AI and problem-based learning

When AI can produce the artifact, what is the work?


Michael Rowe · University of Lincoln

Learning (used to) leave hard-to-fake traces

To write a good essay or produce a quality report, you had to:

  • Find and access the relevant literature
  • Read and synthesise across multiple sources
  • Organise and articulate an argument
  • Produce a coherent written artifact

The amount of effort required was a reasonable proxy for depth of engagement with the material.

The system made sense. And it wasn't wrong.

We can't directly observe learning

So we observe behaviour and infer understanding.

Engagement is a genuinely useful proxy. (Kuh, 2009) Artifacts — essays, reports, presentations — were meant to be evidence of that engagement. A window into the learner's developing understanding.

But the artifact became the target. Assessment shifted, gradually and imperceptibly, from asking "is this student developing their understanding?" to asking "how good is this product?" (Carless, 2007; Fischer, Bearman & Boud, 2023)

The focus should be on who you're becoming, not what you're producing.

AI separates production from learning

AI has removed one layer of that observation system.

If a student can produce a polished artifact without the engagement the artifact was meant to evidence, the proxy breaks. (Corbin, Dawson & Liu, 2025; Dawson, Bearman & Dollinger, 2024; Nikolic, Sandison & Haque, 2024)

The question isn't "how do we stop students using AI to produce artifacts?"

It's "what does meaningful engagement look like today?"

The learning hasn't gone anywhere

Engagement hasn't disappeared. We can still observe students:

  • Grappling with problems and testing ideas against each other
  • Asking questions that reveal the limits of their understanding
  • Revising their reasoning in response to challenge
  • Challenging each other's assumptions
  • Seeking connections between course concepts and real-world complexity
  • Reflecting on what they don't yet know

AI asks us to be more deliberate about what we're looking for, and how to make it visible.

The alternative is harder than it sounds

Assessing the learner rather than their outputs is genuinely difficult at scale.

Portfolios, oral examinations, observed practice, reflective accounts are partial answers with real limitations: (Masters, MacNeil & Benjamin, 2025)

  • Hard to scale: resource-intensive to design, run, and assess consistently
  • Hard to standardise: subjectivity in marking raises equity concerns
  • Hard to protect: even oral examinations can be rehearsed to the point of inauthenticity

This is unsolved. The direction of travel is clearer than the destination.

Playing the game institutions designed

When the rules reward a product, rational agents optimise for the product.

Institutions set the rules: submit an artifact, receive a grade. Students figured out how to play that game well (Walton, Bearman & Crawford, 2025) but AI changed the cost of playing. (Corbin, Dawson & Nicola-Richmond, 2025)

This isn't an integrity problem. It's an incentive and validity problem; and it's one that was always there.

What PBL does differently

PBL starts with different assumptions

Most curricula start with content, then look for problems to apply it to. PBL inverts that.

  • The problem is the starting point, not the content
  • Students direct their own inquiry rather than following predetermined pathways
  • Knowledge is constructed collaboratively, not transmitted individually
  • The facilitator holds the process, not the answers
  • What matters is how students reason and engage, not just what they produce

Savery (2006)

PBL already shifted the focus

Control was already distributed to the student and the group, guided by the demands of the problem. (Norman & Schmidt, 1992; Schmidt, Vermeulen & van der Molen, 2006)

AI doesn't threaten that structure. It extends what's available within it.

PBL is not the solution to the challenge AI poses to education. But it is solution-shaped
PBL centres engagement over artifact quality, distributes authority to the learner, and makes reasoning visible, which is what this moment demands.

Problems that become possible

The goal isn't to make problems easier. It's to make harder problems accessible.

AI reduces the access barriers that previously kept certain problems out of reach — not the cognitive demand that makes engaging with them worthwhile. Productive struggle remains; what changes is how far into a problem students can go. (Bjork & Bjork, 2009; Lodge, Yang & Furze, 2023)

The question isn't what AI replaces in PBL. It's what problems become possible when access barriers shrink.

What changes when AI joins the group

AI as a member of the group

AI can do more than generate content.

  • Thinking partner: challenges reasoning, offers counterarguments, asks probing questions (Sharples, 2023)
  • Tutor: explains concepts from multiple angles, adjusts to the learner's level (Jurenka et al., 2024)
  • Feedback provider: identifies gaps in argument, suggests alternatives (Henderson, Bearman & Chung, 2025; Jensen, Bearman & Boud, 2025)
  • Research assistant: surfaces relevant ideas and connections across a problem space (Wagner, Lukyanenko & Paré, 2022)
  • Devil's advocate: stress-tests positions, surfaces assumptions, challenges epistemic confidence (Yan, Pammer-Schindler & Mills, 2025)

Example: water contamination

A group of engineering and biology students working on contamination in a local community.

  • One student asks AI to summarise the relevant environmental chemistry
  • Another uses it to find case studies from similar communities
  • A third asks it to take the perspective of a local health official and challenge the group's proposed solution

The group isn't outsourcing their thinking. They're extending their reach into a problem too complex for any of them to hold alone. (Yan, Pammer-Schindler & Mills, 2025)

Keeping the learner in control

The concern is legitimate: if AI is doing so much, where is the student's thinking?

The answer lies in who controls the context. Context sovereignty: the learner maintains agency by controlling what they bring to the interaction. (Rowe & Lynch, 2025)

  1. Persistent understanding: interactions build rather than reset
  2. Individual agency: the learner controls what context is provided
  3. Cognitive extension: AI amplifies the learner's intent; it follows rather than leads

The student as orchestrator

Directing one AI for one task is basic use. The more powerful move is when students are directing multiple agents, each focused on a different part of a complex problem. (Alfaro, Fiore & Oden, n.d.)

  • Higher-order skills operate at a level of complexity most traditional assessments don't reach
  • The student isn't just answering questions — they're setting them
  • Making judgements the AI cannot make is where the learning lives

PBL already trains this kind of thinking. (Messeri & Crockett, 2024)

Example: public health problem

Four agents, one inquiry

  • Epidemiological evidence
  • Community priorities and concerns
  • Cost-effectiveness critique
  • Ethical concerns not yet considered

The student's role

Synthesising across perspectives

Evaluating what matters and what doesn't

Making judgements the AI cannot make for them

The student isn't just using a tool; they're orchestrating an inquiry.

AI literacy is not about prompts

The skill is not prompt engineering.

It is context curation: the judgement to know what to bring to the interaction, what to trust, what to question, and what to do with what comes back. (Bearman, Tai & Dawson, 2024)

And increasingly, orchestration: knowing how to direct different kinds of AI support toward different parts of a problem.

What you can do this week

The facilitator test

If you'd be comfortable making the same request to your facilitator, you're probably fine making it to AI.

  • "Can you explain how the pathogen enters the cell?" → reasonable
  • "Can you help me find the gap in my argument?" → reasonable
  • "Can you write the methods section for me?" → not reasonable

The test works because it anchors AI use to the purpose of learning. Your facilitator helps you think — they don't think for you. That's the standard AI should meet too.

It doesn't resolve every situation. But it resolves many of them quickly.

Is AI making this harder or easier to think?

AI has made content generation easier, but content generation was never the point. (Gerlich, 2025; Yan, Pammer-Schindler & Mills, 2025)

Ask yourself:

  • Is AI removing cognitive effort that was the point of the task?
  • Or is it opening up complexity and challenge you couldn't access before?

If AI is pushing you to think more carefully and engage with problems beyond your current reach, you're probably using it well.

Shift 1: bring your context, not just your question

Before asking AI about your PBL problem, give it your thinking first:

  • Write what you currently understand about the problem
  • Note what you're uncertain about
  • State what you're trying to figure out

Give AI your thinking, not a request for a solution. Notice how differently it responds.

Shift 2: use AI as a thinking partner, not an answer machine

The response is the beginning of a conversation, not the end of a search. (Sharples, 2023)

After AI responds, don't stop:

  • Ask it what the weaknesses are in its own answer
  • Ask it for a counterargument
  • Ask it what it's not telling you

Shift 3: try directing, not just asking

Assign AI a specific role relevant to your problem:

  • A sceptical reviewer
  • A domain expert
  • A community member affected by the issue

The same question gets different responses depending on role and context. This is the beginning of orchestration, a skill worth developing deliberately.

What this moment makes possible

AI has collapsed the barriers to execution. Finding, synthesising, and articulating are no longer the hardest parts of working with knowledge. (Noy & Zhang, 2023)

What remains is judgement: deciding what problems matter, evaluating competing approaches, determining whose interests are served. (Bearman, Tai & Dawson, 2024)

That judgement is what PBL was designed to develop.

The focus should be on who you're becoming, not what you're producing.

AI extends your reach into that becoming. It doesn't replace it.

Thank you

Michael Rowe · mrowe@lincoln.ac.uk

References (1)

Alfaro, G. D., Fiore, S. M., & Oden, K. (n.d.). Externalized and extended cognition: Cognitive offloading for human-machine teaming.

Bearman, M., Tai, J., & Dawson, P. (2024). Developing evaluative judgement for a time of generative artificial intelligence. Assessment & Evaluation in Higher Education.

Bjork, E. L., & Bjork, R. A. (2009). Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning. In Psychology and the Real World.

Carless, D. (2007). Learning-oriented assessment: Conceptual bases and practical implications. Innovations in Education and Teaching International.

Corbin, T., Dawson, P., & Liu, D. (2025). Talk is cheap: Why structural assessment changes are needed for a time of GenAI. Assessment & Evaluation in Higher Education.

Corbin, T., Dawson, P., & Nicola-Richmond, K. (2025). 'Where's the line? It's an absurd line': Towards a framework for acceptable uses of AI in assessment. Assessment & Evaluation in Higher Education.

References (2)

Dawson, P., Bearman, M., & Dollinger, M. (2024). Validity matters more than cheating. Assessment & Evaluation in Higher Education.

Fischer, J., Bearman, M., & Boud, D. (2023). How does assessment drive learning? A focus on students' development of evaluative judgement. Assessment & Evaluation in Higher Education.

Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies.

Henderson, M., Bearman, M., & Chung, J. (2025). Comparing generative AI and teacher feedback: Student perceptions of usefulness and trustworthiness. Assessment & Evaluation in Higher Education.

Jensen, L. X., Bearman, M., & Boud, D. (2025). Feedback encounters in doctoral supervision: The role of generative AI chatbots. Assessment & Evaluation in Higher Education.

Jurenka, I., Kunesch, M., & McKee, K. (2024). Towards responsible development of generative AI for education: An evaluation-driven approach.

References (3)

Kuh, G. D. (2009). The national survey of student engagement: Conceptual and empirical foundations. New Directions for Institutional Research.

Lodge, J. M., Yang, S., & Furze, L. (2023). It's not like a calculator, so what is the relationship between learners and generative artificial intelligence? Learning: Research and Practice.

Masters, K., MacNeil, H., & Benjamin, J. (2025). Artificial intelligence in health professions education assessment: AMEE Guide No. 178. Medical Teacher.

Messeri, L., & Crockett, M. J. (2024). Artificial intelligence and illusions of understanding in scientific research. Nature.

Nikolic, S., Sandison, C., & Haque, R. (2024). ChatGPT, Copilot, Gemini, SciSpace and Wolfram versus higher education assessments. Australasian Journal of Engineering Education.

Norman, G. R., & Schmidt, H. G. (1992). The psychological basis of problem-based learning: A review of the evidence. Academic Medicine.

References (4)

Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science.

Rotgans, J. I., & Schmidt, H. G. (2011). Cognitive engagement in the problem-based learning classroom. Advances in Health Sciences Education.

Rowe, M., & Lynch, C. (2025). Context sovereignty for AI-supported learning: A human-centred approach.

Savery, J. R. (2006). Overview of problem-based learning: Definitions and distinctions. Interdisciplinary Journal of Problem-Based Learning.

Schmidt, H. G., Vermeulen, L., & van der Molen, H. T. (2006). Longterm effects of problem-based learning: A comparison of competencies acquired by graduates of a problem-based and a conventional medical school. Medical Education.

Sharples, M. (2023). Towards social generative AI for education: Theory, practices and ethics. Learning: Research and Practice.

References (5)

Wagner, G., Lukyanenko, R., & Paré, G. (2022). Artificial intelligence and the conduct of literature reviews. Journal of Information Technology.

Walton, J., Bearman, M., & Crawford, N. (2025). How university students work on assessment tasks with generative artificial intelligence: Matters of judgement. Assessment & Evaluation in Higher Education.

Yan, L., Pammer-Schindler, V., & Mills, C. (2025). Beyond efficiency: Empirical insights on generative AI's impact on cognition, metacognition and epistemic agency in learning. British Journal of Educational Technology.