This talk from Different Games 2015 was part of a panel presentation along with Amanda Phillips, Tanya Short, and Michael Cook. The panel prompt was: "can computers be feminists?" In this talk, I argue that computers cannot be feminists because artificial intelligence lacks empathy, but that as designers we have a responsibility to ensure that the algorithms we create are imbued with the knowledge required to behave as though they are feminists.
can computers be feminists? well, no. but let me explain…
when we make generative systems, we often care about controlling the content that they will be able to make: fitting in to the aesthetic goals of the overall system. there are two ways we can do this…
by fiddling with the bits of knowledge that make up the overall piece of content
or by fiddling with the algorithm we use to make the content
or somewhere in between
when we do this, we are creating formal theories of design. two different kinds: first, for the artifact being made (what is a level? what is a quest? what is a weapon? what is a story? what is a character?)…
…second, for the process that should be followed to make it (is design an iterative process? a reflective one? or as in this image, one in which there is no iteration or reflection, but many possibilities that can be chained and explored?)
the meaning that emerges from a game comes, in part, from how the player interacts with these underlying procedural systems, which in turn is controlled by the aspects of player behavior that are surfaced to the system and the ways in which the system can respond
so when we work to create generative systems, we are actually embedding meaning. like this character generator, which is loosely generative, by putting together bits of humans without regard for how they fit together. or the dwarf fortress map generator, which follows a simulated history.
and when we make these systems, we can build them on more complex theories than what we are already thinking about (like this prototype interface for character creation that respects gender and sexuality on a spectrum). and doing this is hard, it requires a lot of iteration with a lot of people, to figure out how to formalize some theories that are really hard to formalize, and there will always be gaps between the human-understood theory and the machine-understood formalization.
really, what we’re doing when we build generative and/or procedural systems, is we’re creating little psychopaths. our AI systems lack empathy, they lack intent, beyond what we give them ourselves. we teach our generative systems how to fake empathy to be able to create content that makes sense to the players. humans interpret intent and meaning.
right now, we’re hearing a lot about learning from “big data” and “the cloud”, leaning on building systems that “learn” generative models from lots of data (such as Mike’s system)… but these can never intentionally embed meaning, because the computer does not have any semantic information to be able to present and reason about it. but they still convey meaning, in the way that the human interprets and plays with the information being presented to them.
and our role is to create simultaneously the most effective and the least harmful little psychopaths we possibly can. we need our systems to be effective at faking empathy so that we can create the experiences we want to create, with systems that are responsive to players, that create meaningful and interesting content (though it is interesting to ponder what happens if we stop having that as a goal). and we have a social responsibility to do it in the least harmful way possible, to be inclusive for players, and to project the meaning we want to be projected.
I want all people who create these systems to realize that they are making choices. That the theories that are implicitly embedded in their systems, via their biases and experiences, do not need to be embedded there. That we cannot shrug our shoulders when the computer does something bad and blame it — it’s a psychopath, it doesn’t know any better, and it never will. But we know better.
a computer cannot act with the intent of a feminist
but in a generative system, the actions taken by the “designer” will be interpreted by players as though the system is a human, so…
it’s on us. we need to be feminists, we need to think about how to embed feminism and inclusivity into our systems, our data, our games, so that players will see it as well.