Assessment Design Through Co-Creation: Student-Staff Partnership in Evaluating the Impact of Artificial Intelligence

Synthesis: Williams and Ingle (2025) report a case study of the AI Co-Creators project at University College London (UCL), a co-creation initiative in which a student partner who had completed her second year of an undergraduate medical sciences program and a staff member collaboratively evaluated ChatGPT-generated output in response to diverse coursework assessments. The pair co-designed seven evaluation criteria grounded in UNESCO and JISC guidance, then examined the challenges surfaced through iterative dialogue, a structured interview, and thematic analysis. The partnership found that no coursework assessment was completely immune to ChatGPT interference and explored the value of the collaboration from student, staff, and institutional perspectives, connecting co-creation and students-as-partners practice to Self-Regulated Learning, enhanced understanding, and student empowerment in assessment design.

AI Co-Creators: A Students-as-Partners Assessment Project

The AI Co-Creators project was part of UCL's broader ChangeMakers initiative, established in 2014 to support dialogue between students and staff on education challenges, including curriculum change and learning enhancement. This case study focused on a single partnership between a staff member and a student partner who had personal experience of the assessments under evaluation — a deliberate choice given the dialogic nature of the work and the aim of drawing on the insight of an experienced student. From the outset the project sought a non-hierarchical, inclusive working environment that positioned staff and student as equal partners, echoing the students-as-partners emphasis on students as co-researchers and co-creators rather than passive subjects of education (Healey, 2014).

The project employed an iterative working strategy of in-person and online meetings with clearly established partner responsibilities. The student partner collated assessment information and evaluated ChatGPT-generated output, while the staff partner provided academic support, mentorship, and feedback; both partners co-created the assessment evaluation criteria and parameters. Throughout, the student's input was given space and priority, and the student primarily drove the project forward — an intentional structure aimed at empowering the student as an active participant.

Co-Designing Evaluation Criteria

Due to the novelty of Generative AI and the recent release of ChatGPT, the partners co-created their own evaluation criteria rather than relying on established frameworks. Drawing on UNESCO guidelines (Holmes & Miao, 2023; Sabzalieva & Valentini, 2023), pedagogical applications, and JISC guidance for Generative AI in assessments (Gamage et al., 2023), they designed seven criteria around three themes: the application of ChatGPT to higher education teaching and learning, challenges and ethical implications, and adapting to ChatGPT in higher education. ChatGPT 3.5 was used to generate responses to coursework instructions, and the outputs were co-evaluated through iterative dialogue against the criteria.

The criteria addressed the potential impact of ChatGPT on academic integrity, key challenges facing coursework assessments, applicability to formative and summative assessment, challenges for students and staff, impact on the student experience, how ChatGPT might be incorporated into existing assessment designs, and an analysis of example ChatGPT-generated output for each coursework instruction. Both partners agreed the main impact would likely fall on academic integrity and the potential use of AI for plagiarism, with the student partner raising particular concerns about equity and fairness in grading.

Key Findings

1. No assessment was immune. The main finding, agreed by both partners, was that no form of coursework assessment was completely immune to ChatGPT interference — with important implications for how assessments are regulated and designed.
2. Co-creation surfaces student-centric applications. The student partner highlighted ways ChatGPT could serve as a more inclusive form of formative assessment, a learning assistant, an auxiliary tool to provide Feedback, and a proofing aid — uses more student-centric than staff-centric.
3. Training is needed. Both partners recommended training for staff and students on the capabilities, limitations, and ethical use of Generative AI.
4. Co-design supports assessment literacy. Co-creation can help align differing assessment expectations and deepen both student and staff assessment literacy (Smith et al., 2013; Boud et al., 2018).

Student Perspective

A structured interview with the student partner, subject to thematic analysis following Braun and Clarke (2006), surfaced three main themes: transforming university assessment through generative AI, empowering education through co-creation, and navigating institutional and ethical landscapes. The student echoed broader concerns about academic integrity and equitable access, while also identifying opportunities to integrate Generative AI into assessments to enhance learning through continued and constructive feedback.

The student partner described the process as highly rewarding, felt she had contributed valuable feedback for designing assessments and teaching plans, and noted the collaboration allowed her to discuss ideas she may not have considered independently. Taking responsibility for elements of the project was a powerful device for empowerment, and the project made her more reflective in her own learning. Positioning the student as an equal partner helped build Trust and fostered a proactive, self-regulated approach to the work — illustrating the link between co-creation and Self-Regulated Learning.

Staff and Institutional Perspectives

From a staff perspective, the partnership gained important insights that would not have emerged independently, particularly the student's personal experience of the assessments and her effectiveness at identifying areas where ChatGPT could impact coursework. Co-creation and student-staff collaborations are argued to be ideally positioned to scrutinize Generative AI use by capitalizing on differing perspectives and expertise (Bovill, 2020). At the institutional level, projects such as AI Co-Creators contribute to policy on the use of generative AI in assessments and the classroom, benefiting existing and future student cohorts and supporting the shift from traditional hierarchical structures toward more democratic, student-voice-oriented higher education (Cook-Sather, 2006; Mercer-Mapstone & Bovill, 2020).

What this means for practice

• Assessment designers. Treat no coursework format as immune: audit existing assessments for ChatGPT interference rather than assuming a given format is a safeguard, since both partners concluded no form of coursework assessment was completely immune.
• Assessment designers. Co-create evaluation criteria with a student partner when existing frameworks do not fit, as these partners did in building seven criteria around UNESCO and JISC guidance.
• Instructors. Give the student partner priority and genuine ownership of the work, because taking responsibility for elements of the project was what produced the empowerment and increased self-reflection she reported.
• Instructors. Provide training for both staff and students on the capabilities, limitations, and ethical use of Generative AI, which both partners recommended.
• Educators. Use co-creation to align differing assessment expectations and build assessment literacy in students and staff alike, and extend the approach to further assessment formats (such as MCQs and short answer questions) and to a more diverse group of students to improve equity and inclusion.

Limitations

• The case study involved a single student partner — who had just completed her second year of one undergraduate medical sciences program — working with one staff member, so the entire evidence base is one dialogic partnership.
• The scope of assessment formats was limited to existing coursework assessments in years 1 and 2 of that program; the authors recommend extending evaluation to year 3 and postgraduate-taught MSc assessments, and to formats such as MCQs and short answer questions.
• Only the student partner was interviewed, and the authors note that a reciprocal interview of the staff partner would have yielded further insights and is recommended for future studies.
• The authors state the findings were not definitive, and the evaluation used ChatGPT 3.5 output alone, so the conclusions describe one tool version's performance against one set of assessment instructions.

Citation

Williams, A., & Ingle, E. (2025). Assessment design through co-creation: Student-staff partnership in evaluating the impact of artificial intelligence. International Journal for Students as Partners, 9(1), 214–226.