A Proposal for Open Learning Practices in Response to Generative Artificial Intelligence

Synthesis: Boysen draws a direct parallel between two integrity crises. When repeated replication failures made psychologists doubt the credibility of their published results, the field responded with open science: preregistration, transparent analysis plans, shared data and materials, posted protocols, and standardized reporting. Higher education in 2025 faces the same kind of doubt about student work, because generative AI can produce the essays, literature reviews, presentations and data analyses that unsupervised assignments once treated as evidence of learning. Boysen's proposal translates each open science practice into an analogous "open learning practice": a shared plan for learning set before a project begins, a learning analysis plan that defines required, allowable and forbidden sources including AI, shared copies and annotations of the sources students actually read, documented drafts and version histories that expose the work process, and final reporting that combines citations, an AI contribution statement and a portfolio. The aim is not to police or detect AI but to relocate evaluation from a final product that can be counterfeited to a documented process that can be observed, so that judgments about integrity rest on visible evidence rather than on unreliable detection.

Key Findings

1. Open science is the template, not a metaphor. The Center for Open Science's Transparency and Openness Promotion Guidelines exist to "increase the verifiability of empirical research claims"; Boysen argues teachers can adopt analogous practices to increase the verifiability of student learning, because both fields face a collapse of trust in unsupervised outputs.
2. Both existing responses keep the focus on products. Banning AI requires technology-free tasks or AI-detection scanning, and "all current methods of AI detection are susceptible to false positives and false negatives"; embracing AI by rewriting learning outcomes to include it still evaluates the finished artifact. The proposal targets the documented process instead.
3. Preregistration becomes a shared plan for learning. Set before students start, it lists every required project step and its timing, extends the Transparency in Learning and Teaching (TILT) advice about purpose, tasks and success criteria, and is authored primarily by the teacher.
4. An analysis plan governs sources, including AI. Teachers must state which AI uses are required, allowable and forbidden, and students document use directly rather than only disclosing it at the end, for example by exporting full chat transcripts with metadata.
5. Data transparency means sharing the actual sources. Students submit links or PDFs, share libraries built in Mendeley or Zotero, expose highlights and annotations through tools such as Perusall, and can record the databases, search terms and limiters used to find sources.
6. Protocol transparency means showing the work process. Brainstorming, outlines, drafts and revisions are preserved through automatic version histories in OneDrive and Google Drive, Track Changes, saved file versions, and explanatory statements naming what informed each revision. Boysen notes this process tracking is what some critics call AI surveillance.
7. Reporting transparency closes the loop. Final work combines conventional citations, an AI contribution statement explaining where AI shaped the product, and a portfolio that gathers evidence of every open practice so teachers can review and give Feedback at any point in the project.
8. The author is candid about cost. Adoption requires new technology, revised assignments and syllabi, teaching students unfamiliar skills, and extra material to collect and grade; low-stakes credit for process can inflate grades, and the model fits large skill-development projects far better than day-to-day knowledge-acquisition activities.

The problem: a final product cannot be attributed

Boysen opens with Kahneman's 2012 warning that social psychology had become "the poster child for doubts about the integrity of psychological research", and its remedy, transparency in scientific practice. He then maps the situation onto higher education in 2025. The traditional contract of college coursework is unsupervised work followed by submission of a final product for evaluation: the generic research paper, but also presentations, unproctored exams, online discussions, literature reviews and data analyses. For all of these, generative AI can now do most of the work, so teachers "can no longer assume that unsupervised student work is an accurate reflection of their ability to achieve traditional learning goals". The quintessential artifact of Assessment no longer carries reliable information about the student who submits it.

The two obvious responses both fail on his account. Banning AI means designing technology-free assignments, closed-book tests, handwritten essays or oral exams, which pose practical difficulties and do not suit every learning goal, or surveilling work with style reading and AI-detection software that is prone to both false positives and false negatives. The alternative, embracing AI as an essential skill, adjusts outcomes so students may use AI, but this too evaluates a product. Boysen's diagnosis is that both approaches remain product-centered, whereas the AI era "calls for a different approach, one that is focused on documentation of the learning process".

The open science analogy and its limits

The analogy is structural rather than rhetorical. Each open science practice exists to make a stage of research inspectable: preregistration makes the plan inspectable before data collection, an analysis plan fixes statistical choices in advance, data and code sharing lets others reproduce findings, protocol and material posting clarifies procedure, and reporting standards such as the APA Journal Article Reporting Standards prescribe what a manuscript must contain. Boysen proposes that the same moves make learning inspectable, and presents a table pairing each science practice with its learning equivalent and the methods that make it transparent.

The limits are worth naming, though Boysen presses the parallel hard. Open science opens work to an external community of peers who can independently reanalyse; open learning opens work mainly to one teacher, and the "data" are a student's sources and drafts rather than a dataset another person could reanalyse to a different conclusion. Boysen also flags a terminological trap in a footnote: his "open learning practices" are derived from open science and are "not directly related" to the older sense of open learning or to open educational practices. The practices he proposes are an integrity device for Assessment, not a movement about open access to course materials.

The practices and what they ask of teachers and students

For teachers, adoption is front-loaded. Preregistration requires writing a plan that enumerates project steps and due dates alongside the final requirements; the analysis plan requires an explicit statement of acceptable sources and an explicit description of AI uses that are required, allowable and forbidden; grading requires deciding how much structure to impose and how to score compliance. Boysen sketches the range: at one extreme a teacher stipulates every detail, at the other a teacher simply asks for a portfolio at the end. Evaluation can be a prerequisite for earning any grade on the work, a specifications-style credit for meeting minimum requirements, or a rubric-scored element when the practices themselves are the learning goal.

For students, the demands are equally concrete and amount to a different way of working. They export AI chat transcripts, maintain shared reference libraries, annotate what they read, save successive drafts with version history intact, and explain what informed each revision, whether teacher comments, a writing center, grammar tools or an AI review. Boysen frames this as self-regulation made visible, and argues the documented steps also reduce procrastination and the incentive to take expedient shortcuts. The shared plan reframes the teacher as a partner who reviews work in progress rather than an examiner who receives a finished object.

Objections, feasibility and what the author concedes

Boysen does not oversell the model. He concedes it cannot eliminate cheating, only make integrity verifiable when work looks suspect, because the documents supplied under open practices are exactly what a teacher needs to check a submission's history. He concedes the workload: new technology, redesigned assignments and syllabi, teaching students skills they may resist, and more material to collect and grade. He concedes that low-stakes credit for process may contribute to grade inflation, and that the approach is "mainly relevant to larger projects aimed at skill development and are less applicable to day-to-day course activities related to knowledge acquisition". He also reports without rebutting the criticism that version-history monitoring functions as surveillance, which sits uneasily with the developmental framing.

His affirmative case rests on three claims: students learn research transparency by enacting it; they build transferable technology, project-management and collaboration skills; and the framing is learning-focused rather than punitive, letting teachers intervene early when a process looks wrong, before a final submission turns a teaching moment into a disciplinary case. The variation section is his answer to feasibility objections, since no research study uses every open science practice and no assignment need use every open learning practice. Institutional conditions are largely left implicit: whether policy and systems support this kind of change is not the paper's subject.

What this means for practice

• Instructors. Publish a shared plan for learning before a project begins by enumerating every required step and its timing alongside the final requirements, extending the Transparency in Learning and Teaching advice about purpose, tasks and success criteria.
• Instructors. Write an analysis plan that states which AI uses are required, allowable and forbidden, and have students document use as they go by exporting full chat transcripts with metadata rather than disclosing it in a statement at the end.
• Instructors. Require and grade process artifacts: shared reference libraries and annotations (Mendeley, Zotero, Perusall) plus preserved version histories in OneDrive or Google Drive, scored as low-stakes credit for meeting minimum requirements, with the portfolio as the record.
• Curriculum designers. Scope the model to larger skill-development projects. Boysen states the practices are "less applicable to day-to-day course activities related to knowledge acquisition" and that adoption costs teachers new technology, redesigned assignments and syllabi, and more material to collect and grade.
• Administrators. Set expectations for the two costs the author concedes before rolling this out: low-stakes credit for process may contribute to grade inflation, and version-history monitoring is criticized as surveillance, so consent and privacy terms for process data should be explicit.

Limitations

This is a position paper, not an empirical study. Boysen offers a conceptual translation and a table of practices; he reports no classroom implementation, no student outcome data, no comparison against the banning or embracing alternatives, and no evidence that documentation actually deters AI misuse or improves learning. The benefits he lists, reduced cheating motivation, better skill development, learning-focused intervention, are plausible mechanisms rather than demonstrated effects. The disadvantages are likewise asserted rather than measured, including the workload estimate and the grade-inflation concern. Feasibility questions the paper does not resolve include student privacy and consent when process data are collected, the position of students without reliable access to versioned cloud tools or AI platforms, and whether the intensive documentation model scales across a full teaching load. The proposal is best read as a framework inviting research rather than a validated intervention, and its value now is the reframing it offers: from detecting AI in products to observing learning in processes.

Connected Concepts

• Academic Integrity — the crisis of trust the proposal answers, and the target of its documentation model
• AI Detection — the surveillance alternative Boysen rejects as unreliable in both directions
• AI Use and Disclosure Statements — AI contribution statements plus exported chat transcripts as the reporting layer
• Assessment — relocated from final product to documented process, which is the paper's central move
• Assessment Validity — the claim that unsupervised work no longer supports inferences about learning
• Authentic Assessment — process evidence, portfolios and revisions as the observable object of evaluation
• Formative Assessment — early, learning-focused intervention on work in progress rather than final judgment
• Generative AI — the capability that broke the assumption of attributable student work
• Higher Education — the sector-wide context of the proposed change
• Pedagogical Partnerships — the shared learning plan as a negotiated agreement between teacher and student
• Research Methods in AIED — the open science heritage the proposal borrows and the empirical work it invites
• Self-Regulated Learning — documented steps as visible regulation of the learning process

Connected Articles

• From authentic products to authenticated processes: a systematic conceptual review of authentic assessment in AI-rich — The product-versus-process reframing that Boysen's proposal operationalizes
• Beyond Detection: Redesigning Authentic Assessment in an AI-Mediated World — Moving past detection toward assessment that can attribute learning
• The End of Assessment? Disruption and Transformation in the Age of AI — How GenAI disrupts established assessment and what replaces it
• The Integrity of Psychology Assessments in the AI Age: A Critical Examination — Assessment integrity in psychology teaching, the disciplinary home of this proposal
• The Impact of Generative AI on Academic Integrity of Authentic Assessments Within a Higher Education Context — Linking generative AI, authentic assessment and integrity
• Purpose Before Policy: Academic Integrity, Generative AI, and Rhetorical Stance — Purpose-driven integrity practice ahead of policy mandates
• "Should I Tell My Teacher?" Student AI Disclosure Practices, Stigma, and Self-Regulated Learning in Higher Education — Student disclosure decisions as the reporting-transparency problem
• The Hidden Cost of Disclosure: A Multi-institutional Study on Undergraduate Students' Generative AI Usage and Faculty Accusations — The friction and costs hidden inside AI disclosure requirements
• Students' Perceptions of Multiliteracies Development Using AI-Assisted Portfolio Assessment — Portfolios as evidence of learning process in AI-mediated work
• Beyond ID Embeddings: Process-Grounded Language Modeling for Cognitive Diagnosis — Diagnostics grounded in the process rather than the product

Citation

Boysen, G. A. (2026). A proposal for open learning practices in response to Generative Artificial Intelligence. Scholarship of Teaching and Learning in Psychology.