This page carries the detail behind our third focus area. For the argument in brief, start with Educational Equity as Leverage.
The Compounding Returns to Learning
Education differs from other development interventions in a crucial way: its benefits compound over time and across generations. A road built today serves those who travel it; a school built today transforms communities for decades. Consider the mechanisms:
Individual Capability Multiplies Across Domains
An educated person does not simply know more facts. They develop transferable capability: the metacognitive skills to learn independently, the critical judgment to evaluate new information, the problem-solving frameworks to navigate complexity. These capabilities apply across every domain of life.
A woman who learns to read does not merely decode text. She gains access to health information bearing on her children’s survival, agricultural guidance that affects yields, financial literacy that makes saving possible, and civic knowledge that bears on how she is governed. Each capability opens others in sequence. The accumulation is not linear; it compounds.
Intergenerational Transmission Accelerates Progress
Educated parents invest differently in their children—not just financially, but cognitively and emotionally. They read to them, answer questions with questions, model curiosity, and create home environments rich in language and reasoning. These investments shape neural development during critical periods, creating advantages that persist throughout life.
A girl whose mother reads tends to enter school with a substantially larger vocabulary than peers whose parents do not—a gap that studies of early language exposure have measured repeatedly, though its size varies considerably by setting. She is more likely to complete secondary school, to delay marriage, to plan pregnancies, to seek prenatal care, and to see her own children educated. The cycle reinforces itself across generations, and trajectories between families and communities diverge accordingly.
Collective Capability Transforms Societies
At sufficient scale, educated populations fundamentally alter social dynamics. Democratic institutions require citizens who can evaluate policy proposals, detect demagoguery, and hold leaders accountable. Markets require workers who can adapt to technological change, entrepreneurs who can identify opportunities, and consumers who can assess quality. Civil society requires organizers who can mobilize collective action, journalists who can investigate power, and professionals who maintain ethical standards.
When a critical mass of people develop these capabilities, societies shift from extraction to investment, from clientelism to accountability, from tradition-bound rigidity to adaptive innovation. This is how education functions as a meta-intervention: it does not solve specific problems directly but creates the capacity for communities to solve problems themselves, including problems that do not yet exist.
Our Research Imperatives
We approach educational equity through four interconnected research streams. Each demands rigorous empirical investigation, and each demands that we report findings that are inconvenient to the positions we hold:
1. Mapping Access Barriers at Every Level
We document not just who lacks connectivity or devices, but the full ecology of barriers that prevent AI-enhanced learning from reaching marginalized populations:
Infrastructure gaps: Rural areas without reliable electricity. Communities where devices must be shared among dozens of students. Networks that cannot support video streaming or real-time interaction.
Linguistic exclusion: AI systems that function only in dominant languages, marginalizing the majority of the world’s population who speak other tongues. Automatic translation that strips away cultural nuance and reproduces stereotypes.
Disability inaccessibility: Platforms built without screen reader compatibility, without captions, without alternative input modalities. Systems that assume users have typical vision, hearing, mobility, and cognitive processing.
Cultural unresponsiveness: Content that centers Western examples, Western historical narratives, Western epistemologies as universal. Assessments that privilege individual competition over collective problem-solving, written expression over oral tradition, abstract reasoning over contextual knowledge.
Economic barriers: “Free” platforms that require premium subscriptions for essential features. Systems that lock schools into proprietary ecosystems with escalating costs. Platforms that monetize student attention through advertising or data sales.
Our mapping work makes visible what techno-optimism obscures: the compounding ways that AI can reproduce and amplify existing inequities when equity is not the foundational design commitment.
2. Designing for Universal Access
We develop and evaluate frameworks for building AI systems that work for everyone, not just the privileged:
Offline-First Architecture: Systems that can run without constant connectivity, caching resources locally and syncing when networks are available. This requires different design choices—lighter models, downloadable content, progressive enhancement—but makes learning accessible in contexts where bandwidth is a luxury.
Multimodal and Multilingual by Default: Not translation as an afterthought, but systems designed from inception to work fluidly across languages, scripts, and modalities. Voice interfaces for those with limited literacy. Visual representations for those learning in languages with limited digital resources. Cultural localization that goes beyond mere translation to respect different ways of knowing.
Radical Accessibility: Universal design principles embedded from the beginning—not retrofitted later. Every feature tested with assistive technologies. Every interaction designed to work with keyboard navigation, screen readers, voice control. Not because it is legally required but because it is morally imperative.
Community-Governed Data: Participatory frameworks where communities decide what data is collected, how it is stored, who can access it, and how value generated from it is shared. This requires shifting power from vendors to users, from extraction to co-governance.
Open Standards and Portability: Learners own their records and can move them across systems without vendor lock-in. Data formats are open, APIs are documented, and migration paths exist. This prevents monopolistic capture and ensures competition serves students, not shareholders.
3. Holding Systems Accountable to Equity Outcomes
Technology alone does not produce equity. It requires policy frameworks that create incentives for equitable design and consequences for inequitable outcomes:
Procurement Standards: Public institutions should purchase only systems that meet accessibility requirements, support multiple languages, work offline, protect privacy, and demonstrate equity impact through disaggregated data. This shifts market incentives toward equity.
Equity Auditing: Independent evaluation of AI systems across demographic groups, geographic contexts, and ability profiles. Performance gaps should trigger remediation or removal, not be dismissed as edge cases.
Weighted Funding: Resources should flow disproportionately to schools and students facing compounded disadvantages—not through charity but through recognition that equity costs more than maintaining privilege.
Teacher Capacity Investment: AI should reduce administrative burden to free teacher time for high-value human work. This requires professional development, protected collaboration time, and compensation that reflects expertise.
Community Voice: Families and students should have meaningful power in decisions about AI adoption—not consultation after decisions are made, but genuine co-governance over what systems are used and how.
4. Measuring What Matters for Flourishing
Standard metrics—test scores, graduation rates, college enrollment—capture only narrow slices of educational success. If we optimize AI systems solely for these metrics, we will get systems that game them while undermining deeper learning.
Our measurement work develops frameworks that assess:
Deep Understanding: Not recall of facts but the ability to transfer knowledge to new contexts, to explain reasoning, to identify misconceptions, to build coherent mental models.
Collaborative Capability: Not just individual achievement but the skills to work across difference, to build on others’ ideas, to navigate conflict constructively, to accomplish together what none could alone.
Wellbeing and Belonging: Do students feel safe, valued, and connected? Do they experience learning as meaningful? Are they developing agency and self-efficacy?
Equitable Outcomes: Are gaps closing or widening? Are students from marginalized groups gaining capability at rates that will enable them to overcome structural barriers? Are opportunities distributed fairly?
We disaggregate all data by race, language, disability, geography, and socioeconomic status because aggregate gains often mask widening inequities. A system that raises average performance while leaving the most vulnerable further behind is not successful—it is failing at its most important function.
Why This Matters: The Stakes for Humanity
Educational equity is not a regional concern, and it is not one policy preference among several. We treat it as a condition bearing directly on whether the challenges now in front of us can be met at all:
Climate Adaptation Requires Universal Capability: Every community will face disruptions—droughts, floods, heat, displacement. Response requires local problem-solving, resource management, collective action. If only elites have these capabilities, adaptation will be a catastrophe of exclusion.
Pandemic Preparedness Depends on Broad-Based Understanding: COVID-19 revealed how misinformation, distrust, and incapacity undermine public health. Future pandemics will arrive with less warning and more lethality. Survival requires populations that can evaluate evidence, coordinate behavior, and trust institutions—capabilities cultivated through education.
Democratic Governance Cannot Function Without Educated Citizens: Autocracy thrives on ignorance, tribalism, and the inability to think critically about power. Democracy requires populations that can deliberate, compromise, hold leaders accountable, and distinguish truth from manipulation. These are learned capabilities, not natural instincts.
Technological Governance Needs Distributed Expertise: AI, biotechnology, nanotechnology—the most powerful technologies humanity has created—cannot be governed wisely by small technical elites while the rest of society remains ignorant. Governance requires broad-based understanding of how systems work, what values they encode, and what futures they enable or foreclose.
Human Dignity Demands Capability: To be denied education is to be excluded from full participation in common life—rendered dependent rather than agentic, spoken for rather than deliberative, managed rather than capable. Where capability increasingly governs access to opportunity, educational exclusion is not a temporary disadvantage. It is a durable one, and it is passed down.
How Society & AI Addresses This
Our commitment to educational equity shapes every dimension of our work:
We Center Equity in All Research: Every study disaggregates by demographic groups. Every design includes accessibility from inception. Every policy recommendation considers impact on the most marginalized. Equity is not a separate workstream—it is the lens through which we evaluate everything.
We Prioritize Under-Resourced Contexts: We conduct research in rural schools, in multilingual communities, in regions without reliable connectivity, with learners who have disabilities—not as afterthoughts but as our primary focus. Solutions that work for the most constrained environments work everywhere.
We Build Capacity, Not Dependency: We do not parachute in with solutions. We work alongside educators and communities to develop context-appropriate approaches, building local expertise so interventions can be sustained and adapted without external support.
We Advocate for Systemic Change: Individual tools matter less than the policies, procurement standards, funding mechanisms, and accountability structures that shape what gets built and who benefits. We engage with policymakers, testify before legislatures, and collaborate with civil society to shift systems toward equity.
We Publish Openly: All research, frameworks, and tools are available under open licenses because knowledge generated in service of equity belongs to everyone, not behind paywalls protecting private profit.
Return to Educational Equity as Leverage, or see the other focus areas: Knowledge & Intelligence, Systems & Complexity, Human Flourishing.