This article has been published at the Dagstuhl Seminar 17471 "Artificial and Computational Intelligence in Games: AI-Driven Game Design". The original publication, along with its bibtex entry and other information can be found here.

AI-assisted Board Game Play
Joint work by:
Antonios Liapis, Mike Cook, Steve Dahlskog, Matthew Guzdial, Mirjam Palosari Eladhari, Emily Short, Gillian Smith, Anne Sullivan, Tommy Thompson

Over the last decade, an increasing amount of tabletop board games use digital media to replace or enhance elements of gameplay. As an example, the first edition of Mansions of Madness [6] required one of the players to take the role of a Keeper as an opponent to the remaining players, handling monster movement and revealing previously hidden information. The second edition of Mansions of Madness [8] "is now guided by a companion app that removes the need for a Keeper player, makes for a fully cooperative game, allows for single player gaming, and makes setup quick and easy" (based on the official FAQs). A plethora of other board games use primarily mobile phone applications for time management as in One Night Ultimate Werewolf [1], initializing and revealing hidden information when appropriate as in Alchemists [4], a combination of both as in X-COM [7] and Space Alert [3], and enhancing the atmosphere of the game e.g. via the sound effects in Stop Thief! [10].

Since board games are seeing a resurgence of popularity and a desire to innovate thematically, mechanically and technologically [5], it is the perfect time to analyze how digital game technologies and artificial intelligence can be applied to enhance analog play. Technology, especially if enhanced with AI, has many possible uses for board games in the near future. On the one end, simple enhancements such as saving or transmitting a game state can allow for games to be played in multiple sessions, or over the Internet. Assistive technology can be used to simulate the board game in a forward model digitally, allowing players to see the results of their actions one or more steps ahead: feedback can be provided to players in abstract form (e.g. whether this move is beneficial or not) or showing the full digitized board in a future state. In such a forward model, opposing players' moves can also be simulated using methods as simple as minimax [11] and as complex as a computational model of each player based on their previous decision-making strategies [9]. Beyond assistive technology for players, games in which a player acts as the game master can benefit from similar technology both for managing non-player characters (e.g. automating or semi-automating decision-making of individual NPCs digitally so that the game master confirms and moves the figures on the board) but also for providing inspiration and assistance during game preparation. On the more ambitious end of the spectrum, board game play mediated by technology can allow for customization of the game’s rules: this can be done based on human choice (e.g. checking which optional rules to use from a list), or automatically based on past game data in this player group or a global player community. Automated rule adaptation can be done when the game starts (ensuring that all players understand and learn the rules of play which remain consistent throughout the session) or in real-time as the game progresses (considering play data in the current game for dynamic difficulty adjustment [12]).

In light of the plethora of options for integrating technology and AI into board game play, a set of 33 usage patterns were identified. The usage patterns largely focused around increasing “time to fun” (e.g. by simulating the effects of a player action as a tutorial), enhancing the atmosphere (e.g. by adding sound effects or by offering additional rendering options through projectors), increasing depth of play (e.g. by managing hidden information or logistics beyond what a human game master can do), or changing the ruleset (e.g. by creating new rules or adapting existing rules). Using the format for describing design patterns followed by Bjork and Holopainen [2], four example usage patterns were chosen and further detailed: (a) tech for creativity support, (b) tech for atmosphere, (c) tech provides player flavoring, (d) tech adapts/creates rules. With this initial set of usage patterns in place, all 33 usage patterns are planned to be mapped in future research on the topic.


[1] Bézier Games. One Night Ultimate Werewolf, 2014.

[2] Staffan Bjork and Jussi Holopainen. Patterns in Game Design. Charles River Media, 2004.

[3] Czech Games. Space Alert, 2008.

[4] Czech Games. Alchemists, 2014.

[5] Owen Duffy. Board games' golden age: sociable, brilliant and driven by the internet. 2014. The Guardian. Online. Accessed 9 January 2018.

[6] Fantasy Flight Games. Mansions of Madness, 2011.

[7] Fantasy Flight Games. X-COM, 2015.

[9] Christoffer Holmgård, Antonios Liapis, Julian Togelius, and Georgios N. Yannakakis. Evolving models of player decision making: Personas versus clones. Entertainment Computing, 16:95–104, 2016.

[8] Fantasy Flight Games. Mansions of Madness, 2nd ed., 2016.

[10] Restoration Games. Stop Thief!, 2017.

[11] Stuart J. Russell and Peter Norvig. Artificial Intelligence: A Modern Approach (2nd ed.). Upper Saddle River, 2003.

[12] P. Spronck, I. Sprinkhuizen-Kuyper, and E. Postma. Difficulty scaling of game AI. In Proceedings of the International Conference on Intelligent Games and Simulation, 2004.

This article has been published at the Dagstuhl Seminar 17471 "Artificial and Computational Intelligence in Games: AI-Driven Game Design". The original publication, along with its bibtex entry and other information can be found here.