation of both our digital infrastructure and our social fabric. The question is not whether AI will reshape our world. It will. The question is whether the people of that world will have shaped the AI.

• Transparency, accountability, and human oversight
• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• AI capacity-building

My four selections reflect a single conviction developed across two decades of research, teaching, and AI systems design: that the governance failures we fear most in AI are not primarily technical failures. They are failures of design philosophy - and they require architectural responses, not regulatory afterthoughts. The social, economic, ethical, cultural, linguistic and technical implications of AI are the foundational domain because governance without a clear theory of harm is governance without direction. My empirical work across many students at multiple institutions, and my research on cognitive failure modes in deployed AI agents, demonstrates that AI's most consequential impacts operate at the level of human agency, equity, and identity - dimensions that technical benchmarks do not capture and that policy frameworks rarely reach. AI capacity-building is inseparable from this. I have directed AI education programs at Indiana University Indianapolis and built the HAILEI multi-agent educational platform specifically because the populations most affected by AI are consistently the least represented in the rooms where AI is governed. Closing that gap is not a parallel priority to governance - it is a precondition for legitimate governance. Safe, secure and trustworthy AI is where my architectural work is most concentrated. My Enlightened Intelligence framework and the CRFS diagnostic model demonstrate formally that trustworthiness is not a property AI systems acquire through compliance requirements. It must be built in: through recursive ethical self-evaluation, continuous cognitive health monitoring, and governance mechanisms that operate inside agent reasoning rather than above it. Transparency, accountability, and human oversight complete the architecture. Transparency that residents can interrogate, accountability records that cannot be retroactively modified, and human oversight boards that engage precisely where system confidence is lowest - these are not values to aspire to. They are design specifications to implement. The Dialogue succeeds if it produces governance frameworks that demand the second kind.

The seven listed themes are necessary but share a common limitation: they are designed to govern AI as a product - a system evaluated at a point in time, certified against a standard, and deployed under a framework. What they do not collectively address is AI as a process - a system that evolves, learns, and potentially drifts from its original values across an operational lifetime. Three cross-cutting issues emerge from this gap. First: the cognitive integrity of deployed AI systems over time. My research formally characterizes a class of failure modes - semantic drift, sycophantic conditioning, and echo-chamber convergence - that emerge not at deployment but through sustained operation under recursive feedback. These are not safety failures in the conventional sense; they are governance failures of a reasoning process that degrades silently while producing outputs that appear coherent. No existing theme addresses the continuous monitoring of AI cognitive health as a governance obligation, distinct from and prior to transparency of outputs. Second: civic co-governance as an active right, not a passive protection. The listed themes address human rights (protection from harm) and transparency (legibility of outputs). Neither captures what I term co-agency: the right and institutional capacity of communities to actively interrogate, challenge, and participate in governing the AI systems that shape their lives. This is a democratic governance question, not a rights protection question, and it requires different institutional architecture - community oversight boards with genuine decision authority, not consultation mechanisms with advisory status only. Third: longitudinal accountability for AI systems that learn and change. Governance frameworks are designed for static artifacts. But AI systems deployed over months and years update, adapt, and accumulate precedent. The governance question is not only what a system does at deployment, but what it becomes - and who is accountable for that becoming. These three issues cut across every listed theme. Without addressing them, any governance framework will be obsolete before it is implemented.

In my sector — higher education and AI systems design — the governance gaps across my four priority areas are producing a fragmentation I observe daily and have studied empirically. They manifest not primarily as technical failures but as three distinct and equally damaging human responses to ungoverned AI. The first is AI-phobia: educators, policymakers, and community members who, in the absence of trustworthy governance frameworks, default to avoidance. They neither engage with AI tools nor develop the literacy to govern them. The governance gap here is not safety — it is the absence of frameworks that make AI legible and trustworthy enough to engage with confidently. The second is uncritical over-reliance: individuals and institutions who delegate judgment, creativity, and accountability to AI systems without the capacity or frameworks to evaluate what those systems are actually doing. In education, this manifests as students who cannot distinguish between AI-assisted thinking and AI-replaced thinking. In governance, it manifests as institutions that automate decisions without monitoring whether the systems making those decisions have drifted from their original parameters. The third is principled but ungoverned rejection: communities and professionals who distrust AI entirely and advocate against its use — often for legitimate reasons, but without institutional channels to translate that distrust into governance input. Their concerns are real; their exclusion from governance processes makes those concerns invisible. All three postures converge on the same outcome: misuse, abuse, and productive non-use of AI. The opportunity is precisely here. Governance frameworks that build trustworthiness architecturally, invest in genuine capacity-building, and create civic co-governance structures give all three populations a constructive path forward — neither fear, nor deference, nor rejection, but informed, participatory engagement. The Dialogue cannot afford frameworks that speak only to those already fluent in AI. It must create the conditions for everyone else.

The dynamic operates as follows. Countries and communities that lack AI infrastructure, literacy, and institutional capacity cannot participate meaningfully in international governance discussions. Governance frameworks designed without them reflect the priorities and risk tolerances of those who built AI, not those who live under it. The AI systems those frameworks enable are consequently not designed for the languages, cultural contexts, or developmental needs of excluded populations. This widens the adoption gap — which further reduces governance participation — which produces the next generation of governance frameworks with the same exclusions built in. The digital divide is not a background condition for AI governance. It is an active governance failure that compounds with each iteration. The Dialogue is positioned to interrupt this loop in a way no prior forum has been, because it is the first genuinely universal platform where every Member State arrives with equal formal standing, before frameworks are finalized. The same three dysfunctional responses I observe in my sector — fear, uncritical adoption, and principled rejection — manifest at the national and regional level, and for the same reason: populations excluded from governance design have no constructive path for engaging with AI on their own terms. International cooperation on AI governance will be meaningful only when it is designed to include, not merely extended to include after the fact. The Dialogue's role is to make that architectural commitment the precondition for every framework that follows it.

The existing landscape of AI governance mechanisms — UNESCO's Recommendation on the Ethics of AI, the ITU's AI for Good framework, the Global Digital Compact, and regional instruments such as the EU AI Act — represents a substantial and necessary policy and technical governance infrastructure. The AI Dialogue should build on all of it. But it should also recognize what this landscape does not yet contain: civic AI infrastructure that enables communities across cultures and civilizations to develop their own governance wisdom through dialogue, rather than receiving governance decisions through consultation after they have been made. The Timeless Agora Institute and its HAILEI platform represent this missing layer. Inspired by the ancient civic practice of the Agora — where free inquiry, ethical deliberation, and collective reasoning were not procedures but a way of life — TAI reimagines this tradition as a global, AI-orchestrated space where learners, educators, and Enlightened AI agents from every cultural and civilizational background engage in structured dialogue on humanity's most pressing questions. This is not a consultation mechanism. It is a civic culture system: anchored in reflection, shared ethical imagination, and what I call intellectual co-becoming across difference. What makes this model globally significant for the Dialogue's purposes is precisely its civilizational breadth. Governance frameworks derived exclusively from Western liberal, technocratic, or geopolitical traditions will not produce legitimate global AI governance. The Dialogue needs platforms that draw on Ubuntu, Islamic ethical traditions, Confucian social philosophy, and Indigenous knowledge systems as genuine co-equal inputs — not as footnotes to a framework already written. The added value the Dialogue can bring is becoming the connective tissue between these two levels: translating the governance wisdom generated in civic dialogue platforms into policy frameworks, and making policy frameworks legible and contestable in civic dialogue spaces. Neither level is sufficient without the other.

The format and structure of the AI Dialogue is not a logistical question. It is a governance question. A Dialogue structured around annual plenary sessions with observer status for non-governmental stakeholders will reproduce the same exclusions that have characterized every prior international AI governance process — regardless of how inclusive its mandate declares itself to be. Format is policy. Different stakeholders require genuinely different contribution pathways, not the same pathway at different volumes. Governments and technical bodies are well-served by formal negotiation and standard-setting tracks; these should continue. But practitioners — educators, community organizers, researchers working with affected populations — contribute most effectively through structured deliberative formats where experiential knowledge is treated as governance-relevant evidence, not testimony. Individuals and communities, particularly from the Global South, require AI-facilitated pre-consultation mechanisms that generate genuine input before Geneva sessions begin — not post-hoc reporting on what was decided. Three structural recommendations follow from this. First: establish pre-Dialogue civic consultation rounds, using AI-orchestrated dialogue platforms to convene learners, educators, civil society, and community practitioners across linguistic and cultural contexts in the months preceding each session. The outputs of these consultations should formally inform the co-chairs' agenda, not supplement it. Second: create tiered contribution tracks — policy, practitioner, and community — each with genuine input authority rather than observer or advisory status only. Third: invest in between-session continuity. Two days annually cannot produce deliberation. AI-facilitated ongoing dialogue infrastructure between sessions transforms the Dialogue from a periodic event into a living governance process. This is the architectural principle my work applies to AI systems themselves: governance embedded in the process, not applied above it after decisions have been shaped. The Dialogue's format should reflect the same standard it asks of the AI systems it governs.

The most consequential form of underrepresentation in global AI governance is not demographic. It is epistemic. Governance forums can achieve visible diversity while still treating only one way of knowing — Western liberal technocratic discourse — as governance-relevant evidence. Until non-Western philosophical traditions, indigenous knowledge systems, and community-based reasoning are treated as co-equal intellectual inputs rather than cultural supplements, the diversity in the room will not translate into diversity in the frameworks that emerge from it. Five communities are systematically underrepresented in ways that current mechanisms do not adequately address. Educators at every level — not only academics but K-12 teachers, community educators, and adult literacy practitioners — witness AI's impact on cognition, equity, and learning daily and are almost never in governance rooms. Young people and students, who will live longest under the decisions being made now and who represent the primary population experiencing AI-phobia, uncritical over-reliance, and principled rejection, have no meaningful participation track. Workers in AI-disrupted sectors carry frontline knowledge about AI's social consequences that no policy paper can replicate. Scholars from Ubuntu, Islamic ethics, Confucian social philosophy, and Indigenous traditions are invited as representatives of culture, not as intellectual peers whose frameworks should shape governance architecture. And communities whose distrust of AI is principled and knowledge-based have no institutional channel through which that distrust becomes governance input rather than protest. Inclusion requires three structural changes. Deliberate multilingual, AI-facilitated pre-consultation mechanisms — not translated summaries of decisions already made, but genuine upstream input. Tiered participation tracks with binding input authority for practitioner and community voices, not advisory status. And civic AI dialogue platforms, such as the Timeless Agora Institute model, specifically designed to convene across epistemological difference — where Ubuntu and Confucian frameworks sit alongside Western ethics as co-architects of governance, not footnotes to it.

The most powerful statement the AI Dialogue could make about AI governance is to demonstrate it. A global forum convened to shape the future of AI that relies on static plenary sessions, PDF submissions, and two-day summits has a credibility gap between its ambitions and its methods. The format of the Dialogue is not a logistical decision — it is an opportunity to show what ethical, inclusive, AI-facilitated civic engagement looks like when designed with intention. Four innovative formats, layered into a hybrid architecture, could transform the Dialogue from a periodic event into a living deliberative process: First: AI-orchestrated pre-session civic dialogue rounds. In the months before each Geneva session, AI-facilitated structured deliberation convenes learners, educators, practitioners, and community members across 100-plus countries and linguistic contexts. AI synthesizes the governance wisdom generated into formal agenda inputs — not supplementary background documents, but structured contributions that co-chairs are required to address. This is the Timeless Agora Institute model: a proven architecture for AI-facilitated cross-cultural philosophical dialogue. Second: asynchronous deliberation tracks. Synchronous sessions exclude everyone whose time zone, work obligations, or caregiving responsibilities prevent participation. Structured asynchronous dialogue — AI-moderated, with rolling synthesis over days rather than hours — transforms participation from attendance into genuine deliberation. Third: cross-civilizational dialogue pods. Small groups of eight to twelve participants, deliberately composed across geographic, cultural, and epistemological difference, AI-facilitated with philosophical framing that treats Ubuntu, Confucian, and Indigenous governance frameworks as co-equal deliberative inputs alongside Western traditions. Fourth: living documentation with real-time AI synthesis. Replace static co-chair summaries with a continuously updated, publicly interrogable synthesis of all contributions — translated, organized thematically, and accessible to all stakeholders between sessions. Together these constitute a hybrid architecture: maximum participation, genuine deliberation, and continuous governance rather than periodic summitry.

Effective AI governance requires working examples at three distinct levels: how AI systems learn with and for people, how AI systems govern their own decisions in real time, and how communities govern AI together. Most referenced governance examples - the EU AI Act, UNESCO's Recommendation, NIST AI RMF - operate at the regulatory layer above deployed systems. They are necessary but not sufficient. The Dialogue should also inventory and amplify the architectural approaches that demonstrate governance embedded in practice. At the educational level, HAILEI (Human-AI Agentic Learning and Education Intelligence), developed at Indiana University Indianapolis, provides an empirically validated model. Deployed across many students, instructors, and 4 institutions, HAILEI demonstrated a 27% increase in learning engagement, a 67% reduction in equity gaps across demographic groups, and an 87% improvement in instructor autonomy - achieved through multi-agent AI orchestration governed by ethical alignment principles rather than performance optimization alone. HAILEI demonstrates that AI systems designed with conscience produce measurably better and more equitable outcomes than those designed for efficiency only. At the technical governance level, the Cognitive Recursive Fragmentation Syndrome (CRFS) diagnostic framework, validated empirically across 120-plus autonomous agents, provides a formal monitoring architecture for detecting semantic drift, sycophantic conditioning, and ethical dissonance in deployed AI systems before these failure modes propagate to outputs. Paired with the Virtue Ledger - a real-time, five-dimension ethical scoring and immutable audit architecture - CRFS monitoring offers a concrete approach to the trustworthiness and accountability requirements that current governance frameworks articulate but do not operationalize. At the civic level, the Timeless Agora Institute provides a working model for AI-facilitated cross-cultural philosophical dialogue across civilizational traditions - demonstrating that inclusive, multilingual, multi-epistemological governance engagement is architecturally achievable, not merely aspirationally desirable. These three levels constitute a governance stack. The Dialogue needs all of them.