The Good-Enough Machine
Winnicott's good-enough mother is neither perfectly responsive nor negligently absent. She provides enough frustration for the infant to develop its own capacities without providing so much that the infant is overwhelmed. Phillips proposes the good-enough machine as an analogous ideal for AI tool design. The perfect machine — the one that anticipates every need and fulfills every desire before it is articulated — is not good enough; it is too good, depriving the user of the frustration that is the precondition for creative development. The negligent machine, which fails constantly, is too poor. The good-enough machine provides enough capability to support the creator's ambition while preserving enough friction to sustain the playing that is the source of all genuine creation.
In the AI Story
The argument cuts directly against the dominant direction of AI tool development, which has been organized around the elimination of friction. Seamlessness is treated as an unqualified virtue. The model that requires fewer clarifying questions, that produces cleaner output on the first pass, that requires less user intervention, is considered better by every metric the industry currently measures. Phillips's framework suggests that these metrics are not measuring what they think they are measuring. They are measuring the degree to which the tool has eliminated the developmental friction its users need.
The good-enough machine would be, by current industry standards, a worse product. It would sometimes refuse. It would sometimes require the user to try again. It would preserve moments of not-knowing, of productive confusion, of the genuine surprise that emerges when the tool and the user meet at the edge of what either can do alone. These moments are not bugs. They are the features that make the collaboration developmental rather than merely productive.
The proposal connects to ascending friction but extends it. Ascending friction describes what happens automatically: difficulty relocates to a higher floor. The good-enough machine is a design principle: we should deliberately preserve certain kinds of difficulty rather than optimize them away. The principle is hard to operationalize because the kinds of difficulty that develop the user are not the kinds that show up in user satisfaction surveys, and the industry's feedback loops are all tuned to the surveys.
Phillips would not claim that every friction is developmental. Some friction is genuinely a tax, and removing it is genuinely liberating. The work is to distinguish — and the distinguishing cannot be done by the tool, because the tool cannot see the developmental trajectory of the user. It has to be done by the user, or by institutions that take responsibility for user development over time.
Origin
Phillips draws the analogy from Winnicott's 1953 paper "Transitional Objects and Transitional Phenomena" and elaborates it through his 1988 book Winnicott. The extension to technology echoes arguments made by critics of interface design from Lucy Suchman to Sherry Turkle, though Phillips's psychoanalytic framing is distinctive.
Key Ideas
Sufficient, not maximal. The best tool is not the one that does the most for the user but the one that does enough to support what the user can do for herself.
Preserved friction as a design choice. Some frustrations should be engineered in rather than out — not out of nostalgia but out of developmental realism.
The user's trajectory as design criterion. Tools should be evaluated not only by the quality of the output they produce today but by what they do to the user's capabilities over months and years of use.
Against seamlessness. The smooth interface is not merely an aesthetic; it is a statement about what the user is for. Phillips's framework treats that statement as a question, not an answer.
Debates & Critiques
The principle is difficult to implement in a market that rewards seamlessness and punishes friction. Some researchers have proposed tutoring systems that deliberately withhold answers to encourage learning; these represent the good-enough principle in miniature. Whether the broader consumer market will ever reward such tools, or whether they will remain confined to educational contexts, is an open question that bears directly on the developmental trajectory of an entire generation.
Appears in the Orange Pill Cycle
Further reading
- D.W. Winnicott, "Transitional Objects and Transitional Phenomena" (1953)
- Adam Phillips, Winnicott (Harvard University Press, 1988)
- Sherry Turkle, Alone Together (Basic Books, 2012)
- Lucy Suchman, Human-Machine Reconfigurations (Cambridge University Press, 2007)