Teaching Game AI and Social Justice
Yesterday was our first day of classes at Northeastern. So over the last week, I’ve been tweaking my syllabus for Game Artificial Intelligence. Adjusting assignments that didn’t work well last year, adding in time for discussion on concepts that needed more time, agonizing over what to remove. It’s one of the hardest parts of teaching: identifying and making time for what is important.
The last few weeks have really been the crescendo of misogyny in games. It’s hard to be surprised by the abuse hurled upon Zoe Quinn and Anita Sarkeesian, and it’s also hard to work past my anger that not only is this still happening, but it’s getting worse. And as I see friends post angry articles on Facebook, and sign on to open letters arguing for basic human decency as though it is a revolutionary concept, I wondered for maybe the billionth time… what can I do as a teacher to try to make a difference?
At the GDC education summit a few years ago, Ian Schreiber shared a provocative image–a pipeline of raw sewage, feeding into a murky brown river—and posited that we in higher education are that pipeline. It is our responsibility as educators to push for change in our community. To encourage underrepresented groups to bring their ideas to a problematic industry at the same time as teaching those in privileged groups that these issues are very real and that they can help fight them. And there are some really great classes that address these topics explicitly, teaching about games and societal issues, and more broadly about gender studies, social justice movements.
But often the education of students about such problems ends in those classes. It is the nature of most universities that classroom learning is compartmentalized—we teach about algorithm efficiency in one class, iterative prototyping in a second, and institutionalized racism in a third. Each of these classes tackles difficult material and students have to work hard to learn it. But we rarely explicitly teach the hardest part of all: connecting the dots between all of these concepts.
And thus I decided: we can spend one less class period talking about evolutionary algorithms for procedural content generation, and one more class period talking about social justice, ethics, and game AI. I will call this class period a success even if students come out just being more aware of the problems being faced in our industry. But what I would really love is if they can think through how the decisions they are making at the low level in their AI systems (e.g. making gender a boolean variable, or the myriad and potentially problematic ways that PC-NPC relationships are modeled) reflect and promote systemic biases in games as a whole. And if they can use this knowledge to guide their AI projects moving forward in the semester: what is the role of AI in games that explicitly advocate for change? what could they be considering in regard to gender, race, and sexuality when designing AI for more “traditional” entertainment games? I hope this will be as much a learning experience for me as it is for them.
We’ll be doing this early in the semester, just a couple of weeks from now. I wanted to get this in early, to make time for what is important. This shouldn’t be an idea that gets lost in the end of semester rush, and this should be something they can keep in mind throughout the semester… even though they are still figuring out what Game AI even *is* right now.
I would love to hear from people about examples of games that may be relevant that we can talk about in class, or any work that is explicitly looking at the role of AI in how women and minorities are represented in games. I’ll report back.