Shaping integrity: why generative artificial intelligence does not have to undermine education (Myles Tan and Nicholle Maravilla); There is no scientific evidence to support the idea that people can be classified as visual, auditory or kinesthetic learners. (Luc Rousseau); What the Research Says: Early Lessons from AI Tutoring That Matter Today (Rose Luckin)
Shaping integrity: why generative artificial intelligence does not have to undermine education
The integration of generative artificial intelligence (GAI) in education has sparked intense debate, with concerns about academic integrity at the forefront. However, as Tan and Maravilla (2024) argue, “GAI, when integrated responsibly in education, does not erode academic integrity. Instead, it fosters intrinsic motivation, enhances digital literacy, and supports constructivist learning principles.” Rather than viewing AI as a threat, this perspective encourages a shift toward ethical and pedagogically sound implementation.
One of the key takeaways from their work is the importance of grounding AI use in established ethical frameworks and educational theories. “To navigate this landscape responsibly, it is essential to revisit established ethical frameworks and educational theories. The ethical principles guiding our use of technology in education have remained consistent, even as the tools themselves have evolved. By referencing seminal works and foundational theories, we can demonstrate that the core values of honesty, fairness, and responsibility are timeless.” This assertion highlights the need to align AI integration with long-standing educational values rather than treating it as an entirely new disruption.
Constructivist learning theory provides a particularly useful lens for understanding how AI can be leveraged effectively in education. As the authors note, “Constructivist learning theory posits that learners construct knowledge through experiences and reflections, actively engaging with content to build understanding. GAI, with its advanced capabilities, aligns well with this theory, offering tools that promote exploration, interaction, and personalized learning paths.” This perspective reframes AI not as a shortcut to bypass learning but as a tool that can enrich inquiry-driven education by fostering deeper engagement with content.
Additionally, self-determination theory (SDT) offers another compelling justification for AI’s role in education. According to Tan and Maravilla, “By promoting autonomy, competence, and relatedness, AI tools help students develop a genuine interest in their subjects, reducing the likelihood of dishonest behavior.” AI, when thoughtfully implemented, can empower students by offering personalized feedback and adaptive learning experiences that cater to their unique needs, thereby fostering intrinsic motivation. Rather than incentivizing cheating, well-designed AI systems can encourage students to take ownership of their learning journey.
The practical implications of this argument are significant. Instead of banning AI tools outright, educators should focus on integrating them into curricula in ways that reinforce ethical academic practices. This includes explicit instruction on digital literacy, ethical AI usage, and fostering critical thinking skills that help students engage meaningfully with AI-generated content. Furthermore, institutions must provide educators with training and resources to help them guide students in using AI tools constructively.
The conversation about AI and academic integrity is far from over, but Tan and Maravilla’s work provides a strong foundation for moving the debate forward. Their research affirms that “ethical education and personalized feedback further empower students to navigate the digital world responsibly, ensuring that they use AI tools to enhance their learning rather than as shortcuts.” The challenge ahead is not to resist AI’s presence in education but to ensure its responsible use in ways that enhance, rather than undermine, learning integrity.
There is no scientific evidence to support the idea that people can be classified as visual, auditory or kinesthetic learners.
Over the past week, I have been listening to my Year 9 history students introduce themselves as learners through Microsoft Teams Flip assignments. A striking number of them describe themselves in terms of so-called ‘learning styles’—particularly as ‘visual learners.’ This persistent belief, though widely accepted, is a classic example of a neuromyth that continues to shape education despite overwhelming scientific evidence to the contrary. As Rousseau (2020) argues, “There is no scientific evidence to support the idea that people can be classified as visual, auditory or kinesthetic learners.” This idea, while comforting and intuitive, is fundamentally flawed and has no place in modern pedagogy.
The appeal of learning styles stems from the notion that individuals have unique cognitive predispositions that determine how they best absorb information. However, cognitive research consistently refutes this claim. Rousseau highlights that “human brains have infinitely more commonalities than differences” and that our sensory areas are highly interconnected. When we process information, we do so using multiple sensory modalities simultaneously, meaning that effective learning is not restricted to a single ‘preferred’ style. The VAK (Visual, Auditory, Kinesthetic) model, which has been widely promoted in education, is not supported by empirical evidence. In fact, studies have shown that tailoring instruction to students’ self-identified learning styles does not improve learning outcomes.
Despite this, the belief in learning styles remains deeply ingrained in many classrooms. Rousseau’s research shows that “90 per cent of teachers strongly believe that their students are visual, auditory or kinesthetic” and often attempt to adjust their teaching methods accordingly. This misguided practice diverts valuable time and resources away from strategies that are genuinely evidence-based. Moreover, it risks limiting students by reinforcing a false sense of learning identity—one where a student labeled as an ‘auditory learner’ may feel less capable in subjects requiring strong visual-spatial skills, such as mathematics or geography.
One reason learning styles persist is the power of anecdotal experience and confirmation bias. Teachers observe students engaging with visual materials or listening to explanations and assume that these preferences reflect hardwired learning differences. However, as Rousseau notes, “It is not enough to present scientific evidence to persuade someone to abandon his or her deepest convictions. A frontal attack such as a scientific rebuttal text can even have the opposite effect and amplify the false belief.” This is why debunking the myth requires more than just presenting data—it demands experiential learning. Rousseau’s team has explored interventions where teachers experience firsthand how learning styles fail to impact memory retention, providing them with a counter-anecdote that challenges their prior assumptions.
Instead of reinforcing learning styles, educators should embrace research-backed approaches that support all learners. Presenting information in multiple modalities—combining visuals, text, discussion, and hands-on activities—has been shown to enhance learning because it strengthens neural connections and encourages deeper cognitive processing. This is not because students have rigid sensory preferences but because multimodal learning engages more brain regions and promotes better retention.
Rousseau’s work is a vital call to action for educators. The persistence of neuromyths like learning styles prevents the adoption of effective teaching strategies and misleads students into believing they have fixed ways of learning. The challenge is not only to correct misconceptions but to replace them with approaches that genuinely support student learning. If we are to improve education, we must finally put the myth of learning styles to rest and focus on the science of how students actually learn.
What the Research Says: Early Lessons from AI Tutoring That Matter Today
Is it too meta to write a blog post summarising someone else’s blog post? Perhaps—but if that person is Rose Luckin, then I think I can make an exception. Luckin’s latest contribution to the What the Research Says (WTRS) series dives into the foundational research behind AI tutors, drawing on two seminal studies: “Cognitive Tutors” (Anderson et al., 1994) and “Meta-Reviews on AI in Education” (du Boulay, 2016). She provides links to these studies, which explore the design principles that have made AI-powered learning tools effective over decades of research.
Luckin distills three enduring core principles for AI in education. First, step-based learning remains crucial. Anderson et al. (1994) demonstrated that breaking problems into manageable steps accelerates learning, with AI tutors enabling students to achieve proficiency in one-third the time of traditional instruction. Meta-analyses by du Boulay (2016) confirm that structured, step-based tutoring consistently outperforms conventional classroom teaching, reinforcing the need for AI tools that guide students through problem-solving processes rather than merely providing answers.
Second, immediate and targeted feedback is essential. Research shows that timely feedback enhances learning outcomes, but the quality of feedback matters as much as its timing. Short, specific feedback focused on why an answer is incorrect proves more effective than long-winded explanations or attempts at conversational AI. Du Boulay’s (2016) findings highlight that students engage more deeply when feedback is precise, actionable, and directly tied to their learning goals.
Third, AI tutors must accommodate multiple solution paths. Successful AI learning tools acknowledge that students may approach problems in different, yet equally valid, ways. Anderson et al. (1994) found that AI tutors were most effective when they supported diverse problem-solving strategies while maintaining clear pedagogical objectives. This adaptability remains a cornerstone of modern AI-enhanced education, ensuring that students with varied cognitive approaches can still benefit from structured learning support.
These principles, Luckin argues, should inform the next generation of AI-powered learning tools. By focusing on structured guidance, timely feedback, and adaptable learning pathways, AI can serve as a powerful complement to traditional teaching rather than a replacement for it. The challenge now is to ensure that AI implementation remains grounded in evidence-based pedagogy rather than the allure of technological novelty.
Beyond these principles, Luckin highlights the critical role of thoughtful implementation. AI tools work best when integrated into blended learning environments where teachers can customise and control their role. Effective implementation requires significant teacher training, allowing educators to understand AI’s capabilities and limitations while ensuring that technology enhances, rather than disrupts, the learning process. Additionally, measuring AI’s effectiveness should extend beyond immediate learning gains to include engagement, motivation, and long-term knowledge retention. Ultimately, Luckin’s insights reinforce that AI in education should be seen as an augmentation tool—one that empowers teachers and students alike, provided its use remains firmly rooted in research-driven pedagogy.
Disrupted History 