Naive vs. Purposeful vs. Deliberate Practice
Ericsson distinguished three modes of practice by the quality of engagement between practitioner and domain. Naive practice is repetition within the zone of established competence — the driver who has driven for twenty years without becoming better. Purposeful practice is directed effort toward specific goals, limited by the practitioner's self-knowledge of her own weaknesses. Deliberate practice adds a knowledgeable teacher or coach who perceives what the practitioner cannot and designs activities that target unrecognized weaknesses. The taxonomy is not merely categorical: the three modes produce dramatically different developmental trajectories, and practitioners who begin at the same level diverge exponentially depending on which mode predominates. In the AI era, the default mode of tool-assisted work falls closest to naive practice — productive in output terms, developmentally static in cognitive-architecture terms.
In the AI Story
The classification of AI-assisted work as developmentally naive requires some nuance. The practitioner is not mindlessly repeating an established routine; the output is too sophisticated for that characterization. What makes it naive in the developmental sense is that the practitioner is not engaging with the domain at the boundary of her capability, is not receiving feedback on her own performance, and is not constructing new cognitive structures through the iterative process of effortful engagement. The tool is doing the developing. The practitioner is doing the directing. Directing, while genuinely demanding, builds a different set of representations than doing.
The distinction between evaluative representations (built through directing and reviewing) and constructive representations (built through doing) is consequential. A developer who evaluates Claude's code is building representations of what correct code looks like. A developer who writes code is building representations of how to construct correct code — the problem-solving process itself, including false starts, dead ends, and moments of confusion that eventually resolve into understanding. The first is necessary but not sufficient for the second. A person can become an excellent film critic without being able to direct a film.
The transition from the default naive mode to something more developmental requires active effort against the tool's design. The tool optimizes for smoothness; deliberate practice requires difficulty. The practitioner who wants to develop through AI-assisted work must deliberately introduce friction, actively seek the discomfort the tool is designed to eliminate, maintain an effortful and critical orientation when passive acceptance would suffice. This is psychologically difficult and creates a recursive problem: using AI developmentally requires the metacognitive expertise that deliberate practice itself is supposed to build.
The organizational implication is that purposeful or deliberate modes of AI-assisted work will not emerge spontaneously at population scale. They require explicit structures — mentorship, assessment designed to reveal tool-free capability, cultural norms that value the deliberate seeking of difficulty, protected time for developmental engagement that is measured on growth rather than output. Without these structures, the default mode prevails. The default produces naive practice at scale: hours accumulated, expertise unchanged, the gap invisible until revealed by the situations that demand what AI cannot provide.
Origin
The three-mode taxonomy is an elaboration of distinctions present across Ericsson's research program, made explicit in his 2016 book Peak with Robert Pool. The taxonomy synthesizes findings from dozens of studies on what separates practitioners whose performance continues to improve from those whose performance plateaus.
Key Ideas
Naive practice produces arrested development. Repetition within established competence maintains current performance indefinitely but does not produce growth.
Purposeful practice is self-limited. Self-directed improvement is bounded by the practitioner's capacity to identify her own weaknesses — a capacity that is itself a form of expertise.
Deliberate practice requires external perspective. Knowledgeable teachers perceive and design for needs the practitioner cannot see.
AI defaults to naive. Without active intervention, tool-assisted work produces output without development.
Mode is in the practitioner, not the tool. The same tool can support any of the three modes depending on how the practitioner structures the engagement.
Appears in the Orange Pill Cycle
Further reading
- K. Anders Ericsson and Robert Pool, Peak (2016), ch. 5.
- Angela Duckworth, Grit: The Power of Passion and Perseverance (Scribner, 2016).
- Geoff Colvin, Talent Is Overrated (Portfolio, 2008).