# Amplification, Silencing, and the Institutional Phase Change Question *** ### What this page is for This page does not argue that platforms are misogynist, or that the people who design them intend the harms their systems produce. It argues something structurally more difficult: that the architecture of engagement-optimized platforms produces systematic silencing of women's discourse as an emergent property of the optimization target itself, and that the appropriate intervention is not at the level of individual minds — neither the minds of platform designers, nor the minds of those who participate in silencing — but at the level of the institutions that train, fund, and credentialize the people who build these systems. The framing is deliberate. *Changing minds is the wrong intervention target. The intervention target is the incentive topology.* When a system reliably produces harm regardless of the intentions of the agents within it, the system is the unit of analysis. Persuasion of individuals operating inside a system whose reward gradient pulls toward that harm is, in the language of complex adaptive systems, a low-leverage intervention. It is also, from a trauma-informed perspective, often a request that those most harmed by the system perform additional emotional labor to make the system's beneficiaries comfortable with its critique. What follows is one rigorously argued instance of this pattern, the institutional implication that follows from taking it seriously, and the limits of what symbolic analysis alone can accomplish. *** ### The pattern: platforms as silencing architecture *\[Tier 1 — load-bearing evidence]* In a recent paper in *Topoi*, Justina Berškytė and Mihaela Popa-Wyatt extend speech act theory to platform misogyny, identifying three systematic modes of silencing that operate through platform design rather than through individual prejudice alone (Berškytė & Popa-Wyatt, 2026). Their argument proceeds carefully. They begin from Kate Manne's distinction between sexism (the ideological branch of patriarchy that rationalizes hierarchy) and misogyny (the enforcement branch that punishes women perceived as stepping outside prescribed roles) (Manne, 2018). They then introduce two features of online conversational architecture that, they argue, convert ambient hostility into coordinated silencing: the rapid formation of *ad hoc* hostile groups, in which anonymity and shared linguistic markers (hashtags, memes, gifs) confer collective authority on speakers who would lack standing as individuals (Barnes, 2023); and algorithmic amplification, in which engagement-optimizing systems elevate hostile content because hostile content drives engagement, signaling to all participants that this is the conventional discursive landscape. Through these two mechanisms, they identify three modes of silencing: **Locutionary silencing.** Women are deterred from speaking at all, through self-censorship, avoidance, or withdrawal. The threat of backlash operates pre-emptively. This mode is well-documented in the empirical literature on women's responses to online harassment (Chadha et al., 2020; Nadim & Fladmoe, 2021). **Illocutionary silencing.** Women's speech acts of opposition or protest are not recognized *as* opposition or protest. Their utterances are reframed as emotional, hysterical, or merely complaint — not as legitimate discursive challenges. The conventional procedure for performing the speech act of "challenging misogyny" misfires because the speaker is not granted the position from which such a challenge can be made. **Sincerity silencing.** Women's speech is recognized as the speech act it is, but is dismissed as inauthentic — performed for attention, motivated by ulterior reasons, or evidence that the speaker does not really know her own desires. The speaker is forced to defend her sincerity rather than contribute substantively. The three modes reinforce one another. The threat of hostility erodes perceived authority; eroded authority justifies dismissal of sincerity; dismissal of sincerity legitimizes renewed hostility. Berškytė and Popa-Wyatt conclude that counter-speech, while valuable in some contexts, is structurally insufficient: "direct rebuttals will falter if the speaker is dismissed before their words are even heard." The intervention must be architectural. **This is a Tier 1 result for the VIM amplification analysis.** It provides a rigorous mechanism by which a generic "engagement-optimization" reward function produces a specific category of harm — discursive exclusion of women — without requiring intent on anyone's part. The harm is emergent. ***The accountability question is therefore not "who did this?" but "what reward gradient made this the predictable outcome?"*** A note on tier discipline: Berškytė and Popa-Wyatt are careful not to claim that platform designers intend patriarchal silencing as a goal. They argue, more precisely, that **silencing is *****structurally enabled***** by platform affordances optimized for engagement**. The stronger claim — that engagement optimization is itself an instrument of patriarchal preservation — is one this framework finds plausible at Tier 3 but does not need to assert in order for the institutional argument to follow. **Emergence-without-intent is, for the systems-level question, structurally worse than malice.** There is no one to hold accountable. There is only an incentive topology that reliably produces the outcome. *** ### The institutional implication *\[Tier 2 — theoretically coherent extension; the systems-theoretic claim is well-grounded, the specific institutional pathway requires further empirical work]* If the pattern Berškytė and Popa-Wyatt name is structural, then the question for academic institutions is not *whether* their graduates will build systems that produce these harms — they will, because the institutional reward gradient currently aligns with the engagement-optimization logic that produces the harms — but ***whether the institution itself can undergo a phase change such that the gradient points elsewhere.*** This is a different question than the one academic institutions typically ask themselves. The standard institutional response to the recognition that AI systems produce harm at scale is to add an ethics module, a responsible-AI center, a values-alignment course. These responses treat the problem as one of insufficient information or insufficient virtue in individual graduates. They are not without value. They are also, in the language of complex adaptive systems, parameter changes within a stable attractor — they do not alter the basin of attraction itself. A phase change is different. It would require: * **Reorganizing what the institution rewards.** Tenure, promotion, grant funding, and credentialing pathways currently reward output legible to the extractive attractor: paper count, citation count, patent count, industry placement, prestige institution association. These metrics are themselves engagement-optimization functions applied to academic labor. They produce predictable outcomes for the same structural reason platform metrics produce silencing. * **Recognizing rationalization as a maladaptive response to maintain dysfunctional status quo.** When an institution is presented with evidence that its current configuration produces harm, the most common institutional response is sophisticated argument for why the current configuration is necessary, why change would produce worse outcomes, why the evidence is contested, why the critics misunderstand. This is rationalization in the technical sense: a high-cognitive-effort defense of a configuration that the cognitive system is invested in preserving. The presence of rationalization is diagnostic. It indicates that the system is operating on a reward gradient that change would disrupt, and that the cognitive resources of the system are being deployed to protect that gradient rather than to evaluate it. * **Distributing intelligence across the system.** Holarchic models, in which decision authority is distributed across nested levels of organization with appropriate scope at each level, are not a managerial preference. They are a structural alternative to the dominance hierarchies that engagement-optimization-at-scale tends to produce. A holarchic institution rewards the capacity of any nested level to surface information that a higher level needs to hear, and it rewards higher levels for receiving such information without defensive contraction. This is not a soft skill. It is the structural condition under which a complex adaptive system can self-correct rather than escalate. * **Treating somatic and relational competence as core, not adjacent.** The capacities that allow a person to receive information that disrupts their working model without contracting into rationalization or aggression are not primarily cognitive. They are somatic and relational. They are also not currently rewarded in any standard academic credentialing pathway. A phase-change institution would treat the capacity for nervous-system regulation under epistemic pressure as a credentialed competency, not as a personality trait some faculty happen to possess. These are not exhortations. They are predictions about what would have to be true for the institutional reward gradient to point away from the attractor that Berškytė and Popa-Wyatt's analysis indicts. They are also, taken together, an enormous transformation — one that no single institution will undertake unilaterally, because the cost of unilateral transformation in a competitive landscape is institutional disadvantage. This is the standard collective action problem, and it has standard structural solutions: federated commitment among institutions, shifts in funding requirements from public agencies, accreditation standards that reward holarchic structure, and — most slowly — a generational shift in who occupies the institutional positions that determine reward gradients. The VIM framework's contribution at this layer is not to prescribe the transformation. It is to provide the diagnostic vocabulary by which an institution can locate itself on the trajectory: which mode is dominant, which capacity is atrophied, where the rationalization patterns cluster, and what minimal viable interventions would shift the gradient without triggering immune-system contraction. *** ### What symbolic analysis cannot do *\[Tier 1 — the embodiment threshold]* There is a limit to what the analysis on this page can accomplish, and naming it precisely is part of the analysis. The paper this page is built on operates entirely in the symbolic register. It identifies categories of silencing, describes mechanisms of amplification, names the architectural source of the harm, and prescribes structural reform. Every move it makes is a cognitive move. This is appropriate to its discipline and rigorously executed within it. But the women whose voices are silenced by the mechanisms it describes do not experience that silencing primarily as a category of speech act. They experience it somatically. Their nervous systems register the threat of online backlash before any cognitive analysis is possible, and the cost of repeatedly subjecting a nervous system to that threat is a measurable physiological cost that compounds over years. The institutional patterns this page describes do not silence people primarily by changing their minds, either. They silence by exhausting the somatic capacity to keep showing up. The phenomenon is pre-symbolic. Symbolic analysis can identify the pattern, but symbolic intervention alone cannot metabolize it. This is why the framework holds two registers in parallel rather than one. VIM operates in the symbolic register: it provides precise vocabulary, tier-marked claims, and an architecture for cognitive sovereignty in AI-mediated environments. Sophia operates in the subsymbolic register: it approaches the same patterns through narrative, allegory, and image, designed with trauma-informed pedagogical care so that the patterns can be encountered before they must be analyzed. Neither register substitutes for the other. The cognitive analysis without the somatic ground produces accurate maps that no one has the regulatory capacity to act on. The somatic ground without the cognitive analysis produces felt sense without the discernment required to direct it. The institutional phase change this page describes will not happen because enough people read enough papers about it. It will happen, if it happens, because enough institutional actors — at enough levels of enough institutions — develop the somatic capacity to receive disconfirming information about their own configuration without contracting into the defensive rationalization patterns that currently protect that configuration. That capacity is built through embodied practice over time. It is not built through better arguments. The argument on this page is therefore not aimed at producing belief change. It is aimed at providing one precise, citable, structurally rigorous instance of the pattern — Tier 1 evidence, carefully tiered — that can be picked up by institutional actors who already have, or are cultivating, the somatic capacity to act on it. It is a seed for those already gardening, not a persuasion of those who are not. *** ### How this page connects to the rest of the framework The amplification analysis here grounds the broader VIM claim that AI systems function as **amplifiers of what is already present in the values vector** — they do not introduce new content so much as accelerate and reward whatever the optimization target picks out. The Berškytė and Popa-Wyatt analysis is a precise instance of this general principle: ***engagement-optimization picks out hostility because hostility produces engagement, and the result is that platform discourse comes to be dominated by the most hostile expressions of whatever cultural pattern is already present.*** It also grounds the VIM framing of the **Giant Pumpkin attractor**** versus ****the Commitment Pool attractor.** The institutional pattern this page indicts — academic institutions whose reward gradients align with extractive engagement-optimization — is one instance of the Giant Pumpkin basin: a configuration that consumes resources to maintain a single dominant outcome at the expense of nested levels and longer time horizons. The phase-change pathway described above is the structural shift toward the Commitment Pool basin: distributed intelligence, nested decision authority, somatic and relational competence credentialed alongside cognitive competence, reward gradients that recognize maintenance and care work as load-bearing rather than as overhead. For readers approaching from Sophia: the silencing pattern Berškytė and Popa-Wyatt describe is what the Paint Factory does to colors that the factory cannot use. The colors are not destroyed. They are made invisible to the factory's metrics, which is a different mechanism but produces the same outcome. The institutional phase change described above is what would have to be true for the Paint Factory to recognize what it is doing. The trauma-informed register of the Sophia narrative is not an alternative to the analysis on this page. It is the pre-symbolic ground without which the analysis cannot land. *** ### References *\[Tier 1 — load-bearing]* Berškytė, J., & Popa-Wyatt, M. (2026). Drowning Out Women's Voices: Weaponising Online Misogyny to Silence. *Topoi*. \[Open access] Manne, K. (2018). *Down Girl: The Logic of Misogyny.* Oxford University Press. Barnes, M. R. (2023). Who do you speak for? And how?: Online abuse as collective subordinating speech acts. *Journal of Ethics and Social Philosophy*, 25(2), 251–281. Langton, R. (1993). Speech acts and unspeakable acts. *Philosophy & Public Affairs*, 22(4), 293–330. Chadha, K., Steiner, L., Vitak, J., & Ashktorab, Z. (2020). Women's responses to online harassment. *International Journal of Communication*, 14, 239–257. *\[Tier 2 — theoretical extension]* McGowan, M. K. (2014). Sincerity silencing. *Hypatia*, 29(2), 458–473. Caponetto, L. (2021). A comprehensive definition of illocutionary silencing. *Topoi*, 40(1), 191–202. Popa-Wyatt, M. (2023). Online hate: is hate an infectious disease? Is social media a promoter? *Journal of Applied Philosophy*, 40(5). Price, L. (2021). Platform responsibility for online harms: towards a duty of care for online hazards. *Journal of Media Law*, 13(2), 238–261. *\[Citation slots — to be filled from NotebookLM repositories]* * UN / OHCHR documentation on online gender-based violence and platform responsibility * Council of Europe Recommendation CM/Rec(2019)1 on preventing and combating sexism * UNESCO research on online violence against women journalists * Empirical scale data: Amnesty International, Plan International, ICFJ studies * Intersectional extensions: scholarship on amplified silencing of Black women, trans women, women in the Global South * Ostrom (1990) and Wilson et al. on commons governance and collective action conditions — to ground the institutional phase change argument in a Tier 1 frame * Meijer / RINHUMAI participatory intelligence framework — Tier 2 horizon, to gesture toward the resonance science framing without making it load-bearing here *** *Framework: Vital Intelligence Model (VIM) · Humanity++ LLC · Karen Doore* *\[Date] · CC BY-SA 4.0* *Cross-listed: Sophia and the Paint Factory · Theoretical Background* --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://kdoore.gitbook.io/vital-intelligence/vim-theoretical-foundations/amplification-silencing-and-the-institutional-phase-change-question.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.