In this post we’ll continue where we left off in the previous post trying to answer Jamey Stegmaier’s high-flying questions about artificial opponents in board games.
What if a game had feelings that could change over the course of the game, favoring various players and strategies?
In the game Hostage Negotiator you’re negotiating with a hostage taker, and there’s a track to represent the mood of the hostage taker and his attitude towards you. The lower this track is the more receptive he is to what you’re saying and the higher it is the more volatile and hard to influence he becomes. Mechanically the core of implementing this is that the track determines how many dice you roll when trying to influence the hostage taker, and while it’s a very simple mechanic it’s thematically powerful and can be seen as allowing the opponent as sometimes favoring the player and sometimes doing the opposite.
A prototype board showing the track in Hostage Negotiator that determines the mood of the hostage taker. Image credit: Tracy Baker.
Other than that I’m drawing a blank on examples of artificial opponents that have anything that could be interpreted as feelings, but I’d love to hear of other examples.
What if a game could make choices independent of the players?
I could see two approaches to this question: Either an electronic component like in XCOM: The Board Game where there’s an app that you use while playing that handles the opponent or some mechanical contraption that makes something happen and random or semi-random moments, which might fit well in a real time based game. Personally, I think that as soon as we include software in the equation then we’re venturing into the territory of computer games and out of the scope of this post.
Screenshot from the app for X-Com: The Board Game. Image credit: Fantasy Flight Games.
As for some mechanical contraption that makes random decisions independent of the player at random times, then we might come up with some cool stuff, but logically it wouldn’t really be different from having the player roll a die or draw a card, since it would still just be a mechanical process initiated by the player, just a more complicated one with a larger delay from initiation to result. From a player experience point of view, such a contraption might very well feel much more like the artificial opponent doing something independently of the player. I’m not aware of any games that does anything like this, though.
What if a game could evolve as you played it, or even over the course of many games? What if a game could remember (again, within one game or multiple games)?
I’ll deal with these two questions together, since I see them as two sides of the same coin. First I’ll note that evolving/remembering within one game is basically the same as the opponent reacting to you, which we discussed above, so I won’t spend much time on that here, I will start out with one idea for this though, since that will lead us to mechanics spanning multiple games.
Robinson Crusoe has a mechanic where a decision you make leads to a card being inserted into a deck and if that card is drawn your decision comes back to bite you in the ass. While this is not a personified artificial opponent, instead it’s, as I understand it (I haven’t tried the game), the environment responding and it’s limited to one-off effects, then it can easily be expanded if we realize that it’s actually just a very simple deck building mechanic, where instead of building a deck for the player we’re building a deck for the player’s opposition.
All the artificial In the Automa systems I’ve worked on have had a deck of cards that controlled the Automa’s actions, and it occurred to me that it might lead to interesting gameplay if that deck was being deck built during the game.
Part of this deck building could be in response to player actions as in Robinson Crusoe, so if you for example made a lot of attacks on the right flank then cards would added to the Automa deck that defended the right flank. This could then lead to strategies where you could make feints by making a series of small attacks on the right flank and then when the opponent deck had changed to deal with that you’d launch a full offensive on the left.
The idea of doing deck building for the opponent could easily be extended to work over a series of games, like if you attack a lot in one game, then a defensive card is added to the Automa deck, which over multiple games could feel like the Automa remembering that you play aggressively.
While originally “talking” to Jamey about these ideas it occurred to me that the it might work in a cool way in a time travelling game. Imagine that the Automa represents an opponent with a time machine that tries to achieve some goal and the players must stop it. Every time the players defeat the Automa the Automa uses its time machine to travel back in time and try again having learned from its defeat. This is represented by some cards are added to the Automa deck that counters the strategy the players used to win, so that they would have to come up with a new strategy for the next game.
Another approach would be to take the route of Risk Legacy, where the game components themselves are physically altered, so if for example you shot the Automa in the leg in one game you’d destroy its “run fast” card permanently, so that the Automa would be permanently influenced by the happenings of a previous play in a campaign.
A Risk Legacy board after 15 games. Image credit: Nushura.
What if a game could deceive you?
Deception... that's a tough one, or you could say that if you know all the cards in the Automa deck and you've drawn two thirds of them, you might then know that a specific action would be very likely, but you might still draw a less likely action. That could be seen as deception, and I think it's hard to get closer to deception than this.
Another way to achieve something like this is to have the Automa deck be created randomly form a much larger set of cards and then reshuffle it every time perhaps one third of the cards has been revealed. The random selection of cards in the deck would mean that the player wouldn’t know what cards where in there and the limited number of cards from the deck that he would see before each reshuffled would mean that he could end up with a skewed perception of what’s actually in the deck.
As an example let’s imagine that we have a pool of cards consisting of 50 defense cards and 50 attack cards. From that pool of cards a 30-card Automa deck is randomly constructed to have 10 defense and 20 attack cards. In our hypothetical game you reshuffle the Automa deck every time 10 cards have been drawn.
If by chance the 10 cards you got to see before the reshuffle were 7 defense and 3 attack cards, then you’d after the reshuffle be expected to draw mainly defense cards, while in reality you would be likely to draw mainly attack cards and thus the Automa could be said to have deceived you and launch a surprise offensive.
There’s of course no intention to deceive behind this, but that doesn’t matter as long as the player feels that he was deceived, but whether he would feel that way, I’m not sure.
What if a game could love you or hate you, or show gratitude or remorse?
This question is one of those that to me sounded completely like pie in the sky stuff, but once I thought more about it, it occurred to me that the effects for example hate and gratitude could be mimicked.
The idea of deck building an Automa deck over several games might be used to achieve effects like this. Imagine for example if you sacked it’s home city in one game that would add cards to the Automa deck that would make it try harder to sack your home city in the following games, so that it would seem to hate you and be vengeful.
I think that this idea could be used to simulate a range of emotion including gratitude where we’re could basically do the opposite of the above, where if you spared the Automa’s home city in a game where you had the upper hand cards could be added to its deck that made it more likely not to kick you while you’re lying down.
Combining these two things could lead to some interesting choices where you have to weigh the immediate benefit you get in one game by sacked the city, against the changed behavior of the Automa in the following games.
Hate and gratitude might work even better with in competitive games where there are more than two players (whether the extra players are biological or made of cardboard is not important for this discussion), because you could then gain advantages by getting an Automa to like you.
What if a game wasn’t aware that it was a game, or that it is any different than the players?
OK, Jamey, I’ve played along with you through all of the above questions, but on this one I’ll go back to being the guy who tethers high-flying thinkers and reuse my opening word: “That’s impossible”, “can’t be done”, “infeasible” .
A game AI that isn't aware that it's a game AI? This would require it to be aware in the first place and to be honest I think that's impossible to achieve in cardboard form. The question leads us to ask, what awareness and consciousness actually, which I think is a very interesting question that I enjoyed reading and thinking about back in my days as a student, but to be honest I think it’s a fool’s errand to try to attack this problem with cardboard instead of computers.
All your questions answered
So, that’s it. All your questions have been answered, Jamey .
As said during the introduction, I initially found Jamey’s ideas to be impossible to achieve, but as I dove into them I realized that they only seemed impossible, because I took them literally and considered whether cardboard can be made to hate or be vengeful, which is of course impossible.
My usual approach to making Automas is that they should not strive to be another player or do all that a human player does, instead they should be shells that mimic the effects of another player, so that the human player gets the experience of playing against someone else and not noticing that there’s nothing behind the shell. That approach can also be used to achieve many of Jamey’s goals: The Automa doesn’t need to actually hate you and be vengeful, as long as it mimics the effects of that emotion. Once that rather realization hit it seemed to me that a lot of what had appeared to be impossible to achieve could actually be mimicked. Writing this it occurs to me that this seems to be the same realization that Alan Turing came to long before me, when he came up with his Turing test for artificial intelligence, where what’s going on behind the “curtain” is not important, instead it’s whether the AI feels like a human, when talking to it.
Before letting you go I’ll do a shameless plug for one of my previous blog post series: If you’re interested in cardboard AIs, you might find some use of the series of posts I did on my approach for making artificial opponents for multiplayer games, The Automa Approach I: A method for making solitaire modes for multiplayer games.
A blog about solitaire games and how to design them. I'm your host, Morten, co-designer of solo modes for Scythe, Gaia Project, Wingspan, Glen More II, and others.
25 May 2015
- [+] Dice rolls