From 2002 to 2003, there was a show on The Game Show Network called “Friend or Foe.”  The game was set up in such a way that there were two teams containing two people each.  The team that correctly answered the most trivia questions was the winning team, but they only got to keep their prize money if they passed the final test, which consisted of secretly choosing whether to split the money with their teammate or keep it all for themselves.  If they both decided to keep it all for themselves, then they went home with nothing.  But if they both chose to share, then they both left with 50% of the earnings.  The following table describes the four possible outcomes:

Or rather, putting the outcomes into a “Prisoner’s Dilemma”-type chart, like the ones we are used to seeing in class:

$$$ = total team earnings

Notice that the Nash Equilibrii in this game are (friend, foe) and (foe, friend). Also, there is no dominant strategy for this game (although, one can argue that choosing ‘foe’ could be a dominant strategy because it is a best response if the other player chooses ‘friend’ and it does not give a lower payoff if the other player chooses ‘foe’). Additionally, all strategies are Pareto Optimal except for foe-foe.

This is sort of a variation of the Hawk-Dove Game, where if the two players act like doves and cooperate, then they both get an equal share.  But if one player acts like a hawk and decides to “fight” for the prize, then they can get a higher earning given that the other player chooses to play dove.  If both players act like hawks, then they both fight for the prize and end up leaving with nothing.

This final test created an interesting dynamic for the game. In general, players chose to play ‘foe’ a little over half the time. There was a research paper written in 2003 by Felix Oberholzer-Gee of Harvard Business School, and Joel Waldfogel and Matthew White of Wharton Business School in which they aggregated the outcomes of two seasons’ worth of the game show (315 teams in total) and analyzed player decisions based on various factors. They found that, on average, the types of people most likely to cooperate were women and older players (they define “older players” as people over 27, but this includes many 40, 50, and 60-year-olds as well). These types of players were even more likely to cooperate if they were paired with another player of the same type, so for example, a female-female team would choose foe-foe 21.2% of the time, on average - the lowest rate for any of the groups measured - and 31.8% for friend-friend outcomes, the highest rate of any group. On the other hand, a team with two young players had the highest rate of choosing foe-foe (42.1%) and one of the lowest rates for choosing friend-friend (19.7%). Additionally, they found that of the 93 teams with one white team member and one black team member, only 10.8% chose friend-friend, the lowest rate of any group. Here is a summary of some of their findings, based on player types:

Another factor that was analyzed was the relation between a player’s contribution to the total prize money and their ‘foe’ rate.  Here, there appeared to be a change in strategy from the first season of the show to the second season.  In the first season, the players who contributed more to the total prize money were more likely to choose ‘foe’ and the players who contributed less were more likely to choose ‘friend’.  However, in the second season, this correlation was gone: “This behavior disappears in the second season, perhaps because player’s ability to predict their opponents’ play creates incentives that swamp the desire to reciprocate” (Oberholzer-Gee, Waldfogel, White, 27).  In other words, some players had learned by watching the first season that the players who contributed more were more likely to choose ‘foe’, so there was less of a desire for the less-contributing team member to choose ‘friend’.

Friend or Foe was an interesting game show experiment that showed how different types of people interact in a Prisoner’s Dilemma-type game.  But can we actually apply what we learned from a game show to our knowledge on human behavior?  Maybe or maybe not, but I think the aggregate data shows some interesting correlations and causes us to think about these types of interactions between different people.

Research Paper:

http://bpp.wharton.upenn.edu/waldfogj/pdfs/fof.pdf

Wikipedia article on “Friend or Foe” game show:

http://en.wikipedia.org/wiki/Friend_or_Foe%3F

Citation of the research paper:

Oberholzer-Gee, Felix, Joel Waldfogel, and Matthew White. Social Learning and Coordination in High-Stakes Games - Evidence from Friend or Foe. Business and Public Policy Department. 5 June 2003. Harvard Business School, The Wharton School. 27 Feb. 2009 <http://bpp.wharton.upenn.edu/waldfogj/pdfs/fof.pdf>.