[Paper Review] Decisions and Performance Under Bounded Rationality: A Computational Benchmarking Approach

Dainis Zēgners, Uwe Sunde+1 more|arXiv (Cornell University)|May 26, 2020

Auction Theory and Applications4 citations

TL;DR

This paper introduces a computational benchmark of cognitively bounded rationality using chess engines to evaluate human decision-making in professional chess players. It finds that faster decisions, while more deviant from the benchmark, are associated with better performance, suggesting that intuition and experience enable superior real-time judgment under constraints.

ABSTRACT

This paper presents a novel approach to analyze human decision-making that involves comparing the behavior of professional chess players relative to a computational benchmark of cognitively bounded rationality. This benchmark is constructed using algorithms of modern chess engines and allows investigating behavior at the level of individual move-by-move observations, thus representing a natural benchmark for computationally bounded optimization. The analysis delivers novel insights by isolating deviations from this benchmark of bounded rationality as well as their causes and consequences for performance. The findings document the existence of several distinct dimensions of behavioral deviations, which are related to asymmetric positional evaluation in terms of losses and gains, time pressure, fatigue, and complexity. The results also document that deviations from the benchmark do not necessarily entail worse performance. Faster decisions are associated with more frequent deviations from the benchmark, yet they are also associated with better performance. The findings are consistent with an important influence of intuition and experience, thereby shedding new light on the recent debate about computational rationality in cognitive processes.

Motivation & Objective

• To address the lack of a consensus benchmark for bounded rationality in human decision-making by constructing a computationally constrained alternative to perfect rationality.
• To isolate behavioral deviations from bounded rationality in a real-world, high-complexity environment—professional chess—using algorithmic benchmarks.
• To investigate the causes and performance consequences of such deviations, particularly under time pressure, fatigue, and positional asymmetry.
• To reconcile conflicting views on heuristics and biases by assessing whether deviations from a bounded benchmark impair or enhance decision quality.
• To provide a methodological framework applicable beyond chess for analyzing bounded rationality in complex decision contexts.

Proposed method

• Construct a benchmark of cognitively bounded rationality using the move-selection algorithms of modern chess engines, which mirror human computational limits.
• Compare individual human moves by professional chess players against the engine’s optimal moves under identical time and position constraints.
• Use move-by-move data to isolate deviations from the benchmark and analyze their drivers using regression models with controls for time pressure, position quality, fatigue, and game phase.
• Model decision time as a key variable to assess trade-offs between deliberation depth and performance outcomes.
• Incorporate psychological and situational factors (e.g., being in a winning or losing position, remaining time) as predictors of deviation likelihood.
• Assess performance consequences by linking deviation patterns to game outcomes and move quality, using statistical significance testing (p < 0.05, p < 0.01).

Experimental results

Research questions

• RQ1 What behavioral deviations from a computationally bounded rationality benchmark occur in professional chess players, and what factors drive them?
• RQ2 How do time pressure, fatigue, and positional advantage or disadvantage influence the likelihood of deviating from the bounded rationality benchmark?
• RQ3 Do deviations from the bounded rationality benchmark lead to worse or better performance in chess?
• RQ4 Is there a performance advantage associated with faster decisions, even when they deviate more from the benchmark?
• RQ5 To what extent do intuition and experience mediate the relationship between bounded rationality and decision quality in expert performance?

Key findings

• Faster decisions are associated with more frequent deviations from the bounded rationality benchmark, yet they are also linked to better performance outcomes.
• Deviations from the benchmark are systematically driven by time pressure, fatigue, positional imbalance (gains vs. losses), and game complexity.
• Longer deliberation time counteracts the influence of being in a worse position or under time pressure, reducing the likelihood of deviation.
• Extended decision time amplifies the negative performance effects of time pressure and the positive effects of being in a worse position.
• Prolonged deliberation increases the risk of errors due to fatigue, worsening performance consequences of deviations.
• The results suggest that expert intuition and experience enable superior real-time assessment, allowing faster decisions to outperform the engine benchmark despite higher deviation.