Questions and method
Research
AIPLA investigates how generative AI can support learning and assessment in upper-secondary physics without replacing the reasoning, experimentation, and communication students need to practise themselves.
The unit of analysis is classroom practice: the activity, the teacher's intentions, the representations students use, the interaction within a student group, and the role played by the AI system all matter together.
Four practical research contexts
Solving physics problems
Can an AI tutor offer timely questions, hints, and representations without simply producing the answer? The project studies how tutors can help students identify relevant concepts, connect equations with physical situations, check assumptions, and reflect on a solution.
This includes exam-style problems, but the goal is not an automated answer service. A useful tutor should help students make progress while leaving the important intellectual decisions with them.
Carrying out experiments
Physics knowledge is practical as well as conceptual. Students learn through equipment, measurement, uncertainty, troubleshooting, data processing, and comparison between a physical system and its model.
AIPLA explores whether AI can support experimental preparation and interpretation without displacing contact with the physical situation. Possible roles include asking students to predict a measurement, helping diagnose an inconsistent result, or prompting reflection on uncertainty and model limitations.
Exploring concepts
AI can act as a conversational counterpart when students make predictions, compare representations, or explain a phenomenon. The research asks which conversational patterns help students articulate and revise their understanding.
Conceptual dialogue may be paired with a simulation, graph, concept map, or other workbench. The visible representation gives the conversation something concrete to refer to and test.
Preparing presentations
Oral and written communication are part of Danish physics education and assessment. AIPLA investigates appropriate roles for AI when students plan, explain, critique, and revise physics presentations.
The distinction between assistance and substitution is especially important here. An activity may use AI to question an explanation or identify an unsupported claim while still requiring students to construct and defend the account themselves.
Cross-cutting research questions
Across those four contexts, the project asks:
- Which tasks are neither trivial for AI nor impossible for it to support productively?
- Which scaffolds lead students to reason rather than wait for an answer?
- How do equations, diagrams, graphs, simulations, experimental evidence, and verbal explanations work together in an AI-supported activity?
- How much control and transparency do teachers need when configuring an activity?
- How do students understand the AI's role, limitations, and use of their input?
- What forms of evidence are appropriate for formative and summative assessment?
- How do different students and groups experience the same design?
Principles guiding the work
Learning before automation
Every activity begins with a learning purpose. AI is included only where it can contribute to that purpose, and repetitive work may be automated only when doing so creates space for more valuable reasoning or experimentation.
Teacher control
Teachers choose the activity, source material, learning goals, and tutor configuration. Student sessions take place within that prepared context rather than in an unrestricted chatbot.
Physics-specific design
The project treats equations, diagrams, graphs, experiments, simulations, physical models, and verbal explanations as connected representations. Tutor behaviour is designed around those relationships.
Productive uncertainty
AI outputs can be incomplete, misleading, or physically incorrect. Activities should not hide that limitation. In some cases, evaluating a plausible but incorrect AI representation can itself become the learning task.
Responsible research practice
Data protection, transparency, accessibility, consent, and clear boundaries for AI-generated content are part of the design process. Technical capability is not sufficient evidence that a classroom use is appropriate.
Research material and instruments
Depending on the study and its approvals, research material may include:
- teacher interviews, design dialogues, and workshop feedback;
- classroom observations;
- activity configurations and teaching materials;
- group-based student–AI interaction records;
- student-created representations or submitted work;
- participant perspectives gathered through interviews or questionnaires; and
- pre/post or task-specific measures of physics understanding.
The exact material collected is determined by the approved study design. Public platform documentation does not replace participant information, consent procedures, or institutional review.
From prototypes to research findings
Early prototypes deliberately cover a broad range of activity types. They are used to discover which combinations of interface, tutor behaviour, and classroom task deserve deeper study.
The project then narrows its focus through repeated ADDIE cycles. A working feature is not automatically a research result: findings require an explicit design, appropriate evidence, analysis, and review. This distinction is also why technical progress and published research are reported separately.
Expected research contribution
AIPLA aims to contribute empirically grounded knowledge about when and how generative AI can support physics learning and assessment. It also aims to create practical designs and resources that teachers can adapt, while making the limitations and conditions of those designs visible.
Publications and research outputs will be added to this site as they become available.
- Content status
- Current
- Maintained by
- AIPLA research team
- Last reviewed