TL;DR
This paper explores how people, particularly students, understand and adapt to strategic decision-making in "guessing games" through real-world experiments. By analyzing data from various groups playing variations of the "p-beauty contest," the authors reveal interesting patterns of learning, irrationality, and coordination, suggesting these games are valuable tools for teaching game theory and understanding human behavior.
What Problem Does This Paper Address?
Game theory often models players as perfectly rational agents aiming to maximize their utility. However, real-world decision-making is far more complex, influenced by bounded rationality, heuristics, and social factors. This paper investigates how individuals, especially those new to game theory concepts, actually behave in strategic situations. The core question is: how do different groups of players, from school children to adults, understand and adapt their strategies when faced with unfamiliar strategic interactions, and can these experiments be effectively integrated into educational settings?
Methods
The authors employ a series of four "guessing games" designed to illustrate different game theory concepts. These games were played by participants during scientific popular lectures, aiming for a more naturalistic setting than traditional laboratory experiments. Participants were asked to guess a number within a range (1-100), with winning conditions varying for each game. The games included:
- p-beauty contest: Guess a number closest to 2/3 of the average guess. This game is theoretically solvable through iterated elimination of dominated strategies, leading to a Nash equilibrium of 1.
- Two equilibrium game: Guess a number furthest from the average. This game highlights mixed strategies.
- Coordination with assurance: Guess the number chosen by the plurality, with ties broken by the lower number. This game emphasizes the importance of focal points for coordination.
- No equilibrium game: Guess the smallest unique number. This game is designed to have no clear equilibrium or rational strategy.
Data was collected using Google Forms, and analysis was performed using R for statistical visualization. The study contrasts behavior across different rounds of play and across various participant groups (e.g., different school types, adults).
Datasets
The paper presents data collected from participants playing the guessing games during scientific popular lectures between 2018 and 2019. The participants were categorized into several groups:
- Alternative humanitarian school
- Alternative mathematical school
- Adults (including a Facebook online group and adults attending a data science conference)
- Ordinary high school
- Mathematical schools (multiple instances)
- Ordinary school
- IASA KPI (BSc)
- NAUKMA (MSc)
- Tech School
This data, along with the R scripts used for analysis, is made publicly available in an open repository.
Key Insights and Ideas
- "Breaking the Rules" as Learning: Participants, especially high school students and adults, sometimes make "irrational" choices (e.g., numbers > 66 in the p-beauty contest) not just due to misunderstanding but potentially as a form of experimentation or to "cheat the test." This behavior is more pronounced in adults than in younger children or specialized math students.
- Group Learning is Fast: Despite individual variations, the average winning number across groups consistently decreases with each round in games like the p-beauty contest, indicating rapid group learning and an implicit competition to improve collective decisions.
- Intuitive Mixed Strategies: Participants, even without prior knowledge of game theory, exhibit a remarkable ability to play mixed strategies, as seen in the second game where choices of 1 and 100 were nearly equally distributed. This suggests that concepts like mixed equilibrium can be intuitively grasped.
- Focal Points Drive Coordination: In coordination games, players naturally gravitate towards focal points (like the number 1) to achieve a common goal, even when explicit communication is absent. The presence of a "lesser wins" tie-breaking rule reinforces this tendency.
- Educational Value: The experiments demonstrate that guessing games can be effective tools for teaching strategic thinking and game theory concepts in an engaging, hands-on manner, bridging the gap between theoretical models and real-world behavior.
Key Results
- In the p-beauty contest, winning numbers generally decreased significantly across rounds, with students showing better performance in reducing their choices than adults. However, a notable percentage of "irrational" choices (>66) persisted even in later rounds for some groups.
- The two equilibrium game showed participants choosing 1 and 100 in almost equal proportions, suggesting a natural tendency towards a mixed strategy equilibrium.
- The coordination game saw strong coordination around the number 1, as predicted by Schelling's focal point theory, with other numbers like 7, 50, and 100 also appearing as attempts at coordination.
- The no equilibrium game yielded varied and sometimes unexpected winning numbers, highlighting the difficulty of strategic decision-making in the absence of clear theoretical guidance.
Limitations and Future Work
Limitations: * The "weak sides" of the experimental setup are acknowledged, including the possibility of "garbage" decisions, participants talking to neighbors (especially kids), loss of concentration leading to random choices, and occasional initial misunderstanding of rules due to lecture conditions. * The authors note that while their setup is more realistic than laboratory experiments, it still has weaknesses that are "inevitable in realistic scenarios."
Future Work: * The paper suggests that these experiments reveal interesting behavior that is worth studying, even if not conducted under "pure" laboratory conditions. * The findings have multiple applications, including providing hands-on experience with strategic interactions for children, demonstrating the use of game theory experiments in education, understanding decision-making differences across groups, and replicating classical experiments in modern educational systems. * The authors plan to gather more data and formulate hypotheses for games without theoretically found equilibria, like the fourth game presented.
Summary updated