From research to interaction
This project is an interactive extension of my research on feminist content creation on Douyin, China’s largest short-video platform. The research behind the simulator draws on fourteen in-depth interviews with feminist content creators who described how they navigate visibility, audience engagement, platform governance, misogynistic backlash, and self-censorship while trying to make feminist ideas publicly accessible. In the study, creators consistently described themselves less as conventional influencers than as people trying to cultivate feminist knowledge, encourage reflection, and create space for public discussion.
The simulator translates those findings into a different form. Rather than asking visitors only to read about the pressures feminist creators face, it asks them to make decisions within a simplified version of that environment.
Over the course of one simulated month, the player manages one feminist content account and posts four videos. For each video, the player chooses a topic, decides how critically to frame it, and determines whether to alter the post before publishing.
The project is not designed to reproduce Douyin itself. It is designed to make the choices surrounding feminist visibility tangible.
Why a playable argument?
The central problem in this research is not simply censorship.
It is decision-making under uncertainty.
Creators rarely know with certainty what will happen to a video before it is posted. A topic may attract a large audience but also draw greater scrutiny. A structural feminist argument may communicate an idea more clearly while simultaneously becoming harder to circulate. A change intended to reduce moderation risk may preserve the substance of the argument, or it may gradually make the critique less direct.
Platform governance therefore operates partly through anticipation.
Creators learn to imagine possible reactions before those reactions occur. They revise scripts, alter wording, soften claims, avoid certain expressions, and develop circumvention practices in an effort to remain visible. My research argues that these forms of pre-publication self-censorship do more than reduce risk: over time, they can redirect feminist discourse away from explicit structural critique and toward safer, more individualized narratives.
A linear article can describe this process.
A playable form allows the visitor to encounter its logic.
The player must make a choice before seeing its consequence. The point is not to discover the correct answer, but to experience how apparently small decisions accumulate within an uncertain platform environment.
This is not a simulation of Douyin’s algorithm. It is a simulation of the decisions creators make while trying to anticipate it.
What the simulation models
The simulator separates decisions at the level of an individual video from consequences that accumulate at the level of an account.
Each video begins with a topic.
Topics have different levels of Audience Pull and Scrutiny. Audience Pull represents the relative attention a topic is positioned to attract, while Scrutiny represents its baseline exposure to controversy, reporting, or moderation. These values are deliberately simplified. In practice, visibility depends on many additional factors, including timing, editing, account history, recommendation systems, audience composition, and current events.
After selecting a topic, the player chooses a perspective.
The perspective determines the video’s initial Criticality: how directly the argument names gendered structures, institutions, inequality, or power. A video about the same event can therefore remain at the level of individual experience, identify a broader social pattern, or explicitly interpret the issue through structural feminist critique.
The player can then edit the post before publishing.
Some strategies attempt to reduce governance exposure without substantially changing the argument, such as substituting coded language or phonetic spellings. Others alter the form of critique itself, for example by softening a claim or transforming a structural argument into a personal story.
The posted video is summarized through two video-level measures:
Expressed Criticality describes how much of the original structural critique remains in the final post.
Governance Exposure describes the relative degree to which that version of the post is exposed to moderation, reporting, or reduced visibility within the logic of the simulator.
Neither of these measures carries over to the next video. Every new post begins with a new topic and a new argument.
At the account level, two measures do accumulate:
Audience Momentum represents whether the account is building, maintaining, or losing audience growth across the month.
Account Health represents the account’s cumulative standing as repeated governance exposure creates pressure over time.
This distinction is intentional. A creator’s next video does not inherit the previous video’s criticality, but the account on which it is published does have a history.
Visibility is not the same as success
The simulator draws from a tension that repeatedly appeared in the interviews: creators need visibility in order to circulate feminist knowledge, but they do not necessarily value visibility for its own sake.
Participants often used trending news and gender-related public events to reach larger audiences. Trendjacking allowed them to enter conversations already receiving attention and redirect those conversations toward feminist interpretation.
At the same time, participants distinguished popularity from what we describe in the research as critical visibility. What mattered was not only whether a video was seen, but whether it could generate reflection, dialogue, and feminist understanding.
This is why the simulator does not treat a highly visible account as an automatically successful one.
A widely circulated video may contain very little structural critique. A highly critical video may receive little distribution. A safer post may protect the account while also changing what can be said.
The project is designed to keep those outcomes in tension.
Why criticality matters
One of the core ideas behind the simulator is that feminist critique exists at different levels of directness.
Saying that one woman experienced something unfair is not the same as naming a recurring gendered pattern. Naming a gendered pattern is not the same as identifying the institutional or structural conditions that reproduce it.
The Criticality scale is therefore not intended to rank opinions as morally better or worse. It describes how far an argument moves from individual experience toward structural explanation.
This distinction matters because one of the findings of the research concerns what becomes publicly legible under platform governance.
Creators described having to continually judge how much feminist knowledge an audience could absorb, how directly certain forms of critique could be articulated, and how much complexity could survive the combination of short-video conventions, political boundaries, hostile responses, and moderation uncertainty. The study conceptualizes this problem through topic density: the difficulty of making complex feminist knowledge understandable, survivable, and politically legible within a platform environment.
The simulator does not reproduce topic density as a separate score. Instead, parts of that tension are represented through topic choice, criticality, audience momentum, and governance exposure.
Editing is not neutral
The editing stage is central to the project because self-censorship is not treated here as a simple binary between “speaking” and “being silenced.”
Creators may still post.
The question is what form of the argument remains after they adapt it.
Some strategies preserve the underlying claim while altering its surface form. Replacing sensitive words with symbols or phonetic substitutes, for example, is modeled as a circumvention strategy: the argument remains largely intact even as its presentation changes.
Other strategies modify the discourse itself.
When a structural claim becomes a personal anecdote, or when an explicit criticism becomes a heavily qualified opinion, the post may become easier to circulate—but something analytical may also be lost.
This reflects the research finding that self-censorship can function as a depoliticizing mechanism. Rather than merely hiding forbidden words, it can gradually redirect feminist critique away from patriarchy, institutional inequality, or structural power and toward individualized moral narratives.
The editing stage therefore asks the player to notice not only whether a strategy reduces risk, but what happens to the argument in the process.
How to read the numbers
The simulator uses five-level ordinal scales.
These numbers are not empirical measurements of Douyin’s algorithm, and they should not be interpreted as real probabilities of reach, moderation, deletion, or account restriction.
A five-level scale was chosen deliberately because the underlying research is qualitative. Using precise percentages would imply a degree of predictive certainty that the data cannot support.
The scores instead function as interpretive devices.
They make relationships described in the research visible enough to play with:
- some topics naturally attract more public attention than others;
- some topics are perceived as more sensitive;
- some perspectives make structural critique more explicit;
- some editing strategies preserve criticality while reducing exposure;
- other strategies reduce exposure partly by making the critique less direct.
The relative values were developed from patterns in the research and then discussed with four feminist content creators who reviewed the project and helped calibrate scenarios, trade-offs, and scoring logic. Their input does not make the simulator a predictive model, but it helps ensure that its internal relationships remain recognizable to people familiar with this kind of content production.
The simulated enforcement outcomes are likewise illustrative rather than predictive. Higher Governance Exposure makes intervention more likely and potentially more severe, but no score is intended to determine exactly what Douyin would do.
Opacity and unpredictability are part of the phenomenon being represented.
Audience momentum and account health
The account-level metrics are designed to make the four posts feel connected.
Audience Momentum represents the account’s ability to build an audience over the course of the month. Topics with stronger audience appeal have more growth potential, particularly when the argument remains relatively accessible. Highly critical content can still circulate, but in this simplified model it does not automatically translate into audience growth. Reduced distribution can prevent potential growth from being realized, while removal can interrupt momentum altogether.
This is not a claim that audiences reject feminist critique.
Rather, it translates a recurring problem described in the research: creators work within a short-video attention economy in which trending, accessible content can provide entry points to larger publics, while more complex structural analysis often requires greater interpretive work from viewers.
Account Health represents another form of accumulation.
One high-exposure video does not necessarily damage an account immediately. Repeated exposure, however, can create increasing precarity. The simulation therefore tracks cumulative account pressure across the four videos.
Together, Audience Momentum and Account Health produce two different trajectories:
one concerns whether the creator is building an audience;
the other concerns whether the account remains viable enough to keep speaking.
There is no winning strategy
The simulator intentionally does not contain one optimal path.
A player could finish the month with a healthy account and strong audience momentum while having repeatedly softened structural critique.
Another player might maintain high criticality but build little audience.
Another might make widely appealing posts that encounter reduced distribution or account pressure.
These are not intended as conventional success and failure states.
They represent different arrangements of the same underlying trade-offs.
The project therefore resists reducing feminist platform practice to a strategy game in which the goal is simply to maximize reach while minimizing moderation.
That would reproduce the very visibility logic the research questions.
Instead, the simulator asks:
What kind of feminist knowledge becomes visible?
What must be altered for it to remain visible?
And what happens when the conditions of visibility begin shaping the content of critique itself?
What the project cannot simulate
No interactive model can reproduce the actual operation of Douyin.
Platform recommendation and governance systems are opaque, dynamic, and account-specific. Outcomes may depend on factors unavailable to researchers or creators, including changing moderation rules, user reports, account history, recommendation patterns, political context, and internal platform decisions.
The simulator also compresses a much longer process of creator learning into four videos.
Real creators develop strategies through repeated experimentation, failed posts, informal knowledge-sharing, observation of other accounts, and experience accumulated over months or years.
The model therefore simplifies reality in order to make a conceptual argument visible.
Its purpose is not prediction.
Its purpose is interpretation.
Research basis
This project grows from:
Wang, Guan, and Shaheen Kanthawala. (2026). “Douyin is taming us”: Visibility, governance, and feminist discourses on Douyin. New Media & Society.
The study draws on fourteen in-depth interviews with feminist content creators and examines how they negotiate platform visibility, feminist knowledge production, audience engagement, platformed misogyny, topic density, and anticipatory self-censorship.
The interactive project extends that research through research-creation: it translates relationships identified in the qualitative findings into a set of choices, constraints, and consequences that visitors can navigate for themselves.