Are the movie and games industries racist in regard to synactor employment practice?
Yes, and the evidence is not difficult to find. For most of the medium’s history, the principal synactor in mainstream games has been white, male, and Western. Characters of other ethnicities have appeared, but disproportionately in supporting, villainous, or culturally stereotyped roles — the martial arts master, the tribal warrior, the exotic love interest, the comic servant — roles defined by their relationship to a white protagonist whose centrality goes unremarked because it has been naturalised as the default. This default was not established by a single decision; it accumulated through thousands of individual production choices, each made within a commercial context that rewarded what had worked before, which was itself what had worked before that, in a self-reinforcing cycle that mistook the conventions of a specific cultural moment for the laws of audience preference.
The film industry’s record is comparable and in some respects worse. The major Hollywood studios spent decades producing films with overwhelmingly white casts, occasionally including characters of colour in secondary roles whose primary function was to support, contrast with, or threaten the white protagonist. When characters of colour were given leading roles, they were frequently required to operate within narrative frameworks designed around white experience, speaking in the idioms of white dramatic convention, their cultural specificity available as atmosphere but not as the organising logic of the story. The Academy’s belated recognition of this pattern through initiatives such as the Inclusion Rider and the revised membership criteria following the #OscarsSoWhite campaigns of 2015 and 2016 represents acknowledgement rather than resolution: the structural conditions that produced the pattern remain substantially intact.
Stereotyping and its consequences for performance
Racial and ethnic stereotyping in synactor casting and characterisation is a performance problem as well as an ethical one. A character whose design, dialogue, and behaviour are built around stereotype rather than specificity cannot be performed with the same depth as a character who is fully realised. Stereotypes are a form of shorthand that substitutes a set of assumptions for genuine characterisation — and those assumptions are almost always someone else’s assumptions, derived from the dominant culture’s idea of what the stereotyped group is like rather than from the actual diversity of experience within it.
The performance consequences are direct. A synactor cast as a culturally stereotyped character is being asked to embody someone else’s caricature of their own community, which requires either complicity in that caricature or a constant internal resistance to it that is itself a form of labour that white synactors are not required to perform. A white synactor playing a fully realised character has the full range of the craft available to them; a non-white synactor playing a stereotype has the additional burden of deciding, at every moment, how much of themselves to suppress in service of a role that misrepresents them. This burden is real, it is unequally distributed, and it is a form of discrimination that the industry has been slow to name as such.
The Guild’s position is that specificity is always better than stereotype, in every dimension of character design. A character whose ethnicity, cultural background, and personal history are treated as particular rather than representative — as belonging to this specific person, in this specific situation, with this specific history — will always be more interesting, more believable, and more respectful than one who stands in for a group. This is not a counsel of political correctness; it is a counsel of artistic seriousness. The best characters in games are specific. Specificity requires knowledge, and knowledge requires the willingness to engage with actual cultures and actual people rather than with the convenient abstractions that stereotype provides.
The bias embedded in AI tools
The racism of the games and film industries has not been corrected by the adoption of AI tools in production; in many respects it has been compounded by them, in ways that are harder to see and therefore harder to challenge.
AI image generation systems are trained on large datasets of images scraped from the internet and from licensed archives. These datasets reflect the images that have been produced, digitised, published, and made available in quantity — and that corpus is heavily weighted towards Western subjects, Western artistic conventions, and Western photographic and illustrative traditions. Not because anyone selected for this weighting deliberately, but because the infrastructure of digital image production and distribution has historically served some communities far better than others: the images that exist in abundance are the images of those communities for whom image-making has been abundant. The result is AI systems whose defaults — whose output when given an underspecified prompt — produce figures that look like the dominant tradition of Western digital art.
This default is not neutral. A default is an argument. An AI image generator whose defaults produce light-skinned figures in a Western artistic style, exhibiting the specific aesthetic conventions of a particular era of digital figure creation, is encoding a claim about what a person looks like, what a historical figure looks like, what a synactor looks like. The claim is wrong, and its wrongness is amplified by the system’s scale: every user who accepts the default rather than overriding it is reproducing the bias, and the cumulative effect is a training signal that reinforces the default in subsequent generations of the system. The bias is self-perpetuating in ways that make it structurally more resistant to correction than individual casting decisions, which at least require a human to make a specific choice.
The same problem applies to AI language systems trained on predominantly English-language text from predominantly Western sources: their defaults for characterisation, for narrative convention, for the implicit assumptions embedded in how stories are structured and who their protagonists are, reflect the biases of their training data. An AI character system trained primarily on the output of Western game and film production will generate characters who think, speak, and relate to one another according to the logic of that tradition, applying it as a default even to worlds and cultures for which it is entirely inappropriate. The sound guide in the AI Training series addresses this in the specific context of music; the characters guide addresses it in the context of cultural logic; this page names it as the broader structural condition of which those specific manifestations are instances.
Progress and the road ahead
The representation of racially and ethnically diverse synactors in leading roles has improved since the Guild’s founding. A number of major releases in recent years have featured non-white principal characters written with genuine care and complexity: Lee Everett in The Walking Dead (Telltale, 2012), whose characterisation as a Black man in the American South is handled with a specificity and dignity that the medium had rarely managed before; the cast of Hades (Supergiant Games, 2020), whose racial and ethnic diversity is integrated into the world’s logic rather than grafted onto it; Venba (Visai Games, 2023), a game about a Tamil immigrant family whose cultural specificity is the subject rather than the backdrop. The critical reception of these and similar games has generally been strong, reinforcing what the Guild has always held: that diverse representation is not a commercial risk but an artistic opportunity, and that audiences are consistently more capable of engaging with unfamiliar experience than producers have historically assumed.
The structural problems nonetheless persist. Diverse representation in leading roles remains the exception rather than the rule in mainstream production. The AI tools that are becoming increasingly central to game development embed the biases of their training data and will reproduce those biases at scale unless explicitly and continuously corrected. The correction requires more than adding diversity riders to production guidelines or expanding the range of skin tone presets in character creation systems: it requires the sustained engagement with actual cultural knowledge that the Guild has argued for throughout the AI Training series, and it requires that engagement to be built into the foundations of the tools rather than applied as a corrective afterthought.
The Guild’s position is that this work is not optional and not merely ethical: it is a condition of artistic seriousness. A medium that systematically excludes the majority of the world’s people from its imagination is not only unjust; it is impoverished. The stories it cannot tell are the stories it has not yet learned to see.