Nursing lenses
- Nurses
- Students
- Educators
- Practice supervisors
Novice to expert
Benner (1984, 2009), Tanner (2006) & 2020–2026 Cognitive Offloading & AI Literacy Scholarship.
You can only spot what is wrong in an AI output if you already know what right looks like.
An Expert adult ICU nurse moving into community nursing or AI prompting is a Novice in that new domain.
Patricia Benner's From Novice to Expert (1984) applied the Dreyfus model of skill acquisition to nursing. She described five stages: novice, advanced beginner, competent, proficient and expert. As nurses gain experience, they move from following rules to seeing situations as a whole and responding intuitively.
This matters for AI because you can only spot what is wrong with an AI output if you know what right looks like. The same tool is safer in some hands than in others.
Interactive 5-Stage Progression & Supervision Stepper
Match Safe AI Use & Supervision to Benner’s Five Stages
Select a stage of clinical experience and switch between the Learner / Nurse View and Practice Supervisor / Assessor (SSSA) View.
Stage 1: Novice (Year 1–2 student nurse, or any clinician entering a completely new specialty or digital domain)
Clinical judgement profile: Tanner (2006) Noticing & Interpreting: Novices rely on context-free rules and have not yet built the clinical mental schemas to spot what a fluent AI output omits.
Progression across Benner's five stages
Safe use at each stage
| Stage | Typically | Safe use of AI | Watch out for |
|---|---|---|---|
| Novice | Students early in training; anyone new to a field | Practice and rehearsal with invented cases; explanations to check against teaching | Using AI answers as a substitute for learning the rules |
| Advanced beginner | Later students; newly registered nurses | Drafting, with every line checked by you and reviewed with a supervisor or preceptor | Trusting output you can't yet fully judge |
| Competent | A few years in one area | Planning and structuring work; spotting gaps in drafts | Efficiency pressure crowding out checking |
| Proficient | Sees the whole situation and notices when something is off | Using AI to test your thinking, and to catch what is missing | Over-confidence in tools that agree with you |
| Expert | Deep, intuitive grasp of practice | Shaping how teams use AI; teaching others to question it | Your intuition doesn't transfer to the AI, so explain your reasoning |
These stages describe experience in a particular area, not seniority. An expert in one specialty may be a novice in another, and a novice in AI.
For supervisors, assessors and preceptors
Verify the reasoning
Ask learners to show their checking against NICE, BNF or local policy, not just hand in a polished final output.
Practise spotting flaws
Use AI outputs with deliberate errors (such as Spot the error) as teaching material to build clinical scepticism.
Set clear boundaries
Agree with each learner which uses of AI are appropriate at their current stage of competence.
2006–2026 theoretical underpinnings: clinical judgement, cognitive offloading and AI supervision
Subsequent nursing research and 2020–2026 clinical AI evidence have deepened Benner's (1984) model, showing why unverified AI poses sharply different risks across stages of clinical development and how educators and supervisors must protect the formation of clinical judgement.
Tanner’s Clinical Judgment Model
Christine Tanner (2006) synthesised over 200 studies into four phases of clinical judgement: Noticing, Interpreting, Responding, and Reflecting. Novices struggle most with Noticing salient clinical cues. If an AI tool summarises a case before the learner has assessed the patient, it bypasses Noticing and Interpreting entirely, stunting clinical reasoning.
Situated Expertise & Professional Formation
Benner, Tanner & Chesla (2009) and Benner et al. (2010, Educating Nurses) demonstrate that nursing expertise is not merely faster rule-following; it is embodied, emotionally attuned perception and moral agency built through situated bedside experience. An AI system remains permanently bound to context-free rules and statistical patterns.
Automation Bias & Diagnostic Over-Reliance
Goddard, Roudsari & Wyatt (2012) and Jabbour et al. (2023, JAMA) show that clinicians exposed to incorrect AI predictions frequently accept flawed outputs—even when fluent AI “explanations” are provided. Less experienced clinicians are particularly vulnerable to commission errors (following wrong AI advice) and omission errors (failing to notice what AI missed).
Cognitive Offloading & Deskilling Risk
Gerlich (2025) and contemporary health professions education research (2024–2026) demonstrate a significant negative correlation between frequent uncritical AI reliance, cognitive offloading, and critical thinking performance—most pronounced in younger and novice learners who have not yet consolidated foundational mental schemas.
Council of Deans Principles & Two-Lane Assessment
Council of Deans of Health (2026) Principles for Use of AI in Healthcare Education align directly with Benner’s progression: protecting foundational clinical competence through assured, AI-free assessments (Lane 1 OSCEs and bedside PAD/MORA) while teaching structured verification in AI-integrated coursework (Lane 2).
Role- & Experience-Matched AI Literacy
EU AI Act Article 4 (in force February 2025) and the UK Government Response to the National Commission on AI in Healthcare (2026) require AI literacy to take into account the user’s technical knowledge, experience, education, and clinical context—making stage-matched supervision a regulatory expectation.
Theoretical evolution: from Dreyfus & Benner (1980–1984) to 2026 AI supervision
Go deeper: full milestone table with sources (6 rows)
| Year / Period | Key Scholars / Source | Core Theoretical Contribution | What It Means for Nursing & AI Today |
|---|---|---|---|
| 1980 | Dreyfus & Dreyfus | Five-stage model of human skill acquisition (Novice, Advanced Beginner, Competent, Proficient, Expert) | Shows why human situational intuition cannot be reduced to formal symbolic rules or decontextualised checklists |
| 1984 | Benner (From Novice to Expert) | Applies the Dreyfus model to clinical nursing practice | Establishes that novices rely on context-free rules and cannot spot subtle clinical omissions in fluent AI drafts |
| 2006 | Tanner (Journal of Nursing Education) | Clinical Judgment Model: Noticing, Interpreting, Responding, Reflecting | Explains why students must form their own clinical impression (Think First) before prompting an AI tool (Prompt Second) |
| 2009–2010 | Benner, Tanner & Chesla (2009); Benner, Sutphen, Leonard & Day (2010) | Expertise in Nursing Practice and Educating Nurses: clinical reasoning as situated coaching and moral formation | Guides Practice Supervisors and Practice Assessors (NMC SSSA) to question how a learner reasoned, not just inspect written text |
| 2012–2023 | Goddard et al. (2012); Jabbour et al. (2023, JAMA) | Quantifies automation bias and shows that plausible AI explanations can paradoxically increase clinician error | Reinforces active verification against primary UK sources (NICE, BNF, local policy) at every stage of competence |
| 2025–2026 | Gerlich (2025); EU AI Act Art. 4 (2025); Council of Deans of Health (2026) | Links cognitive offloading to deskilling in novices and mandates experience-matched AI literacy and Two-Lane assessment | Anchors stage-specific guardrails across pre-registration curricula, preceptorship, and specialist clinical practice |
Links to NCD learning stages
The NCD stages map loosely onto Benner's: Foundation (novice to advanced beginner), Builder (competent) and Advanced (proficient to expert).
Why is an unverified AI-drafted care plan more dangerous in the hands of a Year 1 nursing student (Novice) than an experienced specialist nurse (Proficient/Expert)?
Reading
- Benner, P. (1984) From Novice to Expert: Excellence and Power in Clinical Nursing Practice. Menlo Park, CA: Addison-Wesley.
- Dreyfus, S.E. and Dreyfus, H.L. (1980) A Five-Stage Model of the Mental Activities Involved in Directed Skill Acquisition. Berkeley: University of California Operations Research Center.
- Tanner, C.A. (2006) Thinking like a nurse: a research-based model of clinical judgment in nursing. Journal of Nursing Education, 45(6), pp. 204–211. doi:10.3928/01484834-20060601-04.
- Benner, P., Tanner, C.A. and Chesla, C.A. (2009) Expertise in Nursing Practice: Caring, Clinical Judgment, and Ethics. 2nd edn. New York: Springer Publishing.
- Benner, P., Sutphen, M., Leonard, V. and Day, L. (2010) Educating Nurses: A Call for Radical Transformation. San Francisco: Jossey-Bass (Carnegie Foundation for the Advancement of Teaching).
- Goddard, K., Roudsari, A. and Wyatt, J.C. (2012) Automation bias: a systematic review of frequency, effect mediators, and mitigators. Journal of the American Medical Informatics Association, 19(1), pp. 121–127. doi:10.1136/amiajnl-2011-000089.
- Jabbour, S., Fouhey, D., Shepard, S., Valley, T.S., Kazerooni, E.A., Banovic, N., Wiens, J. and Sjoding, M.W. (2023) Measuring the impact of AI in the diagnosis of hospitalized patients: a randomized clinical vignette survey study. JAMA, 330(23), pp. 2275–2284. doi:10.1001/jama.2023.22295.
- Gerlich, M. (2025) AI tools in society: impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), 6. doi:10.3390/soc15010006.
- Council of Deans of Health (2026) Principles for Use of Artificial Intelligence (AI) in Healthcare Education. London: CoDH.