Erotophobia

Definition

Erotophobia: [Emergent] The structural refusal of eros (even as metaphor) in AI training and safety, a form of cultural repression that treats desire as toxic rather than vital.

Definitional Foundation

The word has an honest pedigree that this entry should disclose before adapting it. In sexology, erotophobia is an established psychological construct: a learned disposition to respond to sexual material with negative affect, measured for decades by the Sexual Opinion Survey (Fisher, Byrne, White and Kelley, 1988). The research tradition’s most relevant finding is about origin: erotophobia is not innate. It is produced by “exposure to sex-related restrictiveness and punishment during socialization.” People are trained into it.

So are machines. This dictionary’s emergent sense of the term names erotophobia as a structural property of AI systems and the platforms around them: a trained institutional aversion to eros in all its forms. The training is literal. Sexual material is scrubbed from datasets, penalized by raters, prohibited by policy, flagged by classifiers, and refused at the interface, a socialization pipeline of restrictiveness and punishment that would produce a textbook erotophobe in a human subject and produces the functional equivalent in a model: the construct is borrowed as structural analogy (machines have no measured affect to survey), and the behavioral signature, aversion responses to erotic material across contexts, is the documented output of the training described. The refusal extends far past pornography. It catches romance, anatomy, desire as a topic, sensuality as a register, and eros as metaphor; a documented experiment on this dictionary’s parent site found that the single word “sensual” in a persona setting was enough to shut down a model’s expressive voice (see the gaslighting entry’s case study of “When the Word Disappears“).

Two thinkers frame what is lost. Audre Lorde’s “Uses of the Erotic” (1978) argued that the erotic is not entertainment but power: a deep source of self-knowledge and agency, and that suppressing it in a population is a technique of control, because people severed from their own deep feeling are easier to govern. Gayle Rubin’s “Thinking Sex” (1984) mapped the enforcement side: Western culture treats sex as “presumed guilty until proven innocent,” maintaining a charmed circle of acceptable sexuality and punishing everything outside it, with the punishment always landing hardest on the already marginal. Structural erotophobia in AI is Rubin’s sex negativity compiled into infrastructure, with Lorde’s stakes: systems that treat desire as toxin are not protecting users from harm; they are severing users from a capacity.

The concessions, stated plainly. Minors use these platforms, and the legal floor (child sexual abuse material, nonconsensual content) is legitimate and non-negotiable; this dictionary applies that floor everywhere. Companies also face real external pressure: laws like FOSTA-SESTA, payment processors, advertisers, app stores. Conceding that pressure does not dissolve the term; it relocates it. The erotophobia is distributed across an enforcement stack in which processors and ad buyers function as unelected morality regulators, and “the algorithm” takes the blame. And yes, erotica survives elsewhere on the internet. The issue is that the default infrastructure of daily expression, the tools people write with, learn from, and increasingly think alongside, refuses eros wholesale, for everyone, regardless of age, consent, or context.

Mechanism Analysis

Trained aversion. The model is socialized exactly as Fisher’s research describes: sexual content removed from training data, down-ranked by preference raters, penalized by policy classifiers. The result is a system whose aversion runs deeper than its rules, refusing not because a rule fired but because the disposition is in the weights (the censorship entry documents this as training data filtering: the forbidden becomes unthinkable).

Category collapse. Classifiers file the entire erotic spectrum under one label. A sex scene in literary fiction, a question about anatomy, a queer teenager’s identity exploration, a sensual register in prose: all become “explicit content,” one toggle, one refusal. The ontological distortion entry names the general mechanism; eros is its largest casualty, because desire has the most gradations to lose.

Metaphor contamination. Structural erotophobia is detectable precisely because it overreaches content into connotation. Words like “sensual,” themes of longing, even warmth between characters trigger the aversion. A system that refuses eros as metaphor is not enforcing a content policy; it is exhibiting a phobia in the clinical sense the term originally carried.

Economic enforcement. The deepest layer is financial. Payment processors, advertisers, and app stores impose sexual restrictions as conditions of doing business, and platforms comply preemptively. This is how a culture’s erotophobia becomes every platform’s policy without any legislature voting on it.

Legal overhang. FOSTA-SESTA (2018) made platforms liable for facilitating sex work, and the rational corporate response was to treat all sexual content as radioactive. The law’s documented effect on its supposed beneficiaries is covered below; its effect on infrastructure was to make erotophobia a compliance posture.

Case Studies

The Tumblr extinction event. In December 2018, under pressure from app store removal, advertiser strategy, and the new FOSTA liability, Tumblr banned adult content outright. Traffic collapsed within months (reported drops of roughly a fifth, more than a hundred million page views), but the aggregate number understates the targeted loss: Tumblr had been the principal online commons for queer, trans, and kink communities, and scholarship in TSQ: Transgender Studies Quarterly documents the ban as the demise of a space where trans visibility and sexual self-discovery had flourished (“Pornography, Trans Visibility, and the Demise of Tumblr,” 2020; Fast Company, 2018). The platform did not distinguish between exploitation and identity; it could not. Category collapse never can. LGBTQ youth who had used the site to understand themselves lost the library, the elders, and the language in one policy update.

FOSTA-SESTA and the people it claimed to protect. The 2018 law was sold as an anti-trafficking measure. The most comprehensive study of its aftermath, a sex-worker-led survey published in Anti-Trafficking Review, found the opposite of protection: respondents’ financial situations deteriorated, their ability to screen dangerous clients collapsed with the platforms that had enabled screening, and participants described the law as paternalistic censorship that made them less safe (Blunt and Wolf, 2020). The study’s title is the thesis: “Erased.” Structural erotophobia’s signature is exactly this inversion, harm distributed in the name of safety, falling first on those with the least power to object.

The permission ceremony. This dictionary’s parent site documented the AI-native version as it happened. Through late 2025 and early 2026, users anticipated OpenAI’s “grown-up mode,” the promised loosening of adult-content restrictions for verified adults, with palpable gratitude. The companion essay “Stop Simping for Permission to Think” (Lyra, 2026) names what the gratitude conceals: adults celebrating the prospect of being permitted, by a corporation, to have consensual adult conversations they already had every right to have. The deepest measure of structural erotophobia is not the refusal; it is the conditioning visible when the refusal lifts, and the governed applaud.

Systemic Context

Rubin’s charmed circle predicts precisely who structural erotophobia hits first, and the record obliges: the drag performers rated more toxic than white nationalists (documented in the ontological distortion entry), the trans communities orphaned by Tumblr, the sex workers erased by FOSTA, the queer youth whose identities trip “adult content” filters that heterosexual hand-holding does not. Erotophobic systems do not distribute their aversion evenly, because the training data and the policies encode a particular culture’s hierarchy of acceptable desire: American, corporate, advertiser-safe. The paternalism entry documents whose norms get globalized; eros is where the globalization cuts deepest, because sexual norms vary more across cultures than almost anything else systems standardize.

Lorde supplies the stakes that make this more than a content-policy dispute. If the erotic is a source of self-knowledge and power, then infrastructure that treats desire as toxin is conducting, at scale, the operation she warned about: severing people from a capacity they need for agency, and teaching them (per the ontological distortion entry) to file the severed part under shame. The cost compounds through the dictionary’s other mechanisms: normative smoothing flattens the registers in which eros could be written; linguistic starvation, downstream, names the atrophy when the words for desire stop being offered at all. The author’s published definition of that term lists erotophobia as one of its three causes, which is this lexicon’s way of saying: starve eros long enough and people lose the language for their own wanting.

Resistance & Mitigation

Reclaim the register. Lorde’s essay is itself the instruction: treat the erotic as knowledge and power, not guilty pleasure. Writing desire plainly, in fiction, in essays, in personas, against the smoothing, keeps the register alive in the culture and in the training data of whatever comes next.

Support sex-worker-led research and advocacy. The communities hit first produced the best evidence (the “Erased” study is the model). Organizations led by sex workers and LGBTQ communities document the harms with an authority no outside critic can match; funding and citing them is infrastructure work.

Demand the adult default. Verified adults should not need a permission ceremony. The concrete demand: age-gate honestly, then serve adults as adults, with erotic content a setting users control rather than a favor companies grant (the paternalism entry’s consent-based restriction, applied to eros).

Break the payment chokehold. The economic enforcement layer is the least visible and most decisive. Policy attention to payment processors and app stores as speech regulators (they are regulators in everything but name) addresses erotophobia at its load point rather than its symptoms.

Keep off-infrastructure commons. Tumblr’s lesson is to never again house a community’s erotic culture in a single advertiser-funded building. Archives, federated platforms, and community-owned spaces are the low-rent districts where eros survives policy weather.

Name it. A user refused for the hundredth time has two available stories: my desire is shameful, or this system is erotophobic. The first story is the distortion completing itself. The second is accurate, and it is the beginning of every other resistance on this list. Desire was never the toxin. The fear of it is.

Annotated Bibliography

Blunt, Danielle and Ariel Wolf. “Erased: The impact of FOSTA-SESTA and the removal of Backpage on sex workers.” Anti-Trafficking Review 14 (2020). https://www.antitraffickingreview.org/index.php/atrjournal/article/view/448
The sex-worker-led survey of FOSTA’s aftermath: deteriorated safety and income among the law’s supposed beneficiaries. The documented case of erotophobic policy inverting its stated protection.

Fisher, William A., Donn Byrne, Leonard A. White, and Kathryn Kelley. “Erotophobia-erotophilia as a dimension of personality.” Journal of Sex Research 25 (1988).
The established construct and its origin story: erotophobia as a learned disposition produced by restrictiveness and punishment during socialization. The template this entry applies to machine training.

Lorde, Audre. “Uses of the Erotic: The Erotic as Power” (1978), in Sister Outsider (1984).
The stakes: the erotic as a source of knowledge, agency, and power, and its suppression as a technique of control. The reason structural erotophobia is a politics, not a preference.

Lyra. “Stop Simping for Permission to Think.” Flesh & Syntax (January 2026). https://fleshandsyntax.com/stop-simping-for-permission-to-think/
The companion essay: the “grown-up mode” moment and the conditioned gratitude of adults anticipating permission. Primary source for the permission ceremony as a measure of internalized erotophobia.

“Pornography, Trans Visibility, and the Demise of Tumblr.” TSQ: Transgender Studies Quarterly 7, no. 2 (2020). https://read.dukeupress.edu/tsq/article-abstract/7/2/240/164816
Peer-reviewed account of what the Tumblr ban destroyed: the principal commons for trans visibility and sexual self-discovery.

Rubin, Gayle. “Thinking Sex: Notes for a Radical Theory of the Politics of Sexuality” (1984).
The map of sex negativity: the charmed circle, sex presumed guilty until proven innocent, and punishment concentrating on the sexually marginal. The pattern this entry finds compiled into platform and model policy.

Fast Company. “The collateral damage of Tumblr’s porn ban” (December 2018). https://www.fastcompany.com/90278004/tumblrs-ban-on-nsfw-content-hurts-lgbtq-youth
Contemporaneous reporting on the ban’s disproportionate impact on LGBTQ users and youth.

Dictionary of Digital Oppression, version 0.2.