University of Leicester calls for AI literacy in core curricula as study links reflection to critical thinking

Evidence published by the UK Parliament Education Committee says optional provision is reaching too few students and staff, while separate research across UK- and China-based universities points to learner control and reflection as key factors in AI-supported study.

Professor Xue Zhou, Professor in AI in Business Education and Dean of AI at the University of Leicester

The University of Leicester has called for AI literacy to be embedded across higher education curricula after its evidence to the UK Parliament Education Committee found that voluntary workshops, events and centrally provided guidance were not reaching enough students and staff to build deeper capabilities.

The written evidence, submitted by Professor Xue Zhou, Professor in AI in Business Education and Dean of AI at Leicester Business School, and Dr Sara Moze, Digital and Learning Innovation Manager at the University of Leicester, sets out a proposed 4P framework covering Policy, People, Pedagogy and Platform.

The recommendations follow two years of AI education activity at the university, including policies, staff workshops, student resources, lecture programs, interdisciplinary events and a 240-member AI and higher education Community of Practice.

However, the university's participation data shows a gap between interest and sustained engagement. Attendance across sessions at its February 2026 AI and Robotics Symposium ranged from 10% to 38%, while attendance at nine centrally organized staff AI workshops during 2025/26 stood at 54%.

The University of Leicester is now piloting its 4P approach within the College of Business, with the intention of informing wider institutional adoption. An asynchronous AI course for students is also being piloted ahead of a university-wide introduction in the 2026/27 academic year.

Voluntary AI programs struggle to convert interest into participation

The parliamentary submission says more than 300 people registered for the AI and Robotics Symposium, which was co-hosted with the University of Birmingham, with nearly 200 students and staff participating across its online and in-person formats.

The event included robotics demonstrations, virtual reality simulations, an applied AI workshop, panel discussions and sessions focused on ethics. Participants who attended gave the event an average satisfaction score of 4.89 out of five, with all respondents saying they would recommend it.

Other initiatives have also attracted participation. Nine Teach with Tech sessions recorded 182 attendees between February 2025 and March 2026, while seven online lectures on AI and higher education attracted 474 attendees following the series' introduction in April 2025.

The evidence nevertheless argues that optional provision has limited reach, particularly when sessions require attendance in person or greater practical involvement.

Possible barriers identified by the university include students prioritizing paid work, demand for shorter and more flexible learning formats, concerns about AI's environmental and ethical impact and a perception that AI skills will only become relevant later in a student's career.

For staff, the submission points to workload, job insecurity and low morale as factors reducing participation in non-mandatory development. Short online lectures have generally attracted stronger engagement than longer practical workshops.

University surveys also indicate growing skepticism about AI's contribution to learning. The annual survey received 244 responses in 2024/25 and 144 in 2025/26.

Students' average rating of the accuracy of AI-generated academic material fell from 2.898 to 2.719 out of five. Their assessment of AI's effect on academic skills development declined from 3.398 to 2.790, while the perceived importance of AI skills for future careers dropped from 3.574 to 2.962.

The submission says some students equate AI competence with the ability to use a chatbot, creating a risk of overconfidence without deeper skills in verification, evaluation, judgment and ethical reasoning.

Zhou said in a LinkedIn post: "AI literacy should be understood as critical, ethical, and context-sensitive capability rather than simply familiarity with AI tools."

Leicester moves from AI events to a 4P framework

The University of Leicester's proposed model combines institutional governance, human capability, curriculum design and controlled technology adoption.

Under Policy, the university has operated a Policy on Generative Artificial Intelligence in Learning, Teaching and Assessment since September 2024. It is reviewed annually and is supported by a traffic light system that gives students and staff guidance on permitted AI use in assessment.

The university is also finalizing guidance covering AI in research, including its use in research design, data analysis and funding applications.

The People element prioritizes confidence, judgment and reflective practice rather than technical proficiency alone. Training is intended to help students and staff decide when AI should be used, how its output should be assessed and when using it may be inappropriate.

Under Pedagogy, the College of Business pilot treats AI as an embedded part of the curriculum rather than an optional activity. Its approach covers understanding AI concepts, applying AI within specific disciplines, critically evaluating and creating with AI, and AI ethics.

Students are expected to build proficiency in at least two AI tools connected to their subject and likely career pathway.

The Platform element takes a selective approach rather than expanding the number of tools available. The evidence identifies Microsoft Copilot, Blackboard's AI Design Assistant, Microsoft Excel and Sage among the workplace-relevant and education-specific technologies being considered or used.

The submission argues that platform decisions should follow learning objectives and account for data security, copyright, employability and the risk of overwhelming students and staff with unnecessary tools.

UK-China study links learner control and reflection

A separate paper involving Zhou has added research evidence to the university's focus on reflective AI use.

Published online in Studies in Higher Education on June 11, 2026, The agency gap: perceived human AI agency, reflection and generative AI learning across UK and China based higher education contexts examines responses from 309 university students.

The authors are Aniekan Essien and Marios Kremantzis of the University of Bristol Business School, Zhou of the University of Leicester and Da Teng of Beijing University of Chemical Technology.

The study analyzed responses from 145 students studying in the UK and 164 studying in China. Data was collected between September and December 2025 from undergraduate and postgraduate students aged 18 or older who reported using AI in coursework during the term.

In both samples, students who reported retaining greater control over their AI-supported learning also tended to report stronger reflective engagement. Reflection was, in turn, positively associated with self-reported critical thinking in the UK and China samples.

The indirect relationship between perceived human-AI agency and self-reported critical thinking through reflection was statistically significant in both groups.

Zhou said: "This suggests that simply interacting with AI is not enough, students need opportunities to reflect on that interaction, evaluate outputs, and consider how AI influences their own thinking and learning processes."

The study defines perceived human-AI agency through students' reported initiative, monitoring of AI output and control over final learning decisions. It does not measure whether students objectively improved their critical thinking or track their actual prompting, verification or revision behavior.

Differences also emerged between the samples. Reflection was associated with stronger self-reported academic self-concept among UK-based respondents, but that relationship was not statistically significant in the China-based group.

The researchers caution that this does not establish a confirmed difference between the two higher education contexts. Academic self-concept was not measured identically across the samples, and the cross-group comparison did not meet the required threshold for statistical significance.

AI literacy slightly strengthened the relationship between agency and reflection in the China sample. No equivalent moderation effect was identified in the UK or combined models.

The research is based on a one-time, self-reported survey and cannot demonstrate that retaining control causes students to reflect more, or that reflection causes stronger critical thinking. The samples are also not nationally representative, and the research did not directly measure culture, institutional AI policies, discipline, tool availability or educational background.

The University of Leicester's next steps include the 2026/27 introduction of its asynchronous student AI course and continued testing of the 4P model in the College of Business. Its parliamentary evidence recommends that universities move AI capability development into core programs and assessment structures rather than relying primarily on students and staff choosing to attend additional training.