I’ve been failing to get around to doing this for a month now. London Meetup report for the Schelling Point Game.
This was based on my post, Schelling Point Strategy Training, which was in turn based on an actual attempt I made to coordinate with someone on picking a film from a miscellaneous selection of DVDs.
To give people a taste of this process before moving onto the game itself, I’d actually brought along two selections of DVDs, one with unambiguously alphabetisable titles, and one containing two films whose titles began with numerals and ‘The’ respectively. Passing these selections of DVDs round the table in a bag, participants had to coordinate on a film. People tried to coordinate on an assortment of criteria including title, colour of the box, and popularity or genre of film.
The first of two main insights from this meetup was made at this point: people will argue at length over the most appropriate procedure for selecting a Schelling Point. It’s not just that your own procedure seems obvious, but other people’s procedures seem inappropriate or even silly. This seems like an important observation when it comes to coordinating with other humans.
The game itself went as follows: two teams of five people were each given ten minutes to (a) come up with a set of objects that would be difficult for the other team to coordinate on, and (b) come up with a strategy for coordinating on arbitrary sets of objects given to them by the other team. Some constraints were in place: the objects were written on index cards (to avoid simple ordering); the objects had to form some sort of natural set; the objects had to be distinct, and distinguishable from one another by the team presenting them. One point would be awarded if an outright majority of the team managed to coordinate, and three points would be awarded for total coordination in a team.
We ran this for two rounds. In the first round, Team A presented “representations of the letter ‘a’”, with minor variations on font, serifs, etc. Team B presented Kanji characters. Both teams settled on a choice procedure involving the object’s position on the index card. In the second round, Team A presented “Scribbles” (literally scribbling over the card, and distinguishing the scribbles with post-hoc identified idiosyncracies of each scribble), and Team B presented 2x2 grid combinations of the addition and multiplication symbols. Each team achieved partial coordination on both rounds.
At this point, the second insight from this meetup was apparent: artificial coordination problems have a “presentation layer” that’s vulnerable to hacking. The index cards were an attempt to circumvent an obvious decision procedure in the “presentation layer” of a list (pick the first on the list), but they just present a new type of presentation layer, and the game becomes about finding an choice procedure for arbitrary objects presented in that manner. This is not a characteristic of real-world coordination problems. We thought of various ways of constructing the game that would do away with this, but they were technically cumbersome for a game in a pub.
The group was split as to whether to continue with the game, so we abandoned it for various other coordination experiments involving writing stuff on index cards. We then played The Resistance, primed up to the eyeballs with thoughts of coodination strategies.
In summary:
People like to argue about coordination strategies
When played with smart people, this game rapidly degenerates into hacking how the options are presented rather than coordinating on the options themselves
It seems like you might need a Schelling meta-point as an injunction: no meta-gaming, or Munchkining, the Schelling point game. This could be important because lessons about coordination problems, and how to avoid them, seem valuable to people who attend meetups to ostensibly learn such lessons, and this is helped by not creating additional coordination problems.
That is, unless, the meetup group actually wants to learn about how they might want to act in peculiar game-theoretic scenarios, where players have information and signaling powers they wouldn’t normally have, in which case, don’t mind me.
“No abusing the rules” probably only works if people can coordinate successfully on “the spirit of the rules”.
I think one direction to explore is to have a games master picking sets that are easy to define (at least roughly), but hard to enumerate. Things like “locations in New York”, “subsets of the integers”, “nonempty finite subsets of the irrational numbers”, “letters in non-Roman alphabets”, “man-made satellites currently orbiting Earth”, “models of jet plane”, “movies released in the 1980s”. Then teams compete to coordinate on the same sets, instead of presenting sets to each other.
You need the GM because problems can be arbitrarily complicated (“{locations in NY} X {subsets of the integers} X …”). I’m not sure how ambiguous-membership would be handled. My first thought was that if everybody in the team agrees that something is in the set, it counts; but you need to be able to disqualify unambiguously-wrong answers, or everybody just agrees to answer “the information desk in Grand Central Station at noon” regardless of the question. I suspect you can just allow the GM to veto such answers on discretion.
I’ve been failing to get around to doing this for a month now. London Meetup report for the Schelling Point Game.
This was based on my post, Schelling Point Strategy Training, which was in turn based on an actual attempt I made to coordinate with someone on picking a film from a miscellaneous selection of DVDs.
To give people a taste of this process before moving onto the game itself, I’d actually brought along two selections of DVDs, one with unambiguously alphabetisable titles, and one containing two films whose titles began with numerals and ‘The’ respectively. Passing these selections of DVDs round the table in a bag, participants had to coordinate on a film. People tried to coordinate on an assortment of criteria including title, colour of the box, and popularity or genre of film.
The first of two main insights from this meetup was made at this point: people will argue at length over the most appropriate procedure for selecting a Schelling Point. It’s not just that your own procedure seems obvious, but other people’s procedures seem inappropriate or even silly. This seems like an important observation when it comes to coordinating with other humans.
The game itself went as follows: two teams of five people were each given ten minutes to (a) come up with a set of objects that would be difficult for the other team to coordinate on, and (b) come up with a strategy for coordinating on arbitrary sets of objects given to them by the other team. Some constraints were in place: the objects were written on index cards (to avoid simple ordering); the objects had to form some sort of natural set; the objects had to be distinct, and distinguishable from one another by the team presenting them. One point would be awarded if an outright majority of the team managed to coordinate, and three points would be awarded for total coordination in a team.
We ran this for two rounds. In the first round, Team A presented “representations of the letter ‘a’”, with minor variations on font, serifs, etc. Team B presented Kanji characters. Both teams settled on a choice procedure involving the object’s position on the index card. In the second round, Team A presented “Scribbles” (literally scribbling over the card, and distinguishing the scribbles with post-hoc identified idiosyncracies of each scribble), and Team B presented 2x2 grid combinations of the addition and multiplication symbols. Each team achieved partial coordination on both rounds.
At this point, the second insight from this meetup was apparent: artificial coordination problems have a “presentation layer” that’s vulnerable to hacking. The index cards were an attempt to circumvent an obvious decision procedure in the “presentation layer” of a list (pick the first on the list), but they just present a new type of presentation layer, and the game becomes about finding an choice procedure for arbitrary objects presented in that manner. This is not a characteristic of real-world coordination problems. We thought of various ways of constructing the game that would do away with this, but they were technically cumbersome for a game in a pub.
The group was split as to whether to continue with the game, so we abandoned it for various other coordination experiments involving writing stuff on index cards. We then played The Resistance, primed up to the eyeballs with thoughts of coodination strategies.
In summary:
People like to argue about coordination strategies
When played with smart people, this game rapidly degenerates into hacking how the options are presented rather than coordinating on the options themselves
I don’t look like a spy
It seems like you might need a Schelling meta-point as an injunction: no meta-gaming, or Munchkining, the Schelling point game. This could be important because lessons about coordination problems, and how to avoid them, seem valuable to people who attend meetups to ostensibly learn such lessons, and this is helped by not creating additional coordination problems.
That is, unless, the meetup group actually wants to learn about how they might want to act in peculiar game-theoretic scenarios, where players have information and signaling powers they wouldn’t normally have, in which case, don’t mind me.
“No abusing the rules” probably only works if people can coordinate successfully on “the spirit of the rules”.
I think one direction to explore is to have a games master picking sets that are easy to define (at least roughly), but hard to enumerate. Things like “locations in New York”, “subsets of the integers”, “nonempty finite subsets of the irrational numbers”, “letters in non-Roman alphabets”, “man-made satellites currently orbiting Earth”, “models of jet plane”, “movies released in the 1980s”. Then teams compete to coordinate on the same sets, instead of presenting sets to each other.
You need the GM because problems can be arbitrarily complicated (“{locations in NY} X {subsets of the integers} X …”). I’m not sure how ambiguous-membership would be handled. My first thought was that if everybody in the team agrees that something is in the set, it counts; but you need to be able to disqualify unambiguously-wrong answers, or everybody just agrees to answer “the information desk in Grand Central Station at noon” regardless of the question. I suspect you can just allow the GM to veto such answers on discretion.