Will you split or steal my Golden Balls?

The prisoner’s dilemma is a classic problem of cooperation. You’ll find lengthy and erudite academic discussions of it in many fields of academia, from evolution to behavioural economics.

I wouldn’t have pegged it as a staple for a television game show.

Golden Balls is a UK game show that always ends with a prisoner’s dilemma decision. Burton-Chellew and West took advantage of this to study the prisoner’s dilemma respond “in the wild.”

Golden Balls goes through several rounds; I have a whole episode in links below. Early rounds give players ample reason and opportunity to lie. The penultimate “test your psychic powers” segment has players randomly selecting balls, which determines the final total they will play for in the end. In the final round, two players are left, and they have to make a decision: split or steal?

Both players make independent decisions. If they both choose “split,” they the pot of money is split 50-50%. If one chooses “Steal,” that player gets all the money (twice what they’d get by splitting) and the other gets nothing. But if both pick “steal,” neither gets anything.

Here is an example of a final, high stakes decision, for “life-changing” money (as the presenter is fond of calling it):



Not all the stakes are quite so high, though:



By watching every episode of the show on DVD, Burton-Chellew and West found there was no overall pattern to splitting or stealing; it was almost exactly split down the middle. This was a bit of a surprise to me, as I’d understood that people tended to cooperate more.

Lots of factors were correlated with the decision, but one of the biggest ones was the amount of money at stake. With more money at stake, contestants were more likely to steal. People got greedy.

The earlier rounds, where players often had to lie, did affect the decisions of the contestants. There were hints of retribution. Lies about the amount of cash the player had made their opponent more likely to steal.

There are few cases where a player will admit to planning to steal the jackpot. Were there ways to tell who was being honest? One signal that seemed to be an honest signal was laughter. Another was reciprocal physical contact. But if only one player touched the other... it was more likely there would be stealing. Sadly, although both of these seemed to be somewhat honest signals, there was very little evidence that players picked up on these reasonably reliable cues from their opposite number!

One limitation of this study is that game show contestants are hardly selected at random. The producers may well have screened their contestants to get both strategies about equally represented in the game. This makes some of the other demographic information (i.e., that women split more) problematic. The authors address the non-random selection, but they focus more on the possibility of certain personalities wanting to be on TV, more than active choices by producers.

Further, game shows are a very strange situation. People may be inclined to think, “What happens on the game show, stays on the game show.”

Still, if you’re wondering if you can trust someone, asking yourself if whether he or she makes you laugh might be a good first question to ask.

Golden Balls episode

• Part 1
• Part 2
• Part 3
• Part 4
• Part 5

Reference

Burton-Chellew M, West S. 2012. Correlates of cooperation in a one-shot high-stakes televised Prisoners' Dilemma. PLoS ONE 7(4): e33344. DOI: 10.1371/journal.pone.0033344