CBS Survivor, in group identity scoring, and a comparison to web search weighted voting
Andrew Hanson’s paper “The Sabermetrics of Survivor – The Role of In-Group Identity to Survival in Reality Television” investigates how the “in-group identity score” of a player relates to their final performance in the game. For those unfamiliar, Survivor is a reality TV show, which places contestants on an island, where they must compete in challenges. Losing teams will then have to vote off a player in a process known as tribal council, with the player receiving the most votes leaving the game. We define a player’s in group identity as follows: a player gains 3 points if they vote for a contestant that is eliminated in that round and 0 otherwise. A player loses 1 point if someone other than who is voted out votes for them. Because each player will need to attend a different number of tribal councils, we normalize in group identity scores by dividing by the number of attended tribal councils. The paper concludes that there is a direct correlation between a player’s success in the game and their in group identity score.
This concept is in many ways a human example which illustrates the intuition behind using hubs and authorities as a web search algorithm. In particular, look at voting for a player that is eliminated like voting for a popular site. Much like when we discussed weighting voters by the extent they hit popular sites, this metric for Survivor players “weights” their scores by the extent they hit popular targets. Other pieces of intuition about this process also hold: for example, if a player receives a vote, they lose a point, making them less likely to succeed in the game under the analysis of this metric. In other words, they have become a more popular target. Thus, we can see that players act as hubs (those who are good at determining who the targets are), and authorities (those that are the popular targets to get voted out), and hubs that successfully determine popular targets see their scores increase. Indeed, the success of the in group identity score of Survivor players is a way to present the idea behind search algorithms in a way that is easy to consume and anyone can understand intuitively.
Interestingly, Survivor has a vast number of applications of material from this course, perhaps because human social dynamics are one of the most inherent manifestations of network dynamics. Since this paper was written, CBS has added in a number of game theoretic elements to complicate social decision making. For example, players are randomly selected to play a game where they can risk their vote to potentially get a reward. If both players risk, both lose their votes and get nothing. If you choose to protect your vote, you will win or lose nothing regardless of what the other player does. But if you risk, while your opponent protects, you win the reward. This is precisely the prisoner’s dilemma and complicates the above analysis of voting records as a quantitative metric of a player’s success. In particular, these games potentially cause players to lose their ability to vote, which interferes with the core idea that being in the “in group” means a player will likely vote for the target and thereby gain 3 points. With this new construct, players very much in the “in” could very well not gain any points in a round. Does this have a significant impact on the idea of in group identity score as a success metric? Only time – and more seasons of Survivor – will tell.
https://openspaces.unk.edu/undergraduate-research-journal/vol21/iss1/6/