The AIIDE-17 Workshop on Experimental AI in Games WS-17-19 Social Simulation for Social Justice Melanie Dickinson, Noah Wardrip-Fruin, Michael Mateas [email protected], {nwf, michaelm}@soe.ucsc.edu Expressive Intelligence Studio Computational Media Department University of California, Santa Cruz Abstract woman (McIntosh 1988). This evidence included many so- cial phenomena, including whether other people are gen- We argue that social simulation can help us understand social erally welcoming, pleasant, and supportive to her in daily justice issues. In particular, modeling certain social dynamics interactions, rather than suspicious, closed off, or even vi- within computational systems can be used to creatively ex- plore and better understand the social and identity dynamics olent. In her words, white privilege afforded her “comfort, of oppression. Writing theories of oppression in code forces confidence, and obliviousness,” which manifested especially us to explicate everything, and question what we leave out or clearly in her interactions with others. what we can’t account for. As an early step in this direction, Since the 1970s, there has also been a vein of this thinking we present an in-progress social simulation of group discus- in psychology, called pscicolog´ıa de la liberacion´ , or liber- sion in activist meetings, developed in the already-existing AI ation psychology, focused on the effects of oppression on system, Ensemble. Through this minimal, highly constrained the mental health of people and their communities (Mart´ın- social arena, we can explore wide-reaching phenomena like Baro,´ Aron, and Corne 1994; Burton and Kagan 2005; privilege, intersectionality, and power dynamics in nonhier- Moane 2010). This body of scholarship seeks to bridge the archical groups, but in a way that’s grounded in concrete, same gap that feminism observed between the personal and person-to-person interactions. We propose that this kind of social simulation can aid in the process of unlearning hege- political, an isolating, depoliticizing assumption of main- monic ways of being, and imagining liberatory alternatives. stream academic knowledge and popular consciousness. The Personal, as theorized in these and many other foun- dational feminist and social justice works, is the arena in Introduction which the effects of much larger systems of oppression are The Personal is Political daily and viscerally experienced by ordinary people. One of the most important practical functions of critical race and Interpersonal social dynamics can be invisible vehicles of feminist theory is to offer marginalized people a more lib- oppression and inequality. Among many reasons oppression eratory, healing lens through which to see the world and is damaging to individuals is that it is experienced everyday, themselves. Critical theories give people an opportunity to in interactions with friends, strangers, colleagues, and even “break with the hegemonic modes of seeing, thinking, and internal streams of thought. This kind of personally felt op- being that block our capacity to see ourselves opposition- pression doesn’t require malicious agents, racists, or sexists, ally, to imagine, describe, and invent ourselves in ways that but rather societal constructs and patterns of thought that in- are liberatory” (hooks 1992). This unlearning of hegemonic fluence everyone invisibly (unless explicitly made visible, worldviews is a deeply personal process, that can helpfully which can be done by scholarly analysis or works of media, inform interpretation of social interactions, identity, and per- including computational media). sonal difficulty. Even if these theories can’t immediately dis- A key tenet of feminism since the 1960s is that “the per- mantle entrenched oppression, their teachings of personal sonal is political,” or that the seemingly private issues ex- and interpersonal liberation can be internally healing, and perienced by marginalized individuals are often patterned can be enacted for mutual healing in our immediate social and systemic, and therefore not just personal. In her sem- surroundings. inal 1969 essay by that name, Carol Hanisch points to the consciousness-raising, analytical value of talking about per- Personal-Political Simulation sonal problems in an activist group, defending it from some fellow activists who considered it mere “therapy,” irrelevant The Personal is also the domain of expressive social sim- to political action (Hanisch 1969). ulation, which we argue can be designed into playful tools Two decades later, Peggy McIntosh articulated visible, to explore and critically interrogate the political dimension day-to-day evidence of privilege in her own life as a white of the Personal. We can create social simulation systems to augment and facilitate the process of unlearning hegemonic Copyright c 2017, Association for the Advancement of Artificial worldviews, and imagining liberatory alternatives. Intelligence (www.aaai.org). All rights reserved. Social simulation has been defined somewhat broadly and 61 informally for a game design context by Mitu Khandaker- Linear Storytelling about Social Issues Though engag- Kokoris as “anything that allows social interactions with ing, other fictional media representations of these topics only or between [non-playable characters] to meaningfully af- show audiences brief, heavily filtered glimpses of an under- fect the outcome of a situation” (Khandaker-Kokoris 2015). lying social theory. They aren’t working models of that the- These social interactions are computationally under-girded ory. One piece of such media can show just one example by some degree of either hand-written branching, as in Tell- of how specific characters in a specific context experience tale’s The Walking Dead, or emergent autonomous behavior, oppression or act for change. These are hugely valuable for as in The Sims. In her talk, Khandaker-Kokoris advocates for social justice theory communication because they are often greater investment in games about meaningful social inter- emotionally rich and grounded in a specific context. Audi- action in general, whether precisely authored or procedural, ences can therefore understand them and engage with them but places particular emphasis on the latter end of the spec- emotionally. These stories can suggest larger structures and trum. more common experiences, but don’t have to speak in a lan- We take this further, defining a social simulation more guage that accounts for these larger, more abstract ideas, as narrowly as a computational model which affords meaning- more theory-centered work does. ful social interaction with or between autonomous charac- But they also run the risk of audience over-generalization ters. For the purpose of this paper, a social simulation uses of the snapshot they present. Media about social injustice a simulation or emergent system under the hood, and is like Precious, The Color Purple, and Slumdog Millionaire not hand-authored branching. It uses processes to determine are critiqued for their perpetuation of negative stereotypes what interactions occur between characters, the content of about marginalized subjects, while they are acclaimed for those interactions, and how the characters are affected by amplifying stories about real pain that real people experi- the interactions, which in turn impacts what interactions oc- ence due to systemic, entrenched oppression. Both evalu- cur next. For our purposes, a social simulation also doesn’t ations are justified. These kinds of media are often both need to exist in a playable experience or game, though often deeply poignant as well as problematic for depicting only does. Simulations about social interactions allow players or one kind of experience of people at an intersection of iden- users to play or experiment with their existing ideas about tities. If audiences are critical and conscious of the prob- the social world. They still afford experimentation and play- lematic, the snapshot mode of media-making is fine—they ful interaction without a corresponding game experience. already have a relatively complex view of people, identi- It is because of this procedurality that social simulation is ties, and inter-personal politics. More insidious are media uniquely positioned to mediate social justice and injustice. like these in front of audiences who aren’t as aware of iden- Even in her general overview of social simulation in games, tity politics and non-hegemonic ways of thinking and being. Khandaker-Kokoris points to the potential for social simula- They tend to see only the entrenched stereotypical represen- tion to discuss issues of privilege, and the importance of a tations of an oppressed group of people. And so, the depic- diversity of characters in play centered around social inter- tion of a marginalized subject which should evoke empathy action. and make invisible experiences visible, instead reinforces harmful stereotypes. The Case for Social Simulation for Social Justice Systems representations, on the other hand, could poten- tially generate multiple examples of how the same set of ab- Computational media, and social simulation in particular, stract dynamics can play out in grounded contexts, whether can be uniquely leveraged in support of social justice. by player input, non-deterministic processes, given different The expansive systems of oppression which social justice initial states, or by tweaking system parameters. They are seeks to dismantle are notoriously difficult to understand. capable of dealing