Ei-iE

Artificial intelligence (AI) is unlocking both opportunities and challenges for education, with teachers on the front line of this paradox navigating questions of how to best use AI for learning and knowledge gain.

The report I recently authored with Pr. Jason M. Lodge, Artificial intelligence, cognitive offloading and implications for education, investigates a profound new challenge: the risk that students will outsource too much of the cognitive work that is crucial to establishing the knowledge, skill and ‘thinking infrastructure’ that enables both schooling success and lifelong capacity for learning and understanding.

There is a growing body of evidence that using AI can short-circuit the cognitive effort required for sustainable, deep learning, thus creating ‘false mastery’ with potentially long-term consequences. This cognitive offloading from human to AI is especially risky for school students (‘novice’ learners who are building foundational knowledge and skills) when they turn to AI as a tempting substitute, not an amplifier, increase their dependency on the tool and lose access to deeper learning.

The true educational risk of AI therefore is not simply that students will use it to cheat on an essay. The far more profound risk is that AI may fundamentally interfere with the cognitive processes of knowledge construction and verification, the very processes that build the long term memory stores and subsequent skills upon which critical thinking depends.

While there can be both beneficial and harmful offloading with AI, emerging data support the observation that unstructured AI use trends toward detrimental offloading, creating a performance paradox: students’ short-term performance on tasks improves, while their durable, long-term learning is harmed. This trend appears to be driven by the fluency of AI-generated output, which creates an illusion of competence and encourages metacognitive laziness, leading learners to abdicate the generative effort required to build deep knowledge.

AI, when used as an answer oracle, is the ultimate passive review tool. It allows the learner to bypass the generation effect entirely. By providing the answer, the solution, or the essay, it robs the learner of the very cognitive struggle that is necessary to build lasting knowledge.

Cognitive outsourcing also introduces extra equity risks. Research suggests that students who possess high levels of content knowledge and strong metacognitive skills are better able to leverage AI to accelerate and deepen their learning and critical thinking. Conversely, students lacking such skills, often those already experiencing disadvantage, are potentially more susceptible to harmful offloading and missing the learning they need.

The good news is that these harmful effects can be counteracted through purposeful teaching and learning strategies and effective design of AI education technology. These strategies reinforce the importance of quality teaching, with AI in a subsidiary, supporting position. The most powerful use of AI in K–12 education may not be to replace or bypass the teacher with AI tutors, but to augment the teachers themselves.

The pedagogical solution: From cognitive atrophy to augmentation

The evidence for detrimental offloading is compelling, but it is not deterministic; our report outlines three key pathways to promote effective AI use.

Path 1: Beneficial offloading and load reduction instruction

AI can be used for beneficial offloading, managing extraneous load to free up resources for intrinsic learning. This requires an explicit pedagogical framework such as Load Reduction Instruction, which adapts explicit instruction principles useful for managing learning with AI. Drawing on this model, AI can be used to provide scaffolding, structured practice, and feedback, all aimed at managing the cognitive burden on the learner and enabling progressive independence, in other words, helping students to become better self-regulated learners. As validated by recent studies, students explicitly taught this cognitive offload instruction model (such as offloading low-order writing tasks) showed significantly greater gains in critical thinking.

The key is to apply evidence-backed approaches like explicit teaching and cognitive load theory when using generative AI. The NSW Department of Education, Australia’s largest public education system, provides training for teachers in how to incorporate these practices when prompting AI to produce lesson plans and activities that enable progressive independence and mastery. This can involve incorporating case studies or research of load reduction instruction into the prompt, for example.

Path 2: Scaffolding metacognition to counter laziness

The more profound solution is one that directly tackles the core problem of metacognitive laziness. If the problem is that the convenience of AI encourages learners to abdicate their metacognitive responsibilities, the solution is to design AI interactions that explicitly demand and scaffold those responsibilities. While these design parameters may not be within the direct control of educators, these kinds of prompts can be used in a wide range of scenarios to help students develop these capabilities. It will be helpful when AI tools increasingly have these capabilities built in, but technology is not required for teachers to use these approaches.

Teachers can help combat metacognitive laziness by bringing their professional expertise in building student metacognitive capacity and self-regulation explicitly into the AI context. For example: teaching students how to use AI as a tool to check and challenge their reasoning, thus affording opportunities for learning and reflection; getting students to check AI claims by investigating source information; or asking AI to explain a maths process differently, provide worked examples or generate practice problems.

Path 3: Designing AI as a cognitive mirror and verification partner

The most advanced pedagogical and technological design solutions shift the fundamental role of AI from an answer oracle (which invites passive outsourcing) to a tool that provokes intrinsic cognitive load.

• AI as cognitive mirror: The AI is engineered as a teachable novice with a pedagogically useful deficit. It feigns confusion and asks clarifying questions, forcing the human learner into the effortful, generative act of explanation and reflection, thus triggering the generation of knowledge creation.
• AI as Socratic partner: AI is used to create desirable difficulties. Instead of bypassing effort, the AI is used as a cognitive partner to generate retrieval-practice questions, case studies, and Socratic dialogues that force the effortful processing required for durable learning.
• AI as verification partner: Ensuring a model of intelligence equilibrium where the human maintains primary cognitive agency and continuously evaluates and corrects the AI output, guided by a verification mindset.

Conclusion

In a world where cognitive and metacognitive offloading is the norm, the educational imperative is to ready students for it, through deep knowledge and adaptive, transferable skills. This preparation has two non-negotiable components:

• Arming students with the deep, domain-specific knowledge and analytical thinking capabilities they need to think critically about the fluent, unreliable output AI can and does generate
• Fostering the robust metacognitive judgement and self-regulated learning skills they need to think critically with it, avoiding detrimental offloading and abdication

While the extent and scale to which students shift their knowledge and skill acquisition to AI raises fundamental questions for teaching and learning, it also brings an important opportunity to validate and bolster the role of teachers. AI may be a new technological vector for education, but the strategies for its successful integration require a strong pedagogical response: enhancing the central role of teachers; drawing upon well-researched approaches for quality teaching and learning; and ensuring close attention to equity.