This page carries the detail behind our first focus area. For the argument in brief, start with Knowledge & Intelligence.
Our Research Approach
At the Society & AI Research Group, we investigate knowledge and intelligence through three interconnected research streams:
1. Cognitive and Learning Sciences
We study how AI tools affect individual learning processes—memory formation, attention, problem-solving strategies, and metacognition. Our work documents both opportunities and risks:
Opportunities:
- Reduced cognitive load through intelligent tutoring that adapts to learner readiness
- Enhanced working memory through AI assistants that hold context across sessions
- Access to multiple representations of concepts, supporting diverse learning styles
- Scaffolded progression through material calibrated to individual trajectories
Risks:
- Erosion of deep processing when learners skip the productive struggle that builds mastery
- Over-reliance on algorithmic suggestions that undermine independent thinking
- Diminished transfer when learning becomes context-specific to particular AI interfaces
- Loss of metacognitive awareness as AI systems make thinking processes invisible
Our empirical studies track these effects across diverse learner populations, attending particularly to how they differ by prior academic preparation, cultural background, and access to skilled teaching. We report the conditions under which each pattern appears, because we do not think these effects generalize cleanly across settings.
2. Epistemology and Knowledge Systems
We analyze the implicit theories of knowledge embedded in AI educational tools. Every system makes assumptions about:
- What constitutes understanding versus mere performance
- How expertise develops and can be recognized
- Which forms of knowledge are valuable and which are peripheral
- Whether intelligence is a fixed trait or a developable capacity
These assumptions are rarely neutral. They reflect the values and worldviews of system designers, the constraints of training data, and the optimization metrics that define “success.” When these assumptions conflict with educational goals—when systems optimize for engagement rather than learning, or privilege speed over depth, or mistake correlation for causation—they can actively undermine teaching and learning.
We document these conflicts and develop frameworks for epistemic alignment: ensuring that AI systems embody educational values rather than subverting them. This includes creating evaluation protocols that test whether systems:
- Support multiple pathways to understanding, not just procedural competence
- Recognize and value diverse forms of expertise, including community knowledge
- Make their reasoning transparent so learners develop critical evaluation skills
- Acknowledge uncertainty and limitation rather than projecting false confidence
3. Social Dimensions of Knowing
Knowledge is not created in isolation. It emerges through social processes: dialogue, debate, peer review, community validation. AI changes these processes in ways we are only beginning to understand.
Our research examines how AI redistributes epistemic authority:
- When students turn to chatbots for explanations, how does this affect their relationship with teachers and peers?
- When teachers rely on algorithmic dashboards for assessment, how does this change what gets noticed and valued?
- When administrators use predictive models for resource allocation, whose knowledge gets counted and whose gets dismissed?
We study these dynamics not to resist change reflexively, but to establish whether emerging configurations of authority serve equity, transparency, and democratic accountability—and to say so when they do not. This work includes ethnographic study in schools, interviews with educators navigating AI integration, and analysis of how power operates through arrangements that present themselves as merely technical.
Real-World Applications and Implications
The transformation of knowledge and intelligence has immediate, practical consequences:
In K-12 Education
Teachers experimenting with AI tutors report that students receive more individualized support—but also that they struggle to distinguish when AI guidance helps versus hinders. We’ve documented cases where:
- Elementary students use AI to explore mathematical concepts through multiple visual representations, building deeper number sense
- Middle schoolers employ AI writing assistants that provide real-time feedback, improving revision processes
- High school students use AI to reach scientific literature that was previously inaccessible to them, engaging with current research
But we’ve also seen:
- Students gaming systems to get answers rather than building understanding
- AI-generated feedback that is plausible but pedagogically misaligned
- Widening gaps between students with access to high-quality AI tools versus those limited to free, less capable systems
In Higher Education and Research
University researchers now routinely use AI to accelerate literature reviews, generate hypotheses, analyze large datasets, and even draft initial paper sections. This amplifies scholarly productivity—but also creates challenges:
- How do we maintain standards for originality when AI can synthesize existing work so fluently?
- What happens to the serendipitous discoveries that emerge from slow, immersive engagement with a field?
- How do we ensure that AI augments rather than automates the creative, intuitive leaps that drive breakthrough science?
Our work with faculty members across disciplines explores these tensions, developing protocols for responsible AI use that preserve scholarly integrity while embracing legitimate productivity gains.
In Professional Learning and Workforce Development
Knowledge work is being transformed. Professionals who once spent hours searching for precedents, regulations, or best practices can now query AI systems instantly. This frees time for higher-level judgment and creative problem-solving—but only if workers develop the literacy to use these tools critically.
We partner with professional organizations to create training programs that build AI-augmented expertise: the capacity to collaborate effectively with computational systems while maintaining professional judgment, ethical responsibility, and domain mastery.
How Society & AI Addresses This
Our approach to knowledge and intelligence research is guided by several commitments:
1. Augmentation Over Automation
We design and advocate for AI systems that amplify human judgment rather than replace it. In educational contexts, this means:
- Teachers remain the primary decision-makers about pedagogy and assessment
- AI provides suggestions and feedback, but humans always review consequential judgments
- Systems are designed to make reasoning transparent so users can evaluate and challenge outputs
2. Epistemic Justice
We center marginalized voices and knowledge systems that dominant educational structures have historically dismissed. This includes:
- Testing AI systems across languages, dialects, and cultural contexts to identify bias
- Developing frameworks for Indigenous data sovereignty so communities control how their knowledge is represented
- Advocating for training data and algorithms that recognize multiple forms of expertise
3. Methodological Rigor
We employ mixed methods—quantitative analysis, qualitative inquiry, design-based research, and systems modeling—to capture the multidimensional nature of knowledge creation. We publish our protocols, share data (where ethically permissible), and invite replication.
4. Practical Translation
We translate scholarly insights into frameworks, toolkits, and design principles that educators, policymakers, and technologists can implement. Our aim is to close the gap between research and practice so that insights generate actionable change, not merely publications.
5. Open Scholarship
We publish under open-access licenses and make our tools freely available. Knowledge generated through public-interest research should return to the public, not be locked behind paywalls or proprietary restrictions.
Return to Knowledge & Intelligence, or see the other focus areas: Systems & Complexity, Educational Equity, Human Flourishing.