Agency and Authorship (Crawford)
Agency and authorship, in Crawford's philosophical vocabulary, names the distinction between being the cause of effects in the world through one's own skilled execution and being the specifier of outcomes whose execution is performed by systems one neither fully understands nor directly controls. The mechanic who repairs an engine and the director who orchestrates a film both exercise agency, but the forms are categorically different. The mechanic authors the repair: her hands, her judgment, her sensory engagement at every stage determined the outcome. The director specifies a vision that others embody. Both are legitimate. But the experiential quality and cognitive yield of the two forms of agency differ in ways that matter profoundly for what kind of practitioner each produces.
The Romantic Craftsman's Nostalgia — Contrarian ^ Opus
There is a parallel reading that begins from the material conditions of expertise rather than its phenomenology. Crawford's distinction assumes authorship has always been available as a formative practice — that the mechanic who repairs an engine has chosen that path over direction, that both modes have been equally accessible. But most human work has never been authorship in Crawford's sense. The factory worker, the data entry clerk, the line cook following recipes — these roles involve execution without the cognitive autonomy Crawford celebrates. They are neither authorship (too constrained) nor direction (no evaluative power). The vast majority of workers have been subjected to systems they neither author nor direct, performing tasks whose resistance teaches nothing because the tasks permit no interpretation.
Crawford's framework also assumes the authorship that matters is individual rather than collective. The software engineer who writes code is not actually authoring in isolation — she inherits frameworks, libraries, patterns developed by thousands of others. Her 'struggle with the material' is already mediated by layers of abstraction she did not create and does not fully understand. The romantic image of the solitary craftsman encountering raw resistance obscures how expertise has always been socially distributed, how understanding has always been partial. If AI shifts us further toward direction, it may be making visible what was already true: that most work, most of the time, has been specifying desired outcomes within systems we do not control, hoping the mechanisms we only partially comprehend will yield something close to what we imagined.
In the AI Story
AI-mediated production transforms the practitioner from author to director. The engineer who describes a system to an AI and receives an implementation has exercised direction, not authorship. She specified the desired behavior. She evaluated the output. She made architectural decisions about how components should relate. These are genuine cognitive acts — demanding, in some cases, more sophisticated judgment than the implementation itself would have required. But the author-practitioner undergoes the work: she encounters the material's resistance, feels the specific frustration of a function that does not behave as expected, experiences the satisfaction of understanding why. The director-practitioner oversees the work. She evaluates results. She does not undergo the formative process that builds the embodied judgment distinguishing competent direction from brilliant direction.
Crawford's analysis connects to Edo Segal's observation in The Orange Pill about the senior engineer who discovered his twenty percent — the judgment, the taste, the architectural instinct — was everything. Crawford would affirm the discovery and immediately ask where the twenty percent came from. It came from the eighty percent. It was deposited by decades of implementation work, thousands of hours of hands-on engagement with systems that broke in ways that forced revision of his understanding. If the eighty percent is automated for the next generation, the twenty percent is not transmitted — not because it is magical but because it is deposited by a specific process that the automation has eliminated.
The concept has existential as well as cognitive weight. Crawford's 2024 essay AI as Self-Erasure argued that outsourcing cognitive work to AI is a form of voluntary self-absence — a choice not to show up for tasks through which identity is formed and expressed. The father who uses ChatGPT to write his daughter's wedding toast has absented himself from the occasion. The toast may be competent. It may even be moving. But it is not his. It does not bear the mark of his struggle with language, his specific love for his specific daughter. The replacement is not merely of output but of the presence that producing the output would have manifested.
The practical implication is not that AI should be rejected but that the choice between authorship and direction should be made deliberately — with full awareness of what each mode costs and yields. Some tasks are appropriately delegated. Some tasks — the ones through which identity is formed, expertise developed, relationships maintained — require authorship. The practitioner who cannot distinguish the two, or who has never experienced authorship in domains important to her, loses the capacity to make the distinction meaningfully. The discipline of preserving authorship where it matters is, in the AI age, a discipline that must be practiced consciously rather than assumed.
Origin
Crawford developed the concept across his corpus, with particular depth in Shop Class as Soulcraft and Why We Drive. The 2024 essay AI as Self-Erasure extends the framework explicitly to AI-mediated production.
Key Ideas
Authorship as undergoing. The author-practitioner is shaped by the specific process of execution — encounters with resistance that deposit embodied understanding.
Direction as overseeing. The director-practitioner evaluates outputs without undergoing the formative process that produces the judgment the evaluation requires.
Identity through struggle. Certain tasks are constitutive of self-formation; outsourcing them is self-erasure rather than efficiency.
The twenty-percent question. The judgment that remains when implementation is automated was deposited by the implementation — it cannot be inherited without the process that produces it.
Deliberate delegation. The choice between authorship and direction must be made consciously in each domain, with attention to what each mode develops in the practitioner.
Appears in the Orange Pill Cycle
Authorship as Threshold, Not Binary — Arbitrator ^ Opus
Crawford is fully right (100%) that the phenomenological difference between authoring and directing is real and formative — that hands-on struggle deposits understanding in ways evaluation alone cannot. The senior engineer's twenty percent did come from the eighty percent. No amount of architectural judgment emerges from pure direction. But the contrarian view correctly identifies (70%) that Crawford's framework romanticizes a form of authorship most workers have never accessed. The question is not whether to preserve authorship but for whom and in which domains it has ever been available.
The productive synthesis reframes authorship as a threshold rather than a binary state. What matters is whether practitioners cross the minimum threshold of hands-on engagement necessary to develop judgment in a domain — not whether they author every implementation. The surgeon who has performed hundreds of procedures can direct a surgical team effectively because she underwent enough resistance to deposit embodied understanding. The architect who has built models, struggled with materials, felt the difference between theoretical and practical constraints can direct construction. The threshold question is: how much authorship is required before direction becomes competent rather than hollow?
Crawford is right (90%) that AI risks eliminating the threshold-crossing work entirely — that if junior engineers never implement, they never develop the judgment to direct well. But the contrarian view holds (50%) that many roles have always been below-threshold: neither formative authorship nor autonomous direction, but execution within constraints. For those workers, AI's promise is not self-erasure but potential elevation — from constrained execution to genuine direction, if the transition is structured to include threshold authorship rather than bypass it entirely. The fight is not to preserve all authorship but to ensure pathways to threshold-crossing remain open.
Further reading
- Matthew B. Crawford, "AI as Self-Erasure," The New Atlantis (Summer 2024).
- Matthew B. Crawford, Why We Drive (William Morrow, 2020).