14 Why Process-Oriented Assessment Works Well in the AI Era

How This Chapter Fits in Part 2

By this point in Part 2, faculty have clarified outcomes, mapped secured and open work, set assignment-level AI expectations, and redesigned selected tasks for authenticity and access. This chapter focuses on time. Process-oriented assessment asks students to show how their work develops through drafts, feedback, revision, AI transparency, and reflection.

The final chapter, AI Detection and Accountability, will address detection and accountability. Process evidence gives that conversation a better foundation because it creates more ways to discuss learning before suspicion becomes the main lens.

From Product to Process

A student submits an essay. It is well-organized, properly cited, free of errors. It meets every stated requirement. But you cannot tell whether the student wrote it, edited it, or prompted AI to generate it.

That gap points to a real limitation of product-only assessment. When we evaluate only the final deliverable, we lose visibility into how it was made. And the problem runs deeper than academic integrity. Leon Furze argues that when schools prize product over process, generative AI flattens the entire writing cycle. The student skips the messy, generative thinking that writing exists to produce. In education, the process is the learning.

This is a validity problem. When a final product no longer reliably indicates what a student knows or can do, the assessment itself stops measuring what it claims to measure. Furze makes this point directly. Process assessment is one response because it adds evidence about work completed along the way.

The Role of Authentic Assessments and UDL introduced the concept of cognitive offloading, where students hand over the thinking that constitutes learning to an AI tool. In a 2026 working paper, Steven D. Shaw and Gideon Nave use cognitive surrender for a related pattern: accepting AI advice with minimal scrutiny. Their controlled reasoning experiments found that participants frequently followed AI advice even when it was deliberately wrong.

Fernandes et al. (2026) studied U.S. adults completing LSAT-style logical-reasoning tasks. Participants using AI performed better, while their self-assessments remained poorly calibrated. In the randomized second study, both the AI and no-AI groups overestimated their performance, so the study does not show that AI alone caused worse monitoring. Process-oriented assessment gives students occasions to explain and check their reasoning.

The idea itself is old. Writing programs have emphasized process since the 1970s. AI has made process visibility relevant across more disciplines because a finished product supplies limited evidence about how it was produced. Anna Mills documented an agent navigating a learning management system and attempting multi-step assignment workflows. The example shows a current capability, though it does not establish how common or reliable that capability is.

The Role of Authentic Assessments and UDL also drew a distinction between structural interventions (which change what students must do) and discursive measures (which change what students are told). Process-oriented assessment is a structural intervention. It changes the assignment itself so that evidence of learning is built into the workflow.

This approach aligns with TEQSA’s assessment-reform principles (PDF), which recommend multiple, inclusive, contextualized approaches to judging student learning.

What Is Process-Oriented Assessment?

Process-oriented assessment shifts the focus from “what was produced” to “how it was produced” and “what was learned.” Students submit intermediate artifacts alongside their final work. These might include proposals, drafts, feedback responses, revisions, and reflections. The grade accounts for the full arc of that work, from early thinking through to final product.

Research on formative assessment connects feedback cycles to self-regulated learning, while feedback-literacy research emphasizes students’ capacity to make judgments and act on feedback. A meta-analysis of 54 studies found a small-to-medium positive effect of peer assessment on academic performance across educational levels and subjects.

This approach serves multiple purposes.

Process assessment does not prohibit AI use. It gives students a way to explain how they used AI and how their decisions shaped the work. Intermediate artifacts can still be fabricated or incomplete, so they provide context rather than proof of authorship.

Nick Potkalitsky frames process assessment as a countermeasure to cognitive offloading. Its learning value comes from what students do with intermediate work: explain choices, use feedback, revise, and reflect. The artifacts themselves do not guarantee that thinking occurred.

Iterative Drafting

A common form of process-oriented assessment is iterative drafting. Students submit the intermediate artifacts of a larger project for low-stakes credit. The assignment produces a portfolio that includes their process work alongside the final deliverable.

A Typical Sequence

For writing-intensive courses, an iterative drafting sequence might look like this.

• Proposal or outline. Students articulate their topic, research question, or thesis. This establishes direction before significant drafting begins.

• Annotated bibliography or research notes. Students document their sources and their initial engagement with the material.

• First draft. A complete version submitted for feedback, not for a final grade.

• Peer review feedback. Students give and receive structured feedback from classmates.

• Revised draft with reflection. The final submission includes a cover letter or reflection explaining what changed and why.

Each step earns credit. Each step also creates visibility into the student’s process. The final grade weighs both the quality of the end product and evidence of how the work developed.

Some instructors pair this asynchronous process work with in-class activities that reinforce the same skills. Drawing on Matt Beane’s term, Lance Eaton describes a “chimeric” approach that combines human and technological contributions in the same workflow. A student might draft with AI assistance at home, then defend and revise their ideas during an in-class peer workshop. The process tracks both kinds of work.

Adapting for STEM, Clinical, and Applied Fields

The same principles apply outside the humanities. In a STEM project where students design and build something, the sequence begins with a proposal that establishes the problem statement, design constraints, and success criteria. Students then submit a first prototype with initial test results. A feedback round follows, where the instructor, peers, or an AI tool review the work. The student reflects on what they heard. The revised prototype shows how the student responded to that feedback. The final submission pairs the completed product with documentation of what the iterative process taught the student.

In nursing and health sciences, process documentation often already exists in the form of clinical logs, skills check-off sheets, and patient care reflections. A nursing instructor whose clinical competency assessments already require observed demonstrations has a built-in process structure. The AI-era addition is making AI use expectations explicit for the written components that accompany clinical work, such as care plans, case studies, and evidence-based practice papers. In CTE programs where hands-on performance is the assessment, the principle is similar: an observed weld, circuit, or repair is less susceptible to remote AI completion. Process assessment adds value for the planning, documentation, and reflection components that surround the hands-on work.

The principle remains constant. Make the process visible. Create multiple touchpoints for feedback and accountability. Assess the work alongside the finished product.

The PAIRR Framework

The framework was developed by a team of writing instructors at UC Davis and is now expanding across eight California colleges and universities. PAIRR stands for Peer & AI Review + Reflection. A peer-reviewed study by Sperber et al. (2025) involved 654 college students across ten courses and examined how students perceived and responded to peer and AI feedback.

The canonical intervention begins with class preparation about AI and language equity, followed by peer review, AI review, reflection, and revision. The sequence below is one way to place those moves inside a complete assignment workflow.

The framework does a few things at once. It teaches students to treat AI as a feedback tool before they use it as a content generator. It builds critical evaluation skills by requiring students to compare different feedback sources. And it creates accountability, because students must document and justify their revision choices.

This feedback sequence is also a good place to reuse ABC-R. When students compare peer and AI feedback, they can ask whether the AI’s comments are accurate, whether they carry assumptions or bias, whether they fit the assignment context, and whether using the feedback supports the learning purpose. This keeps AI feedback from becoming a second authority figure students obey automatically.

Among 361 post-survey respondents in the Sperber et al. study, 58% preferred combined peer and AI feedback, 36% preferred peer feedback alone, and 6% preferred AI feedback alone. A qualitative analysis of 131 student reflections found that students often perceived peers as more context-aware and valued their emotional support. Students frequently described AI feedback as organized, actionable, rubric-oriented, and attentive to structural patterns. The Reflect step gives students a place to compare those sources and explain which suggestions they will use.

Marc Watkins distinguishes AI feedback from AI grading. He argues that the relational aspect of reading and responding to student work is intrinsically human, and automating it threatens the teacher-student relationship. The framework keeps the instructor in the grading role while giving students an additional source of formative feedback.

The project team has continued to develop the approach. Anna Mills describes updated prompts designed to be more conversational and less overwhelming, refined over a year of testing across multiple institutions. These prompts are publicly available and can be adapted to different disciplines.

Adam Phillips and Alaina Tackitt offer a smaller self-review route in “Structuring AI-Assisted Review to Support Revision” (PDF), part of the peer-reviewed Writing and Rhetoric Studies in the Loop prompt library. Students provide a draft and the assignment instructions, ask whether the draft fulfills the assignment, and use the response to begin a review conversation. The authors reproduce outputs from NotebookLM, ChatGPT, and Gemini, including places where the systems overreach or offer revisions that were not requested. Those examples give faculty something concrete to evaluate before adapting the approach. The student remains responsible for identifying goals, asking follow-up questions, and deciding what to revise.

The reflective cover letter that accompanies the final submission functions as what some call an AI transparency statement. This is a document where students explain how they used AI, what they accepted or rejected, and why.

AI Transparency Statements

An AI transparency statement (also called a disclosure form or acknowledgment statement) is a document where students explain their use of AI tools in completing an assignment. A citation documents a source. A transparency statement documents a process.

Different instructors structure these differently, and the variation reflects different pedagogical priorities. Jason Gulya’s approach emphasizes metacognitive reflection. He asks students to map their work process, mark each step on the AI Assessment Scale, and defend their use or non-use of AI at each stage. Students also reflect on whether they used AI as a “co-pilot” (executing their directions) or “co-thinker” (shaping the ideas themselves). The result is a statement that pushes students to articulate their reasoning alongside their actions.

Christopher Ostro’s mosaic approach centers on documentation and evidence. His transparency statements include an honor code statement explaining expectations, identification of which tools were used (ChatGPT, Grammarly, etc.), a description of how each tool was used, and a link to the document’s version history showing the drafting process. Where Gulya asks students to defend their choices, Ostro asks them to provide a verifiable trail.

Both approaches share a common goal. They ask students to think critically about their process and take responsibility for their choices. Gulya leans toward metacognitive reflection. Ostro leans toward documentation. A faculty member can combine those approaches when both fit the learning goal, asking students to explain their decisions and provide a short process record.

Kari Weaver’s Artificial Intelligence Disclosure Framework provides a more structured option. A short AID statement identifies the tool and describes only the stages or roles in which it contributed. Faculty and students can use the AID Statement Builder to generate a concise statement without maintaining an exhaustive log. The statement documents the process. Source attribution still belongs in citations.

The AI Assessment Scale itself has evolved since its initial release. AIAS 2.1 uses five task types, from No AI through AI Exploration, and distinguishes the task type from whether the assessment is completed under secure conditions. It recommends evidence such as drafts, prompt logs, reflections, and checkpoints when those artifacts fit the task. In unsecured assessments, prohibited AI use cannot be reliably enforced. Process documentation belongs in a broader evidence chain alongside outcomes, student explanations, feedback, and instructor judgment.

Leon Furze adds that frameworks like the AIAS work best as conversation tools, not enforcement tools. They open dialogue between instructors and students about how AI fits into the learning process. That framing aligns with the spirit of transparency statements generally. The goal is reflection and communication, not policing.

Transparency statements work best when they accompany process-oriented assessments. A statement attached to a single final product provides less context because there are fewer process artifacts to compare it with. A statement embedded in an iterative drafting sequence becomes part of a larger accountability structure.

Version History and Documentation

Inspire and Assure introduced version history as one tool for tracking student labor. In the context of process-oriented assessment, version tracking takes on a more specific role. It connects the documentation of how work developed over time to the feedback-and-revision sequence and transparency statements described above.

When students work in cloud-based tools, version history can show saved versions, contributors, timestamps, and changes inside that platform. Google Docs may group revisions, and Word versioning requires the document to be stored in OneDrive or SharePoint. Neither system captures work drafted elsewhere. Version history can complement a student’s reflection or corroborate part of a transparency statement, but it cannot confirm authorship or explain why a change occurred.

Ordinary version history differs from systems that continuously log keystrokes, pauses, pasted text, and typing behavior. Nick Potkalitsky argues that student-owned process documentation can serve learning, while institutional monitoring designed to catch misconduct turns educators from mentors into monitors and reduces students to risk profiles.

Marc Watkins makes a related point, noting that keystroke logging and process monitoring have joined the growing list of surveillance-oriented responses to AI in education. The privacy concerns are real. When process tracking is framed primarily as a detection tool, it can undermine the trust that process-oriented assessment depends on.

Use version history as one element in a larger process-assessment system, paired with reflections, peer feedback, and synchronous touchpoints. Treat it as context for a learning conversation, not as the primary mechanism for catching AI use. If you adopt version tracking, let students use the record to reflect on their revision process. Explain in your syllabus how it connects to the learning goals.

Making Process Work

Several practical considerations affect whether process-oriented assessment succeeds.

One important consideration is communication. Students need to understand how process assessment supports their learning. Jason Gulya identifies common mistakes instructors make when implementing process assessment. Among them is treating process checkpoints as bureaucratic steps that students must complete, which turns the process into exactly the kind of box-checking it was supposed to prevent. Explain the reasoning behind each step along with what students must submit.

This connects to a broader pattern. Students who approach their coursework transactionally may experience process checkpoints as another set of boxes to check. Gulya has written about this problem in AI-era teaching. Explain the purpose clearly and early. Examples of strong process documentation from previous semesters, shared with permission, give students a concrete model.

Anna Mills proposes a layered approach that begins with purpose and motivation before adding structural process elements such as drafting sequences, peer review, and documentation. The sequence is intended to keep those steps connected to the learning goal rather than presenting them as compliance tasks.

Design decisions matter too. Students will not all follow the same path, so building flexibility into your scaffolds helps. You also cannot monitor everything. Use synchronous touchpoints (presentations, conferences, discussions) strategically for high-stakes assessments, and let formative process work stay lower-stakes.

Fully online or asynchronous courses can use alternative accountability structures, although they provide different kinds of evidence. Recorded video reflections let students walk through their work and explain their decisions, but they do not allow the live follow-up of an oral defense. Asynchronous structured peer review with detailed rubrics supports collaboration without requiring simultaneous participation. Short video submissions through an LMS or an approved institutional platform can serve some of the purposes of an in-class presentation. A fully online instructor might pair structured peer and AI review with recorded reflections at selected checkpoints.

A common question is how to weight process work. For illustration, an instructor might assign 30-40% of the grade to process work and the remainder to the final product. Adjust that ratio to the learning outcomes, discipline, workload, and stakes. Tell students what you are looking for in the process artifacts.

Whatever weight you assign to process, share it with students on day one. Students should know which process artifacts count toward the grade and what quality looks like.

One more consideration. Prefer student-owned process documentation whenever possible. Students can collect, curate, and reflect on their own process artifacts. Institutional keystroke tracking or automated process surveillance shifts the activity toward monitoring. Student ownership keeps the focus on self-assessment and intentional AI use.

Looking Ahead

The next chapter, AI Detection and Accountability, addresses AI detection tools directly, examining what the research says about their accuracy and limitations. It also covers communication strategies and accountability structures that reduce the need for detection in the first place.

References

Eaton, L. (2025, May). Ctrl+Alt+Assess: Rebooting learning for the GenAI era. AI + Education = Simplified. Source link

Carless, D., & Boud, D. (2018). The development of student feedback literacy: Enabling uptake of feedback. Assessment & Evaluation in Higher Education, 43(8), 1315-1325. DOI record

Double, K. S., McGrane, J. A., & Hopfenbeck, T. N. (2020). The impact of peer assessment on academic performance: A meta-analysis of control group studies. Educational Psychology Review, 32, 481-509. DOI record

Fernandes, D., Villa, S., Nicholls, S., Haavisto, O., Buschek, D., Schmidt, A., Kosch, T., Shen, C., & Welsch, R. (2026). AI makes you smarter but none the wiser: The disconnect between performance and metacognition. Computers in Human Behavior, 175, 108779. DOI record

Furze, L. (2025, August 9). AI in the writing process: A problem of purpose. Source link

Furze, L. (2025, August 18). Five principles for rethinking assessment with Gen AI. Source link

Furze, L. (2025, September 23). How (not) to use the AIAS. Source link

Gulya, J. (2025, September 10). How to teach with AI transparency statements. Faculty Focus. Source link

Gulya, J. (2025, August 10). If we’re going to adapt to the age of AI, we’ll need to chip away at transactional education. The AI Edventure. Source link

Gulya, J. (2025, October 17). Problems with “process over product” (Part 1). The AI Edventure. Source link

Mills, A. (2025, August 23). Strategies for reducing the likelihood of AI misuse. Anna Mills’ Substack. Source link

Mills, A. (2025, October 19). The time to reckon with AI agents in digital learning spaces is now. Anna Mills’ Substack. Source link

Google. (n.d.). Find out what’s changed in a file. Google Docs Editors Help. Source link

Microsoft. (n.d.). Use versioning with Word. Microsoft Support. Source link

Mills, A. (2026, January 27). AI feedback customized for student writers: The updated PAIRR prompts. Anna Mills’ Substack. Source link

Nicol, D. J., & Macfarlane-Dick, D. (2006). Formative assessment and self-regulated learning: A model and seven principles of good feedback practice. Studies in Higher Education, 31(2), 199-218. DOI record

Ostro, C. (2025). A mosaic approach to academic integrity in the AI era. Pangram. Source link

Perkins, M., Roe, J., & Furze, L. (2025). Reimagining the Artificial Intelligence Assessment Scale: A refined framework for educational assessment. Journal of University Teaching and Learning Practice, 22(7). DOI record

Phillips, A., & Tackitt, A. (2026). Structuring AI-assisted review to support revision. In A. Gupta & B. Gogan (Eds.), Writing and rhetoric studies in the loop: A GenAI prompt library (pp. 383-394). The WAC Clearinghouse. PDF source

Potkalitsky, N. (2025, April 10). Looking for the next world: Possible risks of cognitive offloading in an AI education landscape. Source link

Potkalitsky, N. (2025, April 14). Process tracking is not the answer. Source link

Shaw, S. D., & Nave, G. (2026). Thinking—Fast, Slow, and Artificial: How AI is reshaping human reasoning and the rise of cognitive surrender [Working paper]. The Wharton School. DOI record

Sperber, L., MacArthur, M., Minnillo, S., Stillman, N., & Whithaus, C. (2025). Peer and AI Review + Reflection (PAIRR): A human-centered approach to formative assessment. Computers and Composition, 76, 102921. DOI record

Tertiary Education Quality and Standards Agency. (2023). Assessment reform for the age of artificial intelligence. PDF source

University of California, Davis, University Writing Program. (n.d.). Peer & AI Review + Reflection (PAIRR). Source link

Watkins, M. (2025, March 16). When algorithms watch you write. Rhetorica. Source link

Watkins, M. (2025, October 10). The dangers of using AI to grade. Rhetorica. Source link

Weaver, K. D. (2024). The Artificial Intelligence Disclosure (AID) Framework: An introduction. College & Research Libraries News, 85(10), 407-411. DOI record

Weaver, K. D. (n.d.). Artificial Intelligence Disclosure (AID) Statement Builder. Source link

Further Reading

• Digital Education Council. (2025). The next era of assessment: A global review of AI in assessment design. In partnership with Pearson. Source link

• Elbow, P. (1973). Writing Without Teachers. Oxford University Press. A foundational text on process-oriented writing pedagogy.

• Gunder, A. & Herron, J. (2026, January). AI literacies in practice: A comprehensive playbook for higher education. D2L/WCET. PDF source

• Martineau, K. (2025, May 13). AI is changing how we work: Is it time to change how we credit AI’s involvement? IBM Research. Source link

• Levine, A. (2026, June 13). Transparency of generative AI use in OEAward nominations. OEGlobal. Source link

• Mills, A. (2025, July 19). Getting the most out of AI feedback. Source link. A student-facing chapter from Mills’ OER textbook-in-progress, AI and College Writing.

• Watkins, M. (2025, August 22). Higher education needs frameworks for how faculty use AI. Rhetorica. Source link

• Watkins, M. & Monroe, S. (Eds.). (2025). Building AI literacy [Special issue]. Thresholds in Education. Source link

• Wiggins, G. (1998). Educative Assessment: Designing Assessments to Inform and Improve Student Performance. Jossey-Bass. Connects process assessment to authentic assessment design.