Student use of AI – tools that can adapt based on rules that mimic human intelligence, and can perform learning and problem-solving through machine learning and neural networks – has soared in the last year. This year’s Student Generative AI Survey from the Higher Education Policy Institute found that almost all students (92 per cent) now use AI in some form, up from two-thirds in 2024. 80 per cent of students stated that their UK HE institution had a clear AI policy, and 42 per cent felt staff are somewhat well-equipped to deal with AI (up from just 18 per cent in 2024). However, while students overwhelmingly believe it is essential to have good AI skills, only 36 per cent have received support from their institution to develop them. It’s time to talk more about what ‘good AI skills’ involve, and the role academics – and psychologists – have in helping to build them.
As with any educational tool, AI use comes with multiple perspectives. Proponents of AI in education argue that these technologies can enhance learning by automating routine tasks, thereby enabling students to focus on higher-order cognitive activities. Examples include creating podcasts from readings, summarising documents, and generating code – all of which may help students engage more deeply with material and foster critical analysis. Moreover, such applications provide students with practical skills that are increasingly relevant for the workplace as AI technologies evolve.
How do we use AI in our own workplace, and what issues do we see emerging?
Elizabeth Kaplunov: I use AI very rarely during my academic work. I am ‘old school’ and feel that it is easier to write articles myself, in part because AI writes in a very odd way and because AI can save my ideas and share them with others without my consent.
I sometimes use AI for helping me to plan articles, for example with word count and sections based on the submission guidelines for a journal, or to help me to edit reference lists – I will upload information about Harvard referencing format guide into ChatGPT and then ask it to point out mistakes in the referencing list according to the guide. Recently, I’ve discovered an AI tool which can paraphrase sections of text to suggest alternatives. This is used by my colleagues for writing emails, where they write the basics of what the email should say and then ask AI to write it using a formal or more friendly tone, or to use higher-level vocabulary.
Athina Ntasioti: I often use AI in my academic work when I need clarification; however, I have found that AI tools are not always trustworthy. For example, references are frequently inaccurate, and the information provided often requires double-checking for validation.
As a researcher focusing on AI bias and ethical concerns, I encourage students to use AI during their academic journey, but with mindfulness and caution. It is important that they do not lose their capacity for critical thinking. While AI may support language development, academic tone, and writing – particularly for students from diverse linguistic backgrounds – the foundation of academic engagement should remain with scholarly sources such as Google Scholar and other academic databases. Critical thinking, therefore, is an essential part of their writing, and is often what I see missing when AI is used.
University policy
Many higher education institutions have policies that students are not allowed to use any AI tool to write work for them. However, students are allowed to use AI to check spelling, to help them in their research, or to plan work.
In our experience, students struggle to understand the difference between AI-generated and AI-refined content. We have seen students paste entire ChatGPT-generated paragraphs into assignments, assuming its responses are academically sound. Turnitin has an AI check which tells staff exactly where AI was used within the student assignment. When asked about sources or reasoning during Academic Misconduct Panels, they often cannot explain or even remember the content.
Students tend to use AI to write their work when they don’t understand the task, or when they have not done the work on time. While the time-saving aspect can be useful in planning, few students stop to ask where the information came from or whether it is appropriate for their assignment. The content produced by AI is largely very generic, shallow and factual, and so to the lecturer’s trained eye, it is highly obvious when students have used AI extensively. This often results in an Academic Misconduct Panel, or student work not meeting Learning Outcomes due to the assignments not following the brief.
Ultimately, students need to consider how to use AI in a way that produces critical and thoughtful outputs, as well as being able to paraphrase in their own voice.
Deeper into critical thinking
Critical thinking is, then, emerging as a concern both for us as lecturers and in the wider Higher Education policy over AI. As Brookfield (2013) defines it, critical thinking involves questioning assumptions, interpreting facts, evaluating the strengths and weaknesses of arguments, and forming balanced judgements. Encouraging students to interrogate the sources of AI-generated information and to recognise potential biases is therefore essential. To strengthen critical engagement, educators must not only review the ways in which AI use can hinder critical thinking, but also provide strategies for using AI in more reflective and purposeful ways.
Studies have indeed shown that reliance on AI for information retrieval and decision-making can diminish reflective problem-solving and independent analysis. For instance, a 2025 study by Michael Gerlich highlighted a significant negative correlation between AI tool usage and critical-thinking scores, noting that younger participants (ages 17–25) were more dependent on AI tools and scored lower in critical thinking compared with older groups. This suggests that overreliance on AI can weaken essential intellectual skills.
Psychology offers some explanation and conceptual framework for why this might happen. Daniel Kahneman’s work, summarised in the popular 2011 book Thinking, Fast and Slow, suggests that there are two types of systems working on decision making. System 1 is fast, instinctive and emotional. System 2 thinking is slow, deliberate and more logical. This is where critical thinking is formed.
AI use is aligned with System 1 thinking, as it speeds up and simplifies tasks, mimicking intuition by autocompleting text and making recommendations. AI also reduces mental effort by making quick and low-effort decisions. Students may view AI as a System 2 tool which has all the possible information, but we would argue they are being nudged toward shortcuts and away from deeper reflection.
Critical thinking over critical thinking…
Unfortunately, some of the research around AI and critical thinking, from within Psychology and allied disciplines, arguably suffers from a lack of critical thinking itself.
For example, Kosmyna et al. (2025) conducted a study at the Massachusetts Institute of Technology using EEG technology to monitor students’ brain activity during essay-writing tasks. Participants who used ChatGPT showed significantly lower neural engagement in regions associated with memory, attention, and executive control, compared to those who used only a search engine or no assistance at all. These students also produced less original content and demonstrated reduced recall of their written material. The study described this phenomenon as ‘metacognitive laziness’, highlighting the potential of generative AI to diminish sustained cognitive effort and deeper analysis.
However, the study has faced substantial criticism. As a preprint, it has not undergone full peer review, and commentators have questioned whether EEG patterns alone can reliably measure cognitive engagement. Reduced neural activation in some regions may reflect alternative cognitive strategies rather than ‘laziness’, and that small sample sizes further limit the generalisability of findings.
Similarly, Zhong et al. (2024) examined AI-assisted versus human-led problem-solving among programming students, finding that AI-led participants tended to accept solutions uncritically. Yet, this study is highly context-specific, with a small sample, making it difficult to extrapolate results across disciplines. These critiques emphasise the need for cautious interpretation rather than alarmist conclusions.
Interestingly, in their comparative study, Zhong and colleagues found that students working with AI tools often accepted outputs uncritically unless explicitly prompted to question them. This emphasises the importance of structured support and critical prompts to help learners move beyond passive acceptance and into evaluative thinking.
Encouraging thoughtful use
Despite the cautionary findings, there are ways to encourage thoughtful and productive AI use in academic contexts. Embedding AI within pedagogical frameworks that emphasise metacognition, reflection, and ethical engagement can support critical thinking rather than undermine it. Students can be guided to treat AI as a collaborative partner, one that requires their scrutiny, reflection, and contextualisation.
Practical strategies include explicitly teaching students how to interrogate AI outputs, compare them against reliable sources, and integrate them into broader research workflows. For example, students can use AI to generate preliminary ideas or outlines, then critically evaluate and refine the outputs through independent research. AI can also serve as a prompt for self-questioning, encouraging learners to justify choices, recognise assumptions, and explore alternative perspectives. We can design assignments that require students to annotate or reflect on AI-generated outputs, embedding prompts such as ‘Where does this information come from?’ or ‘What would you change and why?’ into coursework.
Practical teaching strategies can also include asking students to compare information from multiple sources, such as examining medical professional practices or policies reported by different media outlets, to highlight differences and encourage critical thinking. Students can be guided to map out their problem-solving process before consulting AI, then compare their reasoning with AI-generated outputs. This generates valuable classroom discussion about the process of thinking, not just the correctness of the answer.
Encouraging students to adopt this reflective approach requires support at multiple levels. Clear institutional policies, scaffolded learning activities, and feedback that emphasises critical engagement are all essential.
Additionally, when conducting individual student support sessions, it helps to teach students why reflection is important, how to do it, and how to break down assignment requirements and expectations in UK higher education. Encouraging peer marking of small sections of the assignment in class helps students to understand the task requirements and engage in critical analysis. Another approach is asking students to work in groups to review the assignment brief, highlight key action words, and summarise the requirements together, which promotes reflection through collaborative discussion.
Institutions, meanwhile, must go beyond detection tools and discipline. They should develop clearer policies distinguishing between AI as a support tool and as a substitute, offer training for staff on ethical and pedagogical uses of AI, and embed AI literacy into induction and academic skills programmes. Students should be required to submit short AI use statements with their assignments. We can also normalise experimentation with AI in low-stakes settings, and develop rubrics that reward critical engagement.
From the student perspective, the message is simple: be curious, not passive. If students use AI to plan or write, they should ask themselves whether they truly understand the content and whether they can explain it without AI. If not, the learning has not happened.
Ultimately, the key is balance. AI must complement – not replace – the intellectual engagement that drives scientific discovery and deeper learning. By thoughtfully integrating AI into the curriculum, higher education can cultivate spaces where technology enhances curiosity, supports ethical awareness, and ensures that critical thinking continues to thrive. We need to teach students not just how to use AI – but how to question it. Used well, AI can support thinking. Used without intention, it can quietly erase the cognitive struggle that deep learning depends on.
- Dr Elizabeth Kaplunov, FHEA, CPsychol, Senior Lecturer at School of Health and Sports Science, Regent College London
- Athina Ntasioti, FHEA, Lecturer at School of Health and Sports Science, Regent College London
