A New Tool to Support Service Learning in Liberal Arts Colleges
The Community Knowledge Encoder inverts the usual relationship between AI and education. Instead of students consuming what AI generates, students teach the AI what their community knows.
Knowledge base depth — the system prompt is the community's institutional memory
The Community Knowledge Encoder: A New Model for Civic Learning
A scalable framework for Liberal Colleges to integrate AI, service learning, and civic infrastructure
Something has changed about how we learn and how democracies are threatened
AI is changing what a finished product looks like
When a student submits an essay today, it may have been written by an AI in under a minute. When they produce a report, analyse a dataset, or draft a policy memo, the same is true. The output looks exactly like learning. It may not be.
This is not primarily a problem of cheating. It is a deeper pedagogical crisis. If the point of an assignment is to help a student develop the capacity to think — to gather evidence, form a position, defend it, revise it — then a tool that short-circuits that process does not help the student learn. It produces the artefact of learning without the thing itself.
For liberal arts colleges, whose educational mission is precisely the formation of thoughtful, independent, civically capable human beings, this is an existential challenge. The tools are now good enough to convincingly simulate the products of a liberal arts education. Which raises an uncomfortable question: what does a liberal arts education need to become, in response?
The submitted essay looks like learning. But it may not be. What the student lacks is not the output. It is the experience of having produced it.
AI is also transforming the information environment that democracy depends on
Outside the classroom, a different kind of AI problem is unfolding — one with higher stakes and faster consequences. Communities across the world are being targeted by what researchers now call AI-Enhanced Reflexive Control (AIRC): sophisticated, self-learning influence operations that do not primarily spread false information, but engineer the emotional and cognitive conditions under which certain conclusions feel inevitable.
These are not simple disinformation campaigns. They are adaptive systems continuously monitoring community responses, A/B testing narratives, updating their content based on what produces engagement, and growing progressively more precise in how they target specific communities. They know what a community fears. They know what it mourns. They know what cultural symbols carry emotional weight and how to weaponise them. A research by Jacob and Angelov (2025) on Kremlin-backed AIRC operations targeting Bulgaria, shows how AIRC builds a "cognitive settlement" around historical memory, cultural identity and collective grievance. By creating networks of false amplifiers that simulate community and consensus, AIRC operations tap into the fundamental human desire to find our place in a coherent social and cognitive landscape.
The problem inside the classroom and the problem outside it are the same problem
This is the insight that shapes everything in this proposal. In both cases, a powerful AI system is producing content that communities consume passively — without the tools to interrogate it, contextualise it, or recognise what it is doing to their thinking. In both cases, the community is positioned as a receiver of generated content rather than a protagonist in knowledge creation. And in both cases, the solution is the same: communities need to become active, equipped, and epistemologically capable agents in their own information environment.
Service learning is the pedagogical tradition that knows how to do this. And AI — specifically, community-encoded AI — is the tool that makes it possible at the scale and precision the moment demands.
Service learning as democratic infrastructure with AI as its instrument
What service learning already knows
Service learning restores what AI-generated content removes: the experience of genuine inquiry. In a well-designed service learning course, students do not consume knowledge, they produce it, through direct engagement with real communities, real problems, and real stakes. The assessment is not a proxy for learning. It is the learning, made visible.
Liberal arts colleges have built exceptional service learning infrastructure precisely because they believe that civic formation: the capacity to think carefully, act ethically, and engage responsibly with democratic life is inseparable from academic education. Service learning is where those two things meet.
What has been missing, until now, is a direct connection between service learning and the information integrity crisis that is the most urgent democratic challenge of this decade. Students have been learning to serve communities without being equipped to defend those communities against sophisticated information attacks.
Integrating AI into the mix — on the community's terms
The Community Knowledge Encoder is a web platform that enables students to encode their community's own knowledge — its vulnerabilities, its cultural touchstones, its economic anxieties, its linguistic signatures of manipulation — directly into an AI system. The AI then approaches that community's information environment as an informed insider rather than a generic outsider.
This is not AI literacy education. It is not a course about AI. It is civic education that uses AI as a practical instrument of community defence — one that students build, that their community owns, and that grows more capable with each passing semester.
In every other AI-in-education application, AI is the authority and students are the learners. In the encoder model, students become the epistemic authorities that the AI must defer to. The community's knowledge is not raw material for the machine. It is the intelligence that makes the machine worth using.
How the Living Epistemic Commons Works and Grows
The encoder is embedded in a required service learning course. Students are not introduced to it as a technology lesson. They encounter it as the tool through which their research becomes useful — the place where what they have learned about their community gets formalised into something the AI can work with.
What students actually do
Learn the AIRC research framework
Students study the AI-Enhanced Reflexive Control framework and develop the ability to identify coordination signatures, distinguish authentic community voice from its weaponised imitation, and map the specific vulnerabilities that manipulation operations exploit.
Conduct original field research
Students research how AIRC specifically operates in their community. They interview residents, monitor local media and social platforms, document economic anxieties and political tensions, identify cultural touchstones, and map the linguistic patterns of both authentic and coordinated content. This provides an opportunity for genuine ethnographic fieldwork.
Encode findings into the platform
Students enter their research into the Encoder's structured framework: community context; demographic knowledge; cultural knowledge; economic manipulation signatures; linguistic indicators; and analytical parameters (more can be added). Each completed section deepens the AI's understanding of that community. The system prompt that results is a formalised record of genuine field research.
Test, compare, and demonstrate divergence
Students submit real content from their community's information environment. The platform runs it through two AI instances simultaneously: generic and community-encoded. It then produces a divergence score showing how much community knowledge changed the analysis. High divergence is evidence that the research matters.
Pass the knowledge base forward
At the end of the semester, the knowledge base is saved and versioned. The next cohort inherits it, reviews what their predecessors built, tests it against new content, identifies what it misses, and deepens it with their own research. The community's AI gets smarter every year.
The divergence score — making community knowledge visible
The encoder's most distinctive feature is not technical — it is pedagogical. When a student submits a piece of content for analysis, the platform runs it through two AI systems simultaneously: one generic, one community-encoded. Both responses appear side by side. A divergence score between zero and one hundred quantifies how much community knowledge changed the analysis.
The growth of the system prompt across versions is the learning record. Faculty do not need an examination to assess whether a student has learned. The depth and sophistication of the community knowledge base they contributed to and the divergence scores it generates are the evidence.
This is the moment
AIRC operations are accelerating — and adapting faster than our defences
The same networks documented in Bulgaria are operating, in culturally adapted forms, in communities across the United States, exploiting genuine economic anxieties, weaponising authentic cultural identities, and embedding manipulation within the real grievances of real communities. The students in your service learning courses live in those communities. Their families are often the targets.
AIRC systems get smarter about communities over time. The tools we use to counter them do not. The Living Epistemic Commons is the first counter-system built on the same adaptive logic as the threat — accumulating intelligence, updating across iterations, growing more valuable with each semester. The adversary accumulates. Now the defence can too.
Liberal arts colleges are uniquely positioned — and uniquely responsible
Liberal arts education exists precisely to form people who can think carefully, act with moral seriousness, and engage responsibly with the conditions of democratic life. No other educational tradition is better placed to address this moment — because no other tradition has maintained the pedagogical infrastructure, the institutional commitment to civic formation, and the service learning frameworks that the encoder requires.
GCRD is convening a pilot cohort of up to ten liberal arts colleges for Fall 2026 and/or Spring 2027. We invite universities and colleges to join this first cohort.
What participating colleges contribute
Commitment of one faculty member to integrate the encoder into a required service learning course. Student cohort for the pilot semester. Institutional support for the community research process. Participation in the cross-cohort learning community.
What participating colleges receive
Full onboarding and faculty facilitation guide. Student orientation materials and AIRC research methodology training. A functional community AI platform, hosted and maintained. Access to the divergence dataset from all participating colleges. Colleges can also join the Democracy Discourse Index research consortium.
What students gain
Genuine civic research skills. The experience of producing knowledge that outlasts the semester and serves their community. A quantifiable, portfolio-ready record of epistemic contribution. And an encounter with AI that positions them as authors, not just consumers.
What communities gain
A locally-grounded, student-built AI resource for information analysis and civic engagement that grows more capable each year. Civic infrastructure owned by the community and maintained by its own students that is not dependent on external institutions or ongoing grant funding.
Why your college should join now
The question is not whether AI should be part of civic education. It already is but in ways that too often position students as users of generated content and communities as objects of analysis rather than agents of knowledge.
The Encoder enables liberal arts colleges to embed AI into service learning in a way that restores student agency, builds community epistemological capacity, and produces civic infrastructure that compounds in value across years. Course artefacts do not expire at the end of the semester but accumulates to strengthen the knowledge base.
The Encoder is designed to position students as protagonists in their own knowledge creation, not simply users of generated content.
I personally invite you to join the Fall 2026 or Spring 2027 pilot cohort. Contact me directly to request a demonstration of the encoder platform.
Reference
Jacob, J.U. & Angelov, G. (2025). The Disinformation Matrix. In J.U. Jacob & R. Narasimha (Eds.), AI and the Future of Democracy: Building Resilient and Inclusive Societies. London: Routledge (Taylor & Francis). DOI: 10.1201/9781003594185-2
The AIRC research has been presented at the European Parliament and in a NATO Cognitive Warfare training.
