An experimental investigation of the secretary problem: factors affecting sequential search behaviour
thesisposted on 2022-03-28, 20:40 authored by Yu Chin Hsiao
The underlying cause of people stop ping their search prior to the theoretical optimal solution in commonly found sequential search problems, often known as the secretary problem, is as yet undetermined. In the secretary problem, an immediate decision must be made from offers that are presented one at a time in a random sequence without recall. The factors that might cause the early stopping bias were examined in a series of experiments. In Experiment One, I experimentally investigated the effect of time search cost in a version of the secretary problem framed as a house-selling context and found that, if each additional search involved a pre-announced time delay, people shortened the search. In Experiment Two, I found that if there were a house-selling context, people searched more optimally than with no context, even without house information. Experiment Three extended the findings from Experiment Two, framed the secretary problem as either a house-selling or a secretary-hiring task, and found evidence suggesting that regret aversion was a cause of the early stopping bias. Experiment Four investigated group decision making and found that groups chose less optimal prices than the aggregate performance of individuals. In Experiment Five, I examined the effect of an incentive structure that rewarded only finding the optimal prices and found that people did indeed make more optimal decisions in this treatment. Overall, these findings suggest that when we make sequential search decisions in life, exhausting and exploring all offers is never optimal, time spent can never be regained, nor can the rejected offer. Taking note of the context and gathering information prior to making decisions does lead to better results. Settling for offers other than the best can save time and effort, and the earnings will not be much different than when only aiming for the best.