The 2026 AI and Academic Integrity conference highlighted significant themes including AI literacy, innovative assessment approaches, and the evolving definition of authorship in the age of AI. Speakers emphasised the importance of embedding AI literacy in education, rethinking assessment models, and fostering transparency about AI’s role in academic work.
AI literacy
Prof Mary Richardson, UCL, championed the AI-literate graduate approach that focuses attention on assessing capabilities. She argues that AI literacy should be embedded in programme-level learning outcomes, and that assessment should include conceptual understandings, ethical awareness and applied AI capabilities.
Interestingly, she notes that what counts as academic misconduct changes over time. This is a helpful prompt to remind us to regularly revisit our position on this! She also argues that universities are not ethically neutral (Tesar et a. 2021), and likewise neither are our procedures and regulations. I was particularly drawn to this position given the challenges faced around inclusion and AI, and problematics such as AI detection equity concerns for autistic students and English as an additional language students.
As we have done at Bristol through our student focus groups and surveys, Prof Richardson reminds us to check our assumptions:
- What data do you have about student use of AI technologies?
- How often to do you talk to students about how they use such technologies?
- How you talked to students about why they might use such technologies?
Centrally-collected and national-level data provides insights in response to these questions, but there is also good reason to ask your own students in your own local context. Checking in regularly means you can stay on top of shifting trends, behaviours and attitudes. Sometimes this works well as casual conversations in weekly seminars, or you could also do a single regular anonymous prompt at the end of lectures.
Programmatic approaches
As you can imagine, assessment is the number one concern tackled by each speaker. Dr Danny Liu, Professor in the Educational Innovation team in the DVC (Education) Portfolio, updated us on the University of Sydney’s adoption of the “two-lane” approach to assessment that is growing in popularity in Australia. He argues that we need to move away from voluntary compliance frameworks that rely on discursive methods of assessment validity. Instead, he states that we should shift to a structural approach. There is both assessment of learning and assessment for learning, but integrity is handled differently. In practice, this means that there are observed in-person assessments, Lane 1, and not observed assessments, Lane 2. Both may allow AI, but do so differently. Lane 1 may or may not allow AI, Lane 2 allow AI as relevant, where the use of AI is scaffolded and supported.
Lane 2 may not directly verify the capability of students. However, within the two-lane approach at a programmatic level, Lane 1 and Lane 2 routes offer overall verification. Lane 1 becomes the hurdle that students must pass that ensures student capabilities. Students are presented with an assessment “menu”; ways that AI might be used in a given assessment, with some highlighted as more or less useful for a particular assessment.
The big take-way from Dr Liu is that, at first, he was uncomfortable with having Lane 2 assessments, as were many of his colleagues. They had to learn to trust the programmatic approach. He had to accept that while his own unit may consist only of Lane 2 assessments, colleagues on other units would oversee Lane 1 assessments, ensuring the verification of student learning across the whole programme.
Lane 1, secure, assessment types include a final written or practical exam or oral assessment, an in-semester test, in-person interactive orans or practicals, or a placement, internship or supervision. Lane 2, open, assessments include quizzes, inquiries and investigations (experimental design, data analysis, case studies, research analysis), production and creation (portfolio, performance, creative work, dissertation) and discussion (debate, conversation, evaluation). These all map onto Laurillard’s learning types.
For Dr Lui, the teacher’s role focuses on: educating students how to learn alongside AI; and designing meaningful Lane 1 assessments.
He usefully noted some “traps” and unintended consequences: some staff still fixate on their isolated unit alone rather than the whole programme; some staff won’t reimagine assessment and revert to pen and paper exams; some staff focus only on “secure writing” instead of learning outcomes; and some see Lane 1 as “no AI” instead of verifying capabilities.
Authorship and AI
Dr Chelle Oldham, Open University, posed some thought-provoking questions around what authorship means since the world opened up to generative AI at scale? She posits that AI is not the author, AI use should always be disclosed, and the human remains accountable for the whole work.
She notes, however, that accountability is philosophically contested and our concept of authorship was built for a pre-AI world. This should prompt us to reflect on who we value and credit when thinking about labour, originality and accountability.
Dr Oldham suggests that we move away from policing and into principles. For example, we should focus on disclosure over detection to declare the role that AI played. We can imagine authorship as a continuum (Perkins et al. 2024) from no AI to AI-assisted, to AI-guided, to AI-created and full AI.
Process and output
Prof Naomi Winstone, University of Surrey focuses on authenticity. She does this by looking at assessment as “wicked problem” as that gives us permission to compromise, to diverge and to iterate. She asks us to take a structural approach – to assess not only the final product, but also the process students use to create it. The metaphor of a flight recorder black box is used – we need educational flight data to trace and authenticate learning. This process-focused approach values the drafting, researching, logging, mistakes and cognitive changes made through the development of a final output, whether or not AI was used.
Again, principles are important. Assessing the process means we are capturing learning, valuing errors as evidence of progress, encouraging transparency and agency, and rewarding the whole learning journey, not just the end point. You can find out more about what this looks like in practice here.
Assessment questions
Prof Lydia Arnold, Harper Adams University, leaves us with some provocative questions for the future:
- Is anonymity serving us well?
- Do we have a duty to reassess academic communication and norms?
- Are we focused too much on assessment in this debate and not enough on learning?
- Are assessment criteria stifling imagination and creativity?
For Prof Arnold, like others, finds that authentic assessment is key. It is relevant for employers and the future, and to individual aspiration (and reflects UoB’s feedback and assessment priorities!). It results in diverse outputs (nice for when we are marking!), and equips students with skills to respond to uncertainty. It also helps students to build metacognition, and look at process as well as product.
She drew attention to Wilde Scott’s book “Raising Dreamers” that states: “the future belongs to the people who are both autoelic and autodidactic…. The idea of spending the first eighteen to twenty-five years of your life simply preparing for a job by following a straight line makes no sense moving forward”. Plenty there to reflect on how we shape our students as both humans and future employees.
With this in mind, she tasks us with look at authentic assessment as an approach to cultivate: identify and voice; assessment with impact; curiosity as the underpinning; many ways to thrive; right to object; and audience-focused. Like other speakers, she foregrounds the need for programmatic approaches and transparency of AI use.
Prof Arnold also challenges us to value education as a relational act, one that embeds co-learning, connecting and being human. And one that applies ethics, criticality and care.
Next steps
Check out our education focused resource via Develop, perfect for teaching and professional service staff. Dip in and out of relevant sections and explore the many case studies included, and tips on prompt engineering.
You can find out a lot more about how AI works in the new Futurelearn course AI Fundamentals.
Leave a Reply