Written evidence by Dr Colin Fu and Dr Athina Ioannou (AIE0151)

Education Committee

The use of Artificial Intelligence and EdTech in Education inquiry

 

1. Introduction and Authorial Authority

1.1 I am Dr Colin Fu, a professional practitioner cum an educator at University College London, specializing in the intersection of Marketing Science, Market Intelligence, and Cognitive AI Frameworks. My career has focused on the architecture of human-machine interaction and the behavioural economics of automated decision-making. My work centres on the practical application of AI in high-stakes commercial environments where the cost of "False Certainty" is high. 

1.2 My research and frameworks have been featured in QS Asia Summit, and I am frequently invited to speak internationally on the ethics and strategy of AI at global forums, including the United Nations (UN) PRME Global Forum on Analytics and keynote summits in China, where I advise on the systemic risks of algorithmic dependency and the future of human-centric AI.

1.3 I am joined by Dr. Athina Ioannou, a Senior Lecturer in Business Analytics at the University of Surrey. Dr. Ioannou brings an extensive academic background in Computer Science and Business Information Systems, with specific research expertise in Privacy and Data Management, AI Adoption, and the Adaptive/Maladaptive uses of IT (such as Technostress). Her work on the social and organizational consequences of emerging technologies provides the empirical grounding for the cognitive and privacy-related risks discussed in this evidence.

1.4 Our joint submission is motivated by a critical systemic risk identified through our combined professional and academic practice: the UK education system is currently treating Artificial Intelligence (AI) as a "Functional Utility" (a tool for output efficiency) rather than a "Cognitive Architecture" (a system that fundamentally reshapes the human brain’s processing of information and validation of truth).

1.5 We are providing evidence on the Certainty Paradox, which is a phenomenon where high-fluency AI outputs create a psychological "Confidence Gap" in human judgment. Without a specialized pedagogical intervention, we risk a generation of students who are technically capable of generating high-quality content but cognitively incapable of verifying, auditing, or applying it within the noisy, uncertain contexts of the modern digital economy.

2. Executive Summary

2.1 The core of our evidence rests upon the observation that as AI systems become more linguistically sophisticated, they create a Certainty Illusion that effectively bypasses the critical inquiry loops of the human brain. We argue that the English education system must pivot from assessing outputs to governing cognitive processes.

2.2 We identify three primary threats to the English education system that necessitate immediate policy intervention:

2.3 To address these threats, we propose five primary pillars of opportunity that define the future of EdTech policy:

3. The Certainty Paradox: The Hidden Cost of AI Accuracy

3.1 In our professional practice within our disciplines, we observe that "Confidence" is often used as a persuasive tool to drive consumer behaviour. Within an educational context, this same confidence creates a dangerous "Automation Bias."

3.2 Large Language Models (LLMs) are essentially high-dimensional statistical predictors. They do not possess an internal model of truth; rather, they predict the next most likely token in a sequence. However, through Reinforcement Learning from Human Feedback (RLHF), these models have been trained to provide answers that are grammatically perfect and authoritative in tone.

3.3 This creates the Certainty Paradox. As the machine’s accuracy increases, the student’s internal "Error Detection" mechanism decreases. When a student receives an answer that is visually and linguistically flawless, they are psychologically nudged to accept it as an absolute truth. This leads to a catastrophic loss of "Productive Struggle," which is the cognitive friction required for the brain to form long-term memories and logical frameworks.

3.4 Without a curriculum that explicitly teaches the navigation of this paradox, we are training a workforce that can execute prompts but cannot identify when a "certain" machine output is fundamentally flawed or logically inconsistent.

4. Market Intelligence and the EdTech Ecosystem

4.1 The Committee must recognize that the EdTech market is not currently optimized for educational outcomes but for "Engagement Metrics." Marketing Science principles such as A/B testing, gamification, and "low-friction" user experiences are being used to maximize student time on platform.

4.2 While "low friction" is a virtue in consumer apps, it is often a vice in education. Learning is an inherently "high-friction" process. By removing the difficulty of research and synthesis, EdTech tools may be inadvertently limiting the cognitive ceiling of the student population.

4.3 Furthermore, we identify a risk of "Geopolitical Cognitive Dependency." If the UK education system relies exclusively on proprietary AI models from international technology firms, we lose the "Neuro-Sovereignty" required to audit the ethical and logical biases embedded in those models. We require a sovereign framework for Cognitive AI that prioritizes British educational values and critical independence.

5. Technostress and Maladaptive IT Use in Education

5.1 Dr. Ioannou’s research into Technostress provides critical evidence regarding the psychological impact of rapid AI integration. As AI tools become embedded in the classroom, both teachers and students face "Techno-Overload" and "Techno-Insecurity."

5.2 Students face a new type of academic pressure: the "Comparison Gap" between their own raw, human efforts and the polished, synthetic outputs of AI. This can lead to a maladaptive reliance on technology to avoid the stress of "human failure," ultimately resulting in a generation that lacks the resilience to engage in independent thought.

5.3 For educators, the lack of support and training leads to "Techno-Complexity." Without a clear framework for when to use AI and when to prohibit it, teachers are experiencing increased workloads as they attempt to manage the "Arms Race" between AI-generated homework and traditional assessment methods.

6. The Five-Level Cognitive AI Framework: Rebuilding Epistemic Agency

6.1 To resolve these challenges, we propose the national adoption of a proprietary pedagogical model for AI interaction. This framework ensures that the student remains the "Critical Architect" of the interaction, preventing the passive acceptance of machine-generated outputs.

6.2 Level 1: Explore (). The student must treat the AI as a "Probabilistic Map." This involves curiosity and hypothesis framing, requiring the generation of multiple variations of a prompt to identify the statistical signal within the machine's noise.

6.3 Level 2: Engage (). The student must participate in an "Iterative Stress-Test" of the machine’s logic through critical dialogue and interrogation. They must challenge the AI’s first draft, moving beyond passive consumption to active intellectual friction.

6.4 Level 3: Examine (). This is the audit and evaluation phase. Drawing on Business Analytics skills, the student performs bias detection and verifies the AI's "Confidence" against external, peer-reviewed sources, documenting the "Noise" in the machine’s output.

6.5 Level 4: Formulate (). The student must "take back the wheel." This stage involves creation, synthesis, and ethical weighting, using raw AI data to construct a human-led original synthesis.

6.6 Level 5: Reflect (). This meta-cognitive stage requires the student to reflect on values and transferability. They must ask if they accepted an answer because it was "right" or simply because it was "easy," evaluating how the machine's "Certainty" influenced their own bias.

6.7 To quantify the efficacy of this pedagogical model, we have developed a mathematical formalization of the learning process (). This model utilizes a weighted geometric mean to demonstrate that total learning is constrained by the weakest stage in the cognitive chain:

where

)

6.8 This formula is currently being experimented with to evaluate the depth of student engagement. The geometric mean structure is critical for policy considerations because it proves that excellence in "Exploration" cannot compensate for a zero-value in "Examination" or "Reflection." If any single stage of the framework is neglected, the total learning value () trends toward zero. This mathematical reality supports our call for a curriculum that treats all five levels as mandatory, non-negotiable components of digital literacy.

7. The Impact on Teaching and Assessment: A Move Toward Cognitive Telemetry

7.1 Empirical Validation: We have rigorously tested this methodology across diverse modules at the University of West London, the University of Surrey, and University College London (UCL). These tests involved students at various stages of academic development, from undergraduate to postgraduate levels. Our findings consistently demonstrate that when students are held accountable for the process rather than just the result, their engagement with the underlying logic of the subject matter increases significantly.

7.2 The Death of Traditional Assessment: The growing availability of EdTech and AI has fundamentally broken traditional assessment methods. We must accept that coursework completed outside the classroom is no longer a reliable or valid measure of a student’s skills or knowledge.

7.3 The Solution: We propose a revolutionary shift from "Output-Based Assessment" to "Process-Based Assessment" using Cognitive Telemetry. In Marketing Science, we do not merely care if a customer made a purchase; we audit the entire "Customer Journey." We must apply this same logic to the "Learner Journey."

7.4 Process-Based Auditing: Cognitive Telemetry involves the automated recording and auditing of the interaction logs between a student and an AI. By measuring the student’s ability to navigate the Five-Level Framework (their prompt history, their critical audits, their revisions), we can grade the "Quality of Thought" rather than the "Quality of Output." This makes AI-generated "cheating" irrelevant, as the assessment is based on the student's human interaction with the machine.

8. Children’s Digital Rights and Cognitive Sovereignty

8.1 Children’s rights must be expanded to include "Cognitive Sovereignty." This is the right to develop an independent mind free from algorithmic "nudging" and profiling.

8.2 Dr. Ioannou’s work in Privacy and Data Management suggests that EdTech companies must be held to a higher standard of transparency. We recommend that the Government mandate "Uncertainty Disclosures" on all AI tools used in schools. This means every AI output should be accompanied by a "Confidence Score" or a visual indicator of its mathematical uncertainty.

8.3 This transparency empowers children to exercise "Mindfulness in Information System," allowing them to maintain a critical distance between the machine’s authoritative tone and their own emerging judgment.

9. Strategic Recommendations for the Government

9.1 National Curriculum Review. The curriculum should be adapted to prioritize "Statistical Thinking" and "Decision Science" over rote content retrieval which we define as the process of memorizing information through sheer repetition without necessarily understanding the underlying logic, cause-effect relationships, or context. In an era where AI can perform instantaneous retrieval of any factual data point, a curriculum based on "memorization" is not only obsolete but actively encourages Cognitive Atrophy. Students must instead be trained to handle "Noisy Data" and evaluate the probabilistic reliability of machine-generated information.

9.2 Regulatory Quality Marks. Establish a "Cognitive Integrity" quality mark for EdTech tools. Tools that prioritize "User Engagement" over "Productive Struggle" should not be approved for use in state settings.

9.3 Teacher CPD. Professional development should focus on training teachers to be "Uncertainty Architects." Their role should shift from being the "Source of Truth" to being the "Auditor of Logic."

10. Suggested Questions for the Committee to ask the Government

10.1 On Output vs. Growth: "How is the Government distinguishing between 'Output Efficiency' (the speed of task completion) and 'Cognitive Growth' (the retention and synthesis of knowledge) in its evaluation of AI in schools? Specifically, how will the Government ensure that EdTech adoption results in genuine learning rather than long-term Cognitive Atrophy?"

10.2 On Pedagogy vs. Utility: "How does the Government intend to regulate AI as a 'Cognitive Architecture' (a system that reshapes how students think) rather than just a 'Functional Utility', ensuring that the use of AI enhances the brain's reasoning rather than bypassing it?"

10.3 On Market Accountability: "What regulatory safeguards are being developed to prevent private EdTech firms from applying Marketing Science techniques, such as frictionless engagement loops and algorithmic nudging, which prioritize platform retention over the 'Productive Struggle' necessary for student mastery?"

10.4 On the Certainty Paradox: "Does the Government recognize the Certainty Paradox as a systemic risk to the UK workforce? How will the curriculum be adapted to teach 'Signal-to-Noise Discernment' so that future employees can audit probabilistic AI outputs instead of falling victim to Automation Bias?"

10.5 On Assessment Innovation: "Given that our empirical testing across UCL, the University of Surrey, and UWL suggests that traditional outputs are now easily spoofed, what investment is the Government making in Cognitive Telemetry to move national assessments toward a process-based model that evaluates the 'health of a student’s logic'?"

10.6 On Neuro-Sovereignty: "Will the Government mandate 'Uncertainty Disclosures' for all AI tools used in state education, ensuring that children’s Neuro-Sovereignty is protected by providing transparency regarding the machine's confidence levels?"

11. Conclusion

11.1 In the digital economy, AI will generate the world’s "outputs." The humans who will succeed are those who possess the "Statistical Thinking" to decide what to trust. If the UK education system rewards "Certainty" and "Correctness" without assessing the "Process of Judgment," we will deskill our workforce for the very era we are trying to prepare them for. We must protect the cognitive sovereignty of the next generation by teaching them to navigate the paradox of certain-looking machine intelligence.

12. Technical Terms

12.1 Core Theoretical Definitions

12.2 Technical & Methodological Definitions

12.3 Market & Regulatory Definitions

12.4 Psychological & Behavioural Definitions

May 2026