Introduction
From what I see daily on social media and read in the specialized literature, much has been said about how artificial intelligence can help teachers teach more efficiently and students learn more efficiently. Yet very little is being said about its potential negative effects on teachers themselves: on their pedagogical competence, professional judgement, creativity and autonomy.
What I see as the greatest potential threat posed by AI to teacher expertise is professional deskilling: the gradual weakening of the knowledge, judgement and decision-making capacities that teachers develop through deep and sustained engagement with planning, teaching, assessment and reflection. AI may make teachers more productive without necessarily making them more competent. When teachers routinely delegate important pedagogical decisions to AI, they risk losing opportunities to develop and refine the expertise underpinning effective instruction.
Ten potential negative effects
1. Erosion of pedagogical content knowledge (PCK)
Teachers develop expertise by continually deciding how to transform subject knowledge into explanations, examples, tasks and representations that learners can understand.When AI routinely makes these decisions, teachers engage less deeply in this process. In MFL, this includes, for instance, choosing which grammatical structures to prioritise, anticipating linguistic difficulties, selecting appropriate examples, etc.
Research principle potentially undermined: Shulman’s (1986, 1987) concept of pedagogical content knowledge, emphasising the integration of subject knowledge and knowledge of how to teach it.
2. ‘Atrophy’ of instructional design skills
Designing effective lessons requires teachers to identify learning objectives, sequence content, anticipate misconceptions, manage cognitive demands and plan opportunities for practice and assessment with their students and learning setting in mind. Outsourcing lesson planning to AI obviously reduces opportunities to exercise these skills, particularly among novice teachers who have yet to develop robust instructional knowledge. This is very concerning, as the crisis we are currently experiencing in recruiting language learners at GCSE and A-level demands teachers with exceptional instructional design skills!
Research principle potentially undermined: Constructive alignment (Biggs, 1996), backward design (Wiggins & McTighe, 2005) and cognitive load theory (Sweller et al., 1998).
3. Weakening of diagnostic expertise
An important chunk of a teacher’s professional growth refers to the develop the ability to interpret student errors, distinguish between different causes of misunderstanding and identify appropriate interventions.If AI increasingly diagnoses learning difficulties and recommends corrective actions, teachers will eventually become less practised at independently interpreting evidence of learning! For example, an MFL teacher might accept an AI-generated explanation of listening difficulties without investigating whether the problem concerns phonological decoding, lexical recognition, grammatical parsing or working-memory demands. Another MFL teacher delegating corrective feedback on a written piece to a bot, will miss out on an incredibly useful window into their students’ interlanguage development.
Research principle potentially undermined: Teacher noticing and professional vision (van Es & Sherin, 2002), and formative assessment as eliciting, interpreting and acting upon evidence (Black & Wiliam, 1998).
4. Reduced capacity for critical pedagogical judgement
AI-generated recommendations appear very convincing, particularly when accompanied by plausible explanations or research references. Teachers may consequently become less inclined to question instructional choices, compare alternatives or scrutinise the assumptions underlying a suggested approach. The risk is a shift from independently evaluating pedagogical options to selecting among solutions generated by an algorithm.
Research principle potentially undermined: Reflective professional judgement (Schön, 1983), educational judgement (Biesta, 2015) and automation bias (Parasuraman & Riley, 1997).
5. Reduced reflective practice
Professional expertise develops to a great extent through reflection on instructional decisions: why an activity succeeded, why students struggled and what should be changed. If AI routinely generates lesson evaluations, identifies weaknesses and recommends improvements, teachers may engage less actively in the reflective processes through which professional understanding develops.
Research principle potentially undermined: Schön’s (1983) reflection-in-action and reflection-on-action, and Kolb’s (1984) experiential learning cycle.
6. Weakening of curriculum design expertise
Curriculum planning requires complex decisions about progression, sequencing, knowledge selection, cumulative practice and assessment. AI-generated units may be superficially coherent while failing to reflect learners’ prior knowledge or the developmental demands of the subject. When teachers increasingly adopt rather than construct curriculum sequences, they may lose opportunities to understand how knowledge accumulates over time.
Research principle potentially undermined: Curriculum coherence, constructive alignment (Biggs, 1996), and spacing and distributed practice (Cepeda et al., 2006).
7. Reduced pedagogical creativity
Professional creativity involves generating, testing and refining instructional approaches in response to particular learners and contexts. AI can support creativity by suggesting alternatives, but habitual reliance on ready-made materials may reduce teachers’ opportunities for inventive problem-solving. A related danger is pedagogical homogenisation, as teachers draw on similar AI-generated lesson structures and activities.
Research principle potentially undermined: Teacher agency (Priestley, Biesta & Robinson, 2015) and adaptive expertise (Hatano & Inagaki, 1986).
8. Declining professional autonomy
As AI platforms increasingly recommend instructional strategies, assessment approaches and learning interventions, decisions traditionally made by teachers may shift towards technological systems. This becomes particularly problematic when schools institutionalise AI-generated planning or algorithmic recommendations, reducing teachers’ discretion to adapt instruction to their learners.
Research principle potentially undermined: Teacher agency and professional autonomy (Priestley, Biesta & Robinson, 2015), and educational judgement (Biesta, 2015).
9. Reduced development of novice teachers
Novice teachers acquire expertise partly by undertaking demanding professional tasks, receiving feedback, reflecting on outcomes and progressively refining their decisions. If AI routinely performs lesson planning, resource development, assessment design and diagnostic reasoning for them, beginning teachers may miss opportunities to build foundational professional skills. They may also lack the expertise required to recognise weaknesses in AI recommendations.
Research principle potentially undermined: Development of teacher expertise (Berliner, 2004), pedagogical reasoning (Shulman, 1987) and deliberate practice (Ericsson, Krampe & Tesch-Römer, 1993).
10. The illusion of professional competence
This is what I keep seeing more and more when I visit schools. AI can make relatively inexperienced or less competent teachers appear highly competent by enabling them to produce sophisticated lesson plans, assessments, presentations and instructional materials. Yet the quality of these products does not necessarily reflect underlying understanding. A teacher may become increasingly efficient at generating materials while remaining unable to explain why particular activities are appropriate, diagnose their limitations or adapt them effectively.
Research principle potentially undermined: The distinction between performance and learning (Soderstrom & Bjork, 2015), applied cautiously to professional learning, and pedagogical reasoning (Shulman, 1987).
Summary: How AI may erode teacher expertise
| Dimension of expertise | Potential negative impact |
| Pedagogical content knowledge | Less independent transformation of subject knowledge into teachable content |
| Instructional design | Reduced practice in sequencing, scaffolding and planning |
| Diagnostic competence | Greater reliance on automated interpretations of learner difficulties |
| Critical judgement | Increased acceptance of algorithmic recommendations |
| Reflective practice | Less independent analysis of instructional successes and failures |
| Curriculum expertise | Reduced engagement with progression and cumulative learning |
| Creativity | Dependence on ready-made solutions and homogenisation of teaching |
| Professional autonomy | Transfer of pedagogical decisions towards AI systems |
| Novice teacher development | Fewer opportunities to acquire foundational professional skills |
| Professional self-evaluation | Confusion between AI-generated outputs and genuine teaching expertise |
What does the evidence actually establish?
There is a crucial distinction between well-established research on how professional expertise develops and direct evidence that generative AI causes teacher deskilling. Research by Shulman, Schön, Berliner and others provides strong theoretical reasons to be concerned about outsourcing professional thinking. Research on automation bias demonstrates that people can become excessively reliant on automated recommendations. However, longitudinal evidence showing that regular generative AI use causes measurable deterioration in teachers’ pedagogical content knowledge, diagnostic competence or professional judgement remains limited. These are credible risks, not established universal outcomes!
The distinction also matters between AI that replaces professional reasoning and AI that stimulates it. A teacher who critically evaluates, challenges and improves AI-generated lesson plans may strengthen rather than weaken their expertise.
The central paradox
AI can dramatically increase the speed at which teachers produce lessons, worksheets, assessments and feedback. But the intellectual work involved in creating those materials is not merely an administrative burden to be eliminated. It is often part of the process through which teachers develop pedagogical expertise.
The danger is that schools may celebrate gains in teacher productivity while overlooking a gradual reduction in opportunities for teachers to practise the complex professional reasoning that makes them effective.
AI may help teachers do more, faster, while giving them fewer opportunities to become better at what they do.
That is arguably the most important long-term concern about AI and teacher professional development.
Selected research references
Berliner, D. C. (2004). Describing the behavior and documenting the accomplishments of expert teachers. Bulletin of Science, Technology & Society, 24(3), 200–212.
Biesta, G. (2015). What is education for? On good education, teacher judgement, and educational professionalism. European Journal of Education, 50(1), 75–87.
Black, P., & Wiliam, D. (1998). Assessment and classroom learning. Assessment in Education, 5(1), 7–74.
Cepeda, N. J., et al. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380.
Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406.
Hatano, G., & Inagaki, K. (1986). Two courses of expertise. In Child Development and Education in Japan.
Parasuraman, R., & Riley, V. (1997). Humans and automation: Use, misuse, disuse, abuse. Human Factors, 39(2), 230–253.
Priestley, M., Biesta, G., & Robinson, S. (2015). Teacher Agency: An Ecological Approach. Bloomsbury.
Schön, D. A. (1983). The Reflective Practitioner. Basic Books.
Shulman, L. S. (1986). Those who understand: Knowledge growth in teaching. Educational Researcher, 15(2), 4–14.
Shulman, L. S. (1987). Knowledge and teaching: Foundations of the new reform. Harvard Educational Review, 57(1), 1–22.
Soderstrom, N. C., & Bjork, R. A. (2015). Learning versus performance: An integrative review. Perspectives on Psychological Science, 10(2), 176–199.
van Es, E. A., & Sherin, M. G. (2002). Learning to notice: Scaffolding new teachers’ interpretations of classroom interactions. Journal of Technology and Teacher Education, 10(4), 571–596.
