Modeling Decision-Making Behavior in a Double Auction Task

In the field of economics, traditional decision-making models are often built on a foundation of extensive assumptions that might not always hold true in practical situations. Alternatively, models that utilize adaptive learning without assumptions are capable of adjusting to varying levels of uncertainty, though this adaptability might come at the expense of efficiency. In scenarios where information is not fully available, players face dual challenges: structural uncertainties about fundamental environmental parameters, and strategic uncertainties regarding the actions of their competitors. To explore how individuals navigate these challenges, we employed a double auction format in various competitive settings to analyze the learning behaviors of buyers. We examined a variety of adaptive learning models, including reinforcement learning, directional learning, and belief learning. The analysis revealed that straightforward models of directional learning, which operate with few assumptions about the informational efficiency of market prices, most accurately depict actual human behavior. This observation supports the concept of bounded rationality, where individuals strive to optimize their chances of successful transactions by applying the simplest effective learning approach.