AI Conference Round-up: Dagstuhl 2015

The life of a researcher is often a strange one: we are in a fortunate position in that we often visit conferences that take place around the world in an effort to showcase our work. While there, you see talks and receive questions from the researchers whose names you cited as you conducted your literature reviews. As a junior researcher, meeting these people can often prove daunting. However, in time you approach it with much greater enthusiasm. The opportunity to catch up with your peers, those who have dedicated their careers to working in the same weird and wonderful corner of the world as you, who in time you come to know as friends. All the while grateful that your career allows for such annual get-togethers.

While the emphasis of conferences is to present research; to showcase what you, your research centre and your department are working on, it is more important that you take the opportunity to sit and talk with people. Networking at conferences is infinitely more rewarding than attending any conference presentation (or delivering one). The chance to sit and discuss your work, that of your peers and generally throw around ideas in the company of intelligent and enthusiastic people. Sadly, we don’t get to spend that much time doing it, given of course the primary focus of the conference (talks, talks, talks). For any junior academic, the more important advice I can give is to embrace the opportunity to ask all those questions you had when reading the paper of a fellow researcher now sitting across the room. Most of us are easy-going people and are happy to discuss our work with others – in some cases it strokes an ego, but in the case of many we’re simply grateful somebody else thought it was cool. More importantly, it can spur new ideas for research in the minds of all involved: you never know how one random discussion can lead to the formation of new research projects.

Schloss Dagstuhl

This all comes to a head in this piece as I talk about a recent trip to Schloss Dagstuhl: a ‘research retreat’, if you will, that sits near the small town of Wadern in the southern German countryside. It is an 18th century castle that has been repurposed as a centre for discussion and collaboration for those working in Computer Science. February of 2015 saw the second Dagstuhl seminar dedicated to Computational and Artificial Intelligence in Games, where a number of folk – sadly not all – from our field come together to discuss the big issues and the even bigger ideas. Only this time, instead of it taking place over a couple of drinks on one or two evenings after the paper presentations, these discussions are the focus of the event.

Most (not all) of the attendees at this years Dagstuhl seminar.
Most (not all) of the attendees at this years Dagstuhl seminar.  Can you spot me?

As noted previously, this is the second seminar under the AI/CI in games banner. Given it is an event attended solely by invitation, I feel truly privileged to have had the chance to attend both events. I openly concede I do not produce as much research as many of my peers (not at the same level of quality), so to be considered is rather humbling. It is unfortunate that the venue cannot accommodate for all researchers who would like to attend: which is arguably a reflection of how the AI/CI in games community has grown over the last 10 years. What was encouraging to note was the number of junior researchers who are now at the end of their doctoral studies in attendance: adding a body of fresh perspectives to the group.


The focus of this event was ‘integration’, looking at how previously established ideas can be brought together to open research avenues. This often resulted in blue-sky topics that were then discussed in groups and subsequently grounded into actual projects with tangible benefits and achievable milestones. This ranged from ideas such as the design patterns for AI-based game design, to adopting AI in game analytics, the future of player and behaviour modelling and how to assess and measure AI research that intersects with the arts. This is undoubtedly the highlight of the event: new research areas were established, others expanded upon. As a community we have identified new areas to tackle and invigorated existing avenues. I would argue that as a community we achieve more at Dagstuhl – in terms of establishing our aims, objectives and impact – in one week than in a year of conference visits. It’s terribly exciting.

Jamming with a bunch of serious looking researchers.  From left to right: Michael Cook, Julian Togelius, Alex Zook and Gillian Smith.
Jamming with a bunch of serious looking researchers. From left to right: Michael Cook, Julian Togelius, Alex Zook and Gillian Smith.

In addition, what stood out for me in this second instance was the notion of experimenting beyond our traditional scientific boundaries: to explore ideas of game design impacted by AI, of how procedural generation techniques intersect or clash with formal design theory suggests a maturity and confidence in our field that continues to evolve. I’ve often felt that an implicit fear existed of exploring such concepts due to the possibility it may be deemed as lacking merit or integrity. However, the wealth of high-quality research that has been generated by this community in the last 10 years has helped release such inhibitions. It’s also aided by new people coming into the field who embrace the indie development culture; seeing themselves equally as game designers, game developers and game researchers.

This in-turn led to one of the biggest differences between this visit to Dagstuhl and the last was the idea of hackathons and research jams. A day was dedicated for teams to go away and work together on new projects and write code that addressed a burning question or desire. Given I was part of the team looking at ‘AI-Based Game Design’, a small crew spent a couple of evenings actually making (small) games that embrace these ideas! We actually made stuff, at a research event. That’s unheard of! While the idea of ‘jamming’ is increasingly common in the indie-gaming scene, the idea of doing such a thing at a conference is unheard of (we barely even have time to play games at conferences nevermind make them). I truly hope that this can be used as means to challenge how we coordinate our conferences and encourage such activities to continue in the future.


The impact of this event, like the last one, will be noted in the years to come. One need only look at general video game playing as an example of Dagstuhl discussion that manifested into a bevvy of research publications and a competition to boot! I suspect that impending conferences, notably the Foundation of Digital Games (FDG), Computational Intelligence and Games (IEEE-CIG) and Artificial Intelligence for Interactive Digital Entertainment (AAAI-AIIDE) will see many publications that are building from the topics, projects and ideas born at Dagstuhl this year.

In closing, I want to thank everyone who I had the pleasure of hanging out with at this event; whether we worked together or had a chat over lunch or coffee, it was a great experience.  I’ll come back to discuss the fruits of our work at a later date, including the games that were jammed and the relevance of why we were even making them in the first place!

Enjoying AI and Games? Please support me on Patreon!
Tommy Thompson Written by:

Tommy is the writer and producer of AI and Games. He's a senior lecturer in computer science and researcher in artificial intelligence with applications in video games. He's also an indie video game developer with Table Flip Games. Because y'know... fella's gotta keep himself busy.