Skip to main content

Electric Dreams, Part 2: Optimists At Heart

What's Wrong With Game Research

"There's an undiscovered country of possibilities out there that we need to explore and create."

It's Monday morning on the first day of Dagstuhl Seminar 15051: "Artificial and Computational Intelligence in Games: Integration" and Michael Mateas is talking about impossible games. You might remember Mateas from the first Electric Dreams article - he was one of the scientific researchers behind Facade, a groundbreaking games experiment in interactive drama and artificial intelligence. Nowadays he runs the Expressive Intelligence Studio at UC Santa Cruz, a nexus of the world's best and soon-to-be-best games researchers. This January around fifty games researchers, including Michael and myself, came together in Germany for a week to talk about the future of our field and to work together to discuss some of the biggest research questions we're facing right now.

Last time on Electric Dreams we talked about the history of artificial intelligence in the games industry. In this second part I want to talk about the present day, and what scientific research has to do with all of this. I’m going to try to shed some light on why I think games research is broken and not benefitting games as well as it could be - but I also want to end on a positive note, and introduce you to the wonderful people and research that is going on right now around the world.

What did Mateas mean when he talked about an 'undiscovered country' of games? He was talking about narrative-driven games and how AI might be used to create games at a scale that humans simply cannot work at, but his words could have applied to any one of us and our research. When we talked about 'AI in games' last week, a lot of commenters' minds immediately went to classic examples of AI adversaries in games: real-time strategy greats like AI War, or leaps forward in first-person AI like Halo and FEAR. There are a lot of other ways AI can impact the future of games design, development and even game playing, though, and a lot of the people trying to make this happen found themselves at Dagstuhl.

Mirjam Eladhari is one such person. Mirjam has seen many sides of games and AI: in 2000 she started out programming mystery games at a company called Liquid Media. From there she gained a Masters and a PhD, and became a researcher. Up until recently she worked at the University of Malta, but at Dagstuhl she was talking excitedly about something new: a chance to go it alone for a few years as an independent researcher/developer hybrid under the label Otter Play. Mirjam has always tried working on small projects in her spare time that combine her research and game ideas, but it's hard to balance with teaching, paper-writing, conferences and the search for funding. "At some point you have to realise you're not superhuman," she tells me in the canteen, "I've worked in this area for many years, I feel like I've earned this."

Mirjam wasn't alone in contemplating a different career at Dagstuhl. To name just a few: Adam Smith, a brilliant young postdoc who uses AI to invent puzzles and educational games, has begun trying out freelance consultancy; Ian Horswill is secretly developing an AI-powered roleplaying puzzle game in his spare time; Tommy Thompson is running a Patreon to fund his inventive and entertaining videos and lectures about AI; even one of the Dagstuhl organisers recently released his own Android game). Many of the people following more traditional career routes seemed pressured and exhausted by the demands put on them by the university system, but didn't see a clear alternative either. Everywhere I looked it seemed I found people who were looking for something a little different from the standard academic career path.

What is it that makes that path so bad for games researchers? Clearly for some researchers it's not bad at all - lots of people have terrific careers balancing teaching, administrative duties, supervision and even a little bit of research on the side. In general, though, academic careers are a bit of a mess, and a big part of this is that as people get better at doing research, we encourage them to do it less often. Lectureships and tenure track jobs at universities are the only way to get reliable and secure employment in public research, but they pile pressure on academics and pull them towards teaching, student supervision and service duties. They also prioritise certain kinds of academic activity over others - writing for particular journals, hitting particular targets. The people who we spent ten years and hundreds of thousands of pounds training are slowly moved into roles where they have less and less opportunity to do what we trained them to do.

On top of these problems with competing demands on researchers' time, the way academic funding works skews research in very specific directions. Researchers apply for funding wherever they can, and for those whose fields are not fashionable at the time this can mean you end up doing research you don't really want to do, or write grants applications you don't really intend to fulfil. Some researchers will tell you this is just how the system works: you apply for money in such a way that your plan overlaps both your research goals and whatever the funding agencies want that particular year. In reality, though, funding does guide and influence what researchers do and how they do it. Economic impact is valued above cultural impact. The social problems identified by funding agencies are more important than the opportunities for benefitting society that researchers may identify. Industry applicability is held up as a holy grail, a grand achievement for engineers and scientists - which leads to the feverish rush to mainstream we talked about in the previous article.

Good research does get through. Tommy Thompson's outreach work runs alongside his exciting work on a Spelunky bot API his students have been working on - paving the way for bot competitions, procedural level generators and more. Noor Shaker is spearheading work on procedural content generation that understands player preferences. Jon Tremblay and his colleagues are repurposing ideas from robotics to analyse and generate levels for stealth games. Gillian Smith is performing digital archaeology on the history of content generation, in order to help us better understand where to go next. Every conference I go to, I am reminded of how vibrant and energetic the world of game research is - and I only see the small sliver of it that intersects with artificial intelligence. But it's a struggle, and it comes at a cost. People are leaving for other industries, or settling down for quieter lives. Those that stay tell jokes over beers in hotel bars about their sleep schedules and travel itineraries, but no-one's laughing, really.

In the middle of the week at Dagstuhl the group went for a walk through the surrounding snowy forests. I spent the hour trek talking to a few researchers about some of the issues I've brought up to you today. The question that hung in the air between myself and one other researcher in particular was what the next step should be. They were positive that research can be changed from within, and that these systems can be improved if we fight and pressure the right organisations. I'm not so sure. Later on in Electric Dreams, I'm going to argue that the changing face of the games industry means that there might be other options, and that it might be possible to create spaces between the games industry and university life where new, exciting research can be done, and new, exciting games can be made. Like every other part of the games industry, games researchers have a contribution to make to the future of games. If we don't make spaces where we can do this work, Michael Mateas' “country of possibilities” may remain undiscovered forever.

For now, let's do something simple: if you're reading this and you identify as a games researcher, whether you're in game studies, AI and technologies, the social impact of games, design, or anything else - please do me a favour and leave a comment below introducing yourself to RPS readers. Even if it's a one-liner with your name and research interests. Let's make the comments section a little meet-and-greet for everyone, and show off the amazing variety of work that's going on out there. In the next part of Electric Dreams we'll be leaving research behind us for a while and looking at the games that stood out lately as particularly good uses of artificial intelligence. What do they do different and what can that tell us about how good technology can lead to great games? We'll find out in two weeks' time.

Read this next