Playing and Producing
Playing and producing, in the Winnicottian-Phillipsian framework, are not merely different activities but different modes of being. Playing is characterized by the absence of a predetermined outcome: the player does not know what she is looking for until she finds it, and the finding is the playing. Producing is characterized by a predetermined outcome: the producer knows what she wants and executes the steps that will deliver it. Both modes can produce artifacts. Only playing produces the kind of artifact that carries the charge of the real — the surprise, the unanticipated connection, the emergence of meaning that no one planned. The machine can produce. It cannot play. And the builder who treats AI collaboration as production rather than as playing forecloses the surprise that makes the collaboration creative rather than merely efficient.
In the AI Story
The distinction becomes urgent in the AI moment because the tool is optimized for production. It is extraordinarily good at generating outputs that meet specifications. It is structurally not good at playing — at remaining in the unstructured state where the outcome is genuinely undetermined. The builder who wants to play with AI must resist the tool's optimization for production, must insist on formlessness when the tool pushes toward form, must treat its outputs as provocations for further playing rather than as finished products.
Adam Phillips's extension of Winnicott emphasizes that playing requires the tolerance of frustration. The child cannot play if her frustration is immediately eliminated by adult intervention. The builder cannot play if her not-knowing is immediately eliminated by AI query. Frustration is not a regrettable obstacle to playing; it is the condition that makes playing necessary and productive. The AI that eliminates frustration with instant answers eliminates the soil from which playing grows.
The organizational implication is that productivity metrics and playing are in structural tension. Measurements of output, efficiency, completion, and delivery are measurements of producing. Playing is invisible to these metrics — it shows up, if at all, as time unaccounted for, as work that seems less efficient than it could be, as processes that resist streamlining. Organizations that want the output of playing (the genuine surprise, the unexpected connection, the real rather than the competent) must protect space for playing against the metrics that would optimize it out of existence.
Origin
Adam Phillips developed the distinction across On Kissing, Tickling, and Being Bored (1993) and subsequent works, drawing directly on Winnicott's analysis of playing in Playing and Reality. The distinction has been taken up in AI discourse by writers engaging with what the tools can and cannot do at the level of creative mode rather than capability.
Key Ideas
Different modes, not different techniques. Playing and producing are distinct ways of being that shape the whole engagement.
Machines produce; they cannot play. Optimization for predetermined outcomes structurally excludes the playing mode.
Frustration is constitutive. Playing requires the tolerance of not-knowing; instant answers eliminate the condition that makes playing possible.
Invisible to productivity metrics. Playing appears as inefficiency and must be protected against the pull of measurable production.
Appears in the Orange Pill Cycle
Further reading
- Adam Phillips, On Kissing, Tickling, and Being Bored (Harvard University Press, 1993)
- D.W. Winnicott, Playing and Reality (Tavistock, 1971)