Many of you will be thinking right now, “…but wait…isn’t artificial intelligence only used in computers?” Depends on your definition. AI, to me, is any set of rules that allows a system to make decisions on its own without human input. Computer games with pre-programmed monsters who follow your every move and trail along behind you are following a pattern determined by AI. It may not be a pre-programmed path, but depend instead on responding to your own decisions. Based on real-time input, it alters its own actions.
AI in Boardgames
So how does this appear in boardgames? Some boardgames allow creatures to move based on simple directions (hug the left wall, move 1 space per turn) with the input being “a new turn”. One creature in Wiz-War depends on players entering its line-of-sight (Fire Imp – old edition) to trigger its own action. This is AI in board games at its simplest. There is an input (turn change, player action), then a variable response from the creature/tool/deck/payout or whatever based on the new input.
In one game design (unpublished) I circumvented the need for a creature AI merely by letting the players move the monsters toward the opponents or their own heroes toward the monsters. I can’t really call this an AI; the players are the active intelligence in such a scheme. But it was a workaround to avoid complicated AI monster movement and action rules. This is a fairly common workaround in game AI; if you really want it to be intelligent, let the player on your right act as your nemesis on your turn.
So what’s this article about? I’d like to take a quick review of a number of AI schemes that various designers have used, techniques to let the game-board make the decisions. This should, hopefully, give you some game design ideas of your own.
AI Levels
AI in board games functions at two levels;

One, it’s just a set of programmed instructions that lets the board do things autonomously, as with the boulder and crushing walls in Adventurers: The Temple of Chac, and the mummy movement in Adventurers: The Pyramid of Horus. The board doesn’t really interact with the player actions; you just don’t want to be standing in the wrong place when things start happening.

At the second level of AI, the AI responds to player actions, as with Fearsome Floors, where the player movement actually changes the decisions of the monsters moving around the board, or with Zombicide, where the Zombies chase you down, attack you, split up in groups, or pay attention to the level of noise you’re making. One can easily argue that the first level of preprogrammed, non-responsive actions are just rules, and not AI at all, and I mostly agree. An AI should, at some level, be responding to the player’s actions. But I’ve included both types here just for the sake of completeness (like including soda at a wine-tasting event).
Note that almost all cooperative games require some sort of AI mechanism to create and operate the opposition forces if they wish to avoid the use of a “game master” to run the scenario.
Let’s look at some of the techniques designers have used:
1. AUTONOMOUS RULE MECHANISMS
Game rules feature some autonomous mechanism that allows things to change in the game without any player input, like moving a mummy or slaver along a fixed path, sometimes with die rolls to randomize direction or action (used in Adventurers, Freedom: The Underground Railroad, Break The Safe, Kill Doctor Lucky, Castle Panic, Dice Crawl, etc.).
2. CARD SUBROUTINES
Cards that create a little subroutine of action to be followed, usually creating or activating a monster or other attacker, providing instructions on how the creature, spaceship, or fighter is supposed to act. This isn’t significantly different from #1, but is provided in bite-sized pieces on cards instead of a big fiddly rule book. (used in Battlestar Galactica as Crisis cards, in Castle Ravenloft, and in Wiz-War)
3. ACTIVE RESPONSE
Game rules that allow the creatures on the board to actively respond to a player action (what I’ve been calling “level 2” AI). You move, they follow. (used in Fearsome Floors and Zombicide and Wiz-War)
4. DUMMY PLAYERS
Fresco, 7 Wonders, Bridge, Poker, Alhambra and Tokaido – these offer dummy players; artificial players that tend to take somewhat random actions, but allow you to play a game that should really have one more player than you have. Sometimes it adds an unknown element to what could be a fairly deterministic game. It’s only embarrassing when the dummy wins. Some games duplicate this with “robotic players”, as with Race for the Galaxy: The Gathering Storm and Power Grid: The Robots.
AI potential
AI in board games is obviously not needed to make a great game. The large majority of games use none at all; the AI in the game is just your human opponent. But it is a handy tool for game designers to keep in their kits. It would be interesting to play a game with a seriously powerful AI in it that doesn’t require turning on a computer. Can you imagine a game of Berserker where the Berserker is an AI that draws a new card each turn, and each card is a set of additional reactive instructions? Don’t you want to just go design a variant of that right now?
I think the concept of AI delivered by card text is underexplored. AI cards could control board movement in a tactical game; something like having a random robot in Roborally programmed by a randomly selected card hand, or a wargame where each action by the enemy is controlled by card directions drawn from a deck. You might have to pump up that enemy’s strength to offset its stupidity, however.
Story of a Wise-Ass Robot
The idea of games merged with AI has always fascinated me. In a story I had published this year by Fox and Raven called Paper Man, I explored the question of whether an artificial intelligence could actually arise from a board game. The story follows a wise-ass robot dealing with his own AI issues as he becomes intrigued by this bizarre game. Not to toot my own horn, but I really like how the story came out and would encourage everyone to go out and spend their hard-earned $0.99 to read it. Toot!
Thanks for Kelsey Domeny, Luke Laurie, Brad Brooks and Norv Brooks for recommending a variety of games mentioned in this article.
Tom Jolly
Latest posts by Tom Jolly (see all)
- Patents, Protections, and Boardgames – Part 1 – June 13, 2016
- Books, Games and Movies; The Value of Entertainment – March 18, 2016
- Making Semi-Cooperative Games Work – August 31, 2015
31 Readers Commented
Join discussionWe’ve got to get out Freedom, the Underground Railroad and give that a good play.
I love the concept of AI in board games. One thing that interests me is the combination of board games and smartphone/tablet apps. While this is somewhat “turning on a computer”, the interface is much more mobile and can be brought to the table very easily.
___________
I can see all kinds of great applications for competitive as well as cooperative board games integrating with apps.
This is an interesting idea. Do you have examples of games that use an app to control a programmed player in a board game? Is there anyway for the app to interact with the set up of the board game (that is, how does the app determine what it should do?)?
I believe one of the best examples I’ve seen lately is Mansions of Madness, 2nd edition. FF really nailed it – in 2nd edition app from your phone or tablet replaces a Keeper (game master in MoM).
Navajo Wars (2013, Joel Toppen, GMT) is an example of rather advanced AI in board games, and it needs to be: you play solo.
The COIN games (2012, Volko Ruhnke, also GMT) have some pretty advanced AI as well, enough so you can play a drop-in game by swapping AI flowcharts for human players. Nothing substitutes for the politics, though.
That’s the hardest part about game AI, imho: politics.
Fascinating topic, sometimes it’s difficult to define AI in board games. A lot of co-op’s offer AI elements as do the “Dummy Players” in other games. A lesser known game that uses a form of AI is Michelangelo. The game is about completing art projects, at the end of each round “Michelangelo” helps finish an art piece based on what players did that round.
I think the first time I was really introduced to this idea and it registered with me was Pandemic. This is a great topic, thanks for posting Tom.
We started a board gaming company this year that solely makes games that take advantage of the current technology, namely smart devices. We believe board games need to be awesome and operational without the smart device. They do however allow a great number of new features in board games, such as a really smart AI. One of our games was played fiercely at The Spiel in Essen in Germany this year and another one was on display. Audience loved both and it’s clear that more games like ours will pop up in the future.
This year introduces a few games that use smart devices like XCOM and Alchemists. GREAT looking games but they both have one sad feature: You can’t play them without the smart device.
This article talks about “cardboard AIs”, but I would’ve liked to see something about computer based AIs as well on this article. It is after all a pretty big thing in the near future.
To a certain extent most board game AI will be scripted and thus offer no surprises to the players. One recent game (that I haven’t played) that seems to make things interesting is Dead of Winter, where (if I understood correctly from a review) event cards present situations or game states which trigger effects. The event cards being randomized, this leads to a different progression in each play through, which sounds very interesting.
That being said, I’m very keen to see what will come forth in the hybrid sphere of games such as Alchemists and X-Com. If put to good use, a device operated AI could make for an interesting design tool in board games.
“I think the concept of AI delivered by card text is underexplored.”
This is nearly the entire premise of Sentinels of the Multiverse. (https://boardgamegeek.com/boardgame/102652/sentinels-multiverse)
It certainly would have been worth mentioning digital AI apps supporting board games, but it didn’t actually occur to me; it’s been around since people start including VHS tapes with certain games, then DVDs. But it’s definitely getting better.
The best AI in a boardgame I have experienced is from the “Ambush!” series of games by published by Victory Games. They used a “programmed paragraph” system that was very enjoyable to play against. Check out:https://boardgamegeek.com/video/41357/ambush/lets-play-ambush-scenario-7-part-1
for a lengthy video “let’s play”. Ambush! was released in 1983, and won the Charles S. Roberts “Best Twentieth Century Game” award that year.
Though it’s difficult to unscript board game AI mechanics without using a randomization mechanic, things like chain events or combos as a result of an effect might be a good way to explore. Inspired by this article I will be looking at some of these ideas with my next product.
Great post!
randomization can be done via event-dice that collectively show the action, or by using cards (which in a sense has memory).
Which is the game in the first pic, the one with homunculus? Tks in advance,
That would be Wiz-War!
Many programmers (including me) are going to cringe at the use of the term “AI” for something that exhibits NO intelligence, even the much-more-complex programming possible on PCs. “Computer (or Programmed) Opponent” is much clearer.
For that matter, if the computer opponent in a video game is too good, players assume it’s cheating. So the C.O. is usually programmed to give the player a hard game and lose, not to be the best player it can be.
Valid point; I probably wouldn’t have said AI unless it had already entered the common jargon of computer programmers; “How’s the game AI?” “Programmed opponent” is much more appropriate, but it wouldn’t trigger the common meme. As for ACTUAL AI in a game, that’s what my short story was about.
Tom,
I had never heard of Beserker before, and given its low rating on BGG, I probably wouldn’t have paid it much attention. Is this an old favorite?
On the subject of AI, what if a combination of Human player and templated cards was used? Each of four players would be given 5 templated cards. On each card would be a question, like “If a player crosses the AI path I will…”. The writer would describe a choice that the AI makes in the given situation. All of the cards are shuffled together to create the event deck. This could provide more intelligent responses AND an element of surprise for the other players when a card is drawn that they themselves did not create.
Berserker wasn’t a very good game in my opinion, though it did have some interesting insights into game AI. It was poorly balance though, and if I remember right, the player with the Berserker almost always won. Interesting idea you have for an AI mechanism; you should make a prototype. I think it would be a hard sell just because players have to come up with interesting actions, and it would be hard to limit and define what they could add on as the AI action. Like, “If a player crosses the AI path, I will instantly kill him.” I tried to create a similar game once, based on A Thousand Blank Cards, where players wrote in a rule on 3 cards, shuffled the cards, then dealt them out. There was a fundamental block of rules (like, each card is worth a point), then everything else was left to the players’ imaginations. I like the idea, but it was pretty hard to make a rigorous framework of rules that the players couldn’t totally screw with.
On BGG a post says Berserker has no chance to even scratch earth’s surface. But a CAI for Berserker would be incredibly hard to imagine.
Interesting article. I did a simple AI card for the Mammoth in my game Moogh. It is a 100% predictable to know the next ai move – which works quite cool in the game. I am considering using dice to control it instead – but then I would loose some AI. (actually my control card is not really AI – because it does not respond to certain actions).
Just came across this article, quite interesting. Earlier this year I was designing something similar to what Tom suggests under ‘AI Potential’ for my game Centauri Rift. The programmed players (or Non-Player Characters) are controlled by drawing a card from their specific model deck (there are 3 different space ship models in the game). The drawn card specifies how many movement points, how much attack damage, and whether the ship finishes by making a small getaway or not. It would be great if it were that simple, but to make it work I also had to add a list of instructions for players to follow since there is still some player thinking involved in fulfilling the programmed player’s turn. The default is ‘move toward and attack the nearest enemy.’ The rest explains how to deal with exceptions (e.g. there are two enemies equidistant). As the designer, I enjoy it, but I fear that it might be too complex for others to use and enjoy. If someone here was willing to take a look at the rules and give some feedback on that, it would be much appreciated.
You can find the rule book at: https://tabletopgeneration.com/project/centauririft/details (pages 25-29 of the PDF are the most relevant and include an example turn).
You can also find out how to contact me by email at: https://incurabledreamer.com/contact.html
This game (by me) offers a fully visible AI ‘script’ that players can control 90%. Luck and other factors has an effect on the ai outcome. I am considering to include some dice in the AI action tree…
which “this game”?
Moogh is the name. It is coop against a mammoth. My current ai is almost without surprises which maybe make it less intelligent 🙂
Having a bigger boardgaming urge than my friends and family I was interested in finding out about Cardboard AI (let’s call it CAI to not have programmers go cringe), and surprised there was almost NO video about them.
I would like first to keep the definition clear and therefore propose to make a distinction between phone/app based AI games vs pure cardboard AI games. I am not interested in spending my time looking at yet another light-emitting screen, so let’s focus on “in-the-board-game” elements that make up a CAI. And to keep it intuitive let’s constrain also that the CAI must be more than static rules, the CAI “opponent” must react in some dynamic way to player actions, in order to make it PvE (as in video games). I consider app-assisted games like that pharao game that has you peek into and describe the dungeon just normal board games, albeit of a novel and interesting kind.
What I missed about the OP’s text was some homework: there is a game called Tiny Epic Galaxies that has several CAI when playing solo; there is on BGG two variants that have CAI on Patchwork (although they do change the core game), BGG also has a post on CAI in Agricola:All Creatures Great and Small; perhaps you know some more.
I have no xp in TEG, but the latter games indeed have a CAI that reacts to the player’s actions (well, at least the last move of the player).
I think the biggest problem still is that people have personalities and no matter how you design a CAI it will have only 1 personality. For a simple game like Port Royal, it is easy to imagine a CAI, but it would in every game play by the same ruleset.
From a designer’s perspective it would be cool to design different CAI and pit them against each other, just to see what would be the strongest one. It also saves a lot of work on the part of playtesters 😉
Of course some games cannot be programmed on a board, that would be too complex, for example I cannot imagine a CAI for 1830 (because of the bidding) or even Forbidden Desert (because of the navigator). A CAI does not need to be a player killer, just provide some sort of challenge and stick to its essence, that is, a surrogate human.
Hi, interesting comments. In my defense, Tiny Epic Galaxies and Patchwork didn’t exist when the article was first written (late October 2014). The Agricola game came out in 2012, but then, this wasn’t really meant to be a survey of all the AI-affected games out there; just the ones I’d played. I was just looking at a sample to get the general patterns of AI games available. I’m sure there is much left to be written on the subject; I barely scratched the surface. Thanks again for your comments.
Euhm , please don’t get me wrong and don’t take this as some kind of personal attack, but how does a mere rule set (whether on the back of a closed card or in a rules book) suddenly become ‘AI’ though? Yes, even if that rule set simply defines NPC interactions, I wouldn’t call that AI, unless there is actually decision making at play. And for that you’d need something more than a simple condition. It means one decision is weight against another, intelligently. Hence artificial intelligence. I can’t say I see that in boardgames.
“An AI should, at some level, be responding to the player’s actions”
No, an AI should make a decision where making a decision would serve a real purpose other than simply being a single condition being met or the rules set dictating what happens next.
“In one game design (unpublished) I circumvented the need for a creature AI merely by letting the players move the monsters toward the opponents or their own heroes toward the monsters. I can’t really call this an AI; the players are the active intelligence in such a scheme. But it was a workaround to avoid complicated AI monster movement and action rules. This is a fairly common workaround in game AI; if you really want it to be intelligent, let the player on your right act as your nemesis on your turn.”
It might be a common solution to some extent, but it’s not very convincing behaviour at that. Sometimes a random behaviour (in boardgame context that should require a dice roll or card being turned, which is luck-based) looks more convincing than a monster that only homes in on the player, based upon location or movement of said player.
“One creature in Wiz-War depends on players entering its line-of-sight (Fire Imp – old edition) to trigger its own action. This is AI in board games at its simplest.”
Is it really though? I still wouldn’t consider such a condition AI either. It’s like stepping in a square that teleports or poisons the player. It’s a simple condition. There’s no decision at play. The Imp will always act within range and there’s probably something somewhat inevitable about meeting it, possibly involving luck to bypass it (it can be a closed card that only reveals the creature when the player is near, it might be a square with a predetermined condition etc.). There’s no difference with the boulder running down a set path to be honest.
“You move, they follow” doesn’t quite qualify as AI either. Think of boardgames that have a boulder roll down depending on the player triggering it. It’s a condition based reaction, but there’s no decision even required.
“Can you imagine a game of Berserker where the Berserker is an AI that draws a new card each turn, and each card is a set of additional reactive instructions?”
The simple reason why that isn’t as appealing (at least for commercial boardgames) as it may sound has to do with how those types of games are generally played with a group of friends. Honestly, it would be super easy to allow for a non-human player to act upon dice rolls and/or cards, but you’d ultimately just end up playing two characters as a player, right? There’s a reason why most boardgames (think Monopoly’s ‘bank’, which doesn’t have to be a player for what it does right) are basically non-active game elements, but game rules still dictate it plays a role. There’s never a decision involved though. It has to do with how you’re playing with or against your friends, not against ‘the board’ other than it’s environment or cards you come across.
So yeah, consider me one of those programmers who cringe at using AI for what is simply how advanced rules set work in boardgames, including card mechanics. As said, where are the decisions that require the ‘intelligence’ in ‘artificial intelligence’??? Heck, not many videogames even have proper AI these days anyway.
“AI in board games is obviously not needed to make a great game.”
You’re not going to like me saying this, but I would even say it would be somewhat pointless in a board game. The ‘surrogate player’ thing would be somewhat of a problem, unless you’re going to base any kind of NPC interactions upon closed cards that need perhaps a dice roll for a selection of possible actions. But…. that’s still not AI. It would be mostly just based upon a mechanic of luck. How exactly would such a thing be different from the boulder that is triggered upon entering a square?
Don’t consider this an attack at all, but I really don’t see why you called this AI. And yes, some board games can be played ‘singleplayer’ , but most if not literally all still don’t have actual AI where decisions are being weighted against each other. Heck, most games have very limited actions upon mere player presence at that. And frankly, for a board game, nothing more is really required.
I like videogames that were based upon board game rules sets, but honestly the traditional board game is a platform that’s mostly incompatible with true AI. ‘Autonomy’ of things like the boulder example, spike traps or moving walls aren’t part of any AI logic.
These are excellent comments! I understand and generally agree with what you’re saying. There really isn’t a good unique term to describe a game element that responds to varying input, and that’s the main reason for the use of the term Artificial Intelligence. For me, the term means “anything that delivers a variable response based on a variable input.” If that’s just a playing card that tells you condition A gives response A, or condition B gives response B, then that fills my definition. This is, as far as I’m concerned, the core of making a decision. You have a different definition for what an AI is, and I appreciate that. But even an AI that’s solving a Go game does the same thing, just at a much more complicated level. I won’t argue the point beyond that. Heck, most human decisions are just fixed responses to given inputs.
Hi Tom,
Great article! I do like the idea of using “AI” with directions based on card commands. I’m exploring multiple ideas currently that are variations of card RPG’s that require no GM to play, it’s a challenge but I really enjoy cooperative RPG’s.