http://www.dagstuhl.de/15051

25. – 30. Januar 2015, Dagstuhl Seminar 15051

Artificial and Computational Intelligence in Games: Integration

Organisatoren

Simon M. Lucas (University of Essex, GB)
Michael Mateas (University of California – Santa Cruz, US)
Pieter Spronck (Tilburg University, NL)
Julian Togelius (New York University, US)

Koordinatoren

Mike Preuß (Universität Münster, DE)

Auskunft zu diesem Dagstuhl Seminar erteilt

Dagstuhl Service Team

Dokumente

Dagstuhl Report, Volume 5, Issue 1 Dagstuhl Report
Motivationstext
Teilnehmerliste
Gemeinsame Dokumente

Press Room

Summary

The research field of artificial and computational intelligence in games focuses on the wide variety of advanced computational techniques used to create, enhance, and improve the experiences of humans interacting with and within virtual environments. By its nature the field is broad and multidisciplinary. People working in it include academic researchers from a variety of disciplines, corporate researchers in the games industry as well as in other industries, and independent game developers. The methods used include symbolic AI techniques such as reasoning, constraint-solving and partial-order planning as well as biologically-inspired techniques such as evolutionary computation and neural networks, statistical techniques such as support vector machines and clustering, as well as special-purpose techniques such as behavior trees. These are applied to games ranging from board and card games to first-person shooters and real-time strategy games as well as abstract mathematical games, and are used to, for instance, play games, model players, tell stories, generate levels and other game content, and match players. Different researchers have different goals, including developing and testing new AI methods, creating interesting and believable non-player characters, improving the game production pipeline, studying game design through computational means, and understanding players and patterns of interaction. Often several goals overlap in the same project.

Recent years have seen considerable progress in several of the techniques used in the field, as well as rapid development in what kind of research questions are asked and what kind of games are studied with which methods. It has become increasingly clear that many of the research goals require a multidisciplinary approach, or at least a combination of techniques that, in the past, were considered separate research topics. For instance, with respect to the behavior of virtual agents, ten years ago researchers mainly aimed at making such behavior more "effective," which can often be achieved with straightforward computational reasoning. Nowadays, however, researchers aim at making the behavior of virtual agents more "natural" in their interaction with humans, which requires contributions not only from computer science, but also from psychology and social sciences, and which requires a wide variety of techniques, such as player modeling, adaptation, reasoning, and computational linguistics.

To move the research field forward, it is therefore of crucial importance to facilitate the integration of the disciplines and techniques that are involved in this research. The various strands, methodological approaches, and research directions need to inform each other and collaborate, to achieve a whole that is more than the sum of its parts. The goal of the second Dagstuhl Seminar on Computational and Artificial Intelligence in Games was to explicitly take the first steps along this path of integration, and investigate which topics and techniques would benefit most from collaboration, how collaboration could be shaped, and which new research questions may potentially be answered.

The seminar was held between January 25 and January 30, 2015. To stimulate interaction between the participants, which is essential in this context, the seminar was structured around workgroups rather than presentations. The organizers started the seminar on Monday morning with a series of brief presentations on potential discussion topics, after which the participants formed their own workgroups around a variety of topics, not necessarily those brought up by the organizers. Workgroups typically consisted of 3 to 10 people from different backgrounds, who worked together for no more than one day. At regular intervals workgroups reported on their findings in a plenary session, after which new workgroups were formed.

At the start of the seminar it was announced that Thursday would be set aside for practical work. Participants could use that day to implement some of the ideas that had come up in the previous days, in the form of a game, a competition, a design document, or a research proposal. While the organizers deliberately gave the participants the option to simply continue with the workgroups if they so wished, all participants actually got involved in the practical work, some of them even working on multiple projects in parallel.

The results of the workgroups and the practical sessions are briefly related in the remainder of these proceedings. The 13 abstracts on workgroups cover automated and AI-based game design; game analytics; interdisciplinary research methods; design of believable characters; general video game playing; creativity facet orchestration; methods and formal design for procedural content generation; design of "fun" gameplaying bots; communication on game AI research, computers that play like humans; and neural networks for games. The 11 abstracts on practical sessions cover the Planet Wars competition; the automatic generation of games, mazes, and text; Twitter bots; sonification of character reasoning; MCTS and representation learning for procedural content generation; two AI-based games; and the design for a board game.

A special issue of the IEEE Transactions on Computational and Artificial Intelligence in Games will be published on the topic of this Dagstuhl Seminar. While this issue is open for submission for any researcher in this field, it is expected that several of the workgroups of the seminar will submit papers on their results.

As organizers, we knew that the first seminar that we organized in 2012 was considered a great success, and we had expected more people to accept our invitations for this second seminar than for the previous one. However, demand for attending the seminar was even greater than we expected. Almost everyone we first invited immediately accepted our invitation. Moreover, everybody who accepted their invitation indeed showed up at the seminar. We were forced by capacity concerns to not invite many people who, by their strength of contribution in the field, should have been present. We are certain that we could easily have doubled the number of participants visiting the seminar, and that each of those participants would have made a strong contribution.

The value of these Dagstuhl Seminars is indisputable. Considering the large number of researchers that should be invited to a seminar that attempts to cover the whole, very broad research field of Computational and Artificial Intelligence in Games, we believe that it is wise for a future seminar to narrow down the topic, so that it can be restricted to a smaller number of participants that are active in the selected subfield. Naturally, considering the fact that "integration" is such an important issue in the research field, care must be taken to ensure that every discipline interested in and involved in the subfield is represented.

License
  Creative Commons BY 3.0 Unported license
  Simon M. Lucas and Michael Mateas and Mike Preuss and Pieter Spronck and Julian Togelius

Dagstuhl Seminar Series

Classification

  • Artificial Intelligence / Robotics
  • Multimedia
  • Soft Computing / Evolutionary Algorithms

Keywords

  • Experimental analysis
  • Meta-heuristics
  • Multi-agent systems
  • Dynamical systems
  • Efficient algorithms
  • Entertainment modeling
  • Player satisfaction
  • Game design
  • Serious games
  • Game theory

Buchausstellung

Bücher der Teilnehmer 

Buchausstellung im Erdgeschoss der Bibliothek

(nur in der Veranstaltungswoche).

Dokumentation

In der Reihe Dagstuhl Reports werden alle Dagstuhl-Seminare und Dagstuhl-Perspektiven-Workshops dokumentiert. Die Organisatoren stellen zusammen mit dem Collector des Seminars einen Bericht zusammen, der die Beiträge der Autoren zusammenfasst und um eine Zusammenfassung ergänzt.

 

Download Übersichtsflyer (PDF).

Publikationen

Es besteht weiterhin die Möglichkeit, eine umfassende Kollektion begutachteter Arbeiten in der Reihe Dagstuhl Follow-Ups zu publizieren.

Dagstuhl's Impact

Bitte informieren Sie uns, wenn eine Veröffentlichung ausgehend von
Ihrem Seminar entsteht. Derartige Veröffentlichungen werden von uns in der Rubrik Dagstuhl's Impact separat aufgelistet  und im Erdgeschoss der Bibliothek präsentiert.