The Proximity Paradox: Why AI Writing Feels Like a Tool, but AI Acting Feels Like a Threat

Generative AI does not enter every creative field in the same form. In writing, people can use language models privately and incrementally—for brainstorming, outlining, revision, or substitution. A study of young people’s writing practices found uses ranging from routine assistance to catalysing thought, alongside concern about voice, dependence, and what still counts as “real” writing (Higgs et al.). In screen performance, a digital replica can instead appear as a visibly separable version of a particular worker’s face, body, or voice. James Earl Jones’s licensed synthetic Darth Vader voice is one prominent example (Vincent).

This essay proposes the Proximity Paradox as an interpretive lens: resistance to synthetic creativity may intensify when the technology appears closer to a worker’s embodied identity, makes replacement easier to see, or reduces that worker’s control. This is a hypothesis for comparing cases, not an experimentally established law about audiences or creative labour.

Three dimensions of proximity

1. Embodiment

Writing reaches readers through recorded language, so the writer’s physical presence is not normally part of the delivered work. That does not make writing disembodied in its creation: experience, health, place, and material conditions all shape prose. It does mean that the published artefact can circulate separately from the writer’s face and voice.

Acting foregrounds a performer’s body, movement, expression, and voice. A digital replica therefore operates on attributes closely tied to both identity and employability. Masahiro Mori’s “uncanny valley” describes discomfort with representations that approach, but do not achieve, human likeness (Mori, MacDorman, and Kageki). That theory may help explain some aesthetic reactions to synthetic performers, but it does not by itself explain labour opposition, consent, or compensation.

2. Visible substitution

The economic effects of AI in writing are real but difficult to reduce to one pattern. Reports of AI-generated “book spam” and bogus books on Amazon show one visible form of market saturation (Tapper). Other effects occur upstream—in pitching, commissioning, translation, copywriting, or editorial budgets—and require different evidence. The earlier version of this essay claimed that AI rarely displaced an identifiable writer from a prestigious commission; the available sources did not support that claim, so it has been removed.

For performers, the substitution can be easier to identify because a replica may be created from a named person and reused in place of new recorded work. AI was one of several central issues in the 2023 SAG-AFTRA negotiations. The resulting contract requires informed consent for creating and using digital replicas and, in most cases, compensation and residuals; it also prevents producers from using a replica simply to avoid hiring a background performer (SAG-AFTRA). Those provisions offer stronger evidence than treating a pre-agreement studio proposal as the settled state of the industry.

3. Control and authenticity

“Authenticity” is too broad to function as a measurable explanation on its own. Readers and viewers care about provenance differently across genres, communities, and contexts. A more precise question is whether the person whose work or identity is being simulated has meaningful knowledge, consent, compensation, and control.

That shift also clarifies the debate for writers. The Authors Guild’s 2023 open letter asked AI companies to obtain permission, credit writers, and compensate them for copyrighted material used in generative systems (Authors Guild). The objection is not simply that machine-made language lacks a “soul.” It concerns who supplied the underlying work, who benefits, and who decides.

What the framework can—and cannot—do

The Proximity Paradox is most useful as a set of questions:

• Does the system reproduce a particular person’s identifiable voice, face, style, or prior work?
• Is it assisting that person, or enabling someone else to avoid employing them?
• Can the affected creator consent to each material use and negotiate compensation?
• Is substitution visible enough for audiences, workers, or regulators to contest it?
• What evidence would distinguish assistance, displacement, and market saturation in this field?

The framework cannot establish that the public generally accepts AI writing or rejects AI acting. It cannot predict individual responses, and it should not turn differences between crafts into a hierarchy of human value.

Conclusion

The contrast between AI writing tools and digital performers is not a clean divide. Both fields contain assistance, appropriation, consent, and replacement. The difference is often one of visibility and control: digital replicas can make a particular performer’s possible substitution unusually concrete, while changes to writing labour may be distributed across platforms and stages of production.

Nicholas Carr’s broader warning about automation remains useful here: tools can change not only efficiency but also the distribution of skill and judgment (Carr). The practical task is therefore not to declare one medium safe and another uniquely threatened. It is to make authorship, consent, compensation, and human responsibility legible wherever generative systems enter creative work.