Thursday, April 16, 2009

Game theory, simulated interaction, and unaided judgement for forecasting decisions in conflicts: Further evidence

Kesten C. Green

Green's article, published in the International Journal of Forcasting (2005), is a validity study that compares the accuracy of game theory, simulated interactions (role playing), and unaided judgment as methods in forecasting decisions in conflicts. It provides further evidence to support his earlier research, which indicated that expert game theorists' forecasts had less accuracy than those of novice (student) role players.

It is important to note here that these methods were tested in experiments involving specific conflict situations only. Not all game theorists recommend game theory for generating forecasts in specific situations.

Green's 2002 study reflected the accuracy of the three methods for forecasting conflict decisions. Participants in the study were asked to select the most likely decision (in 5 conflict situations) from a list. The results of the study did not bode well for the game theory experts (participants, % accuracy):
  • University Students (using unaided judgment), 27%
  • Game Theory Experts, 28%
  • University Students (using role playing/simulated interactions), 61%
Green's 2002 study drew some attention from his peers and game theorists, who commented that perhaps some conflict situations are more appropriate for use with game theory than others. The 2005 study utilized the same basic process, and was administered to participants representing the aforementioned three methods. It was directed at three (actual) conflicts, unrelated to the five used in the 2002 study:
  1. Personal Grievance. An employee of a New Zealand student association felt there was a disparity between her work and the level of pay she was receiving. Upon the administration's evaluation of her situation, it was discovered that she was being compensated above the top-level of pay in her salary bracket. While her manager did not consider cutting her salary, she did arrange for a mediator to set up a meeting between the parties, as it was obvious she would not be elligible for a pay increase in the near future. Participants were to choose from several decisions.
  2. Telco Takeover. This situation presented the participants with the conflict stemming from a corporate takeover battle between Alltel and CenturyTel. In 2001, executives from the much smaller CenturyTel company presented an offer that involved the sale of their mobile phone division to Alltel, which Alltel declined. Soon after, Alltel returned to CenturyTel with an offer to buy all of the company at 40% more than the share price. The board of CenturyTel fought to prevent an Alltel takeover. Participants were to forecast the execution of the deal and predict the basis of terms.
  3. Water Dispute. This conflict raged over Syria's and Iraq's claims to access of the Euphrates River in 1975. The flow of water into Iraq had been slowed following Syria's construction of a dam across the Euphrates to create a reservoir in Syria. Both sides mobilized for war, and Saudi Arabia called both parties together for mediation. Participants in the study were to forecast whether Syria would increase water flow from the dam or whether Iraq would declare war and bomb Syria's dam.
Since the findings of this research were to be compared with those of the 2002 study, Green took steps to evenly match the conflict situations with game theorists' and other decision makers' interests, i.e. types of organizations involved, nature of the disputes, and familiarity of situations.

The findings of this follow-up study are entirely consistent with the 2002 study. Game theory experts and unaided novices had a much lower forecast accuracy than novices who implemented role playing (simulated interactions) in predicting the eight conflict decisions.

No comments:

Post a Comment