Predicting similarity judgments in intertemporal choice with machine learning.

Area of Science:

• Cognitive Science
• Behavioral Economics
• Computational Neuroscience

Background:

• Intertemporal choice models often use similarity judgments for reward amounts and time delays.
• The cognitive processes underlying these similarity judgments remain poorly understood.
• Existing models lack insight into the predictive factors of judgment formation.

Purpose of the Study:

• To investigate the factors predicting similarity judgments in intertemporal choice.
• To determine if decision tree algorithms can model both the outcomes and processes of these judgments.
• To explore the application of machine learning in understanding decision-making mechanisms.

Main Methods:

• Employed machine-learning algorithms to analyze factors influencing similarity judgments.
• Utilized decision tree models to assess prediction accuracy for judgment outcomes and response times.
• Examined the role of numerical differences and ratios in decision-making heuristics.

Main Results:

• Combining numerical differences and ratios of reward values effectively predicts similarity judgments.
• Decision tree structures accurately capture both the outcomes and the process of making these judgments.
• Response times were also predictable using these machine learning approaches.

Conclusions:

• Machine learning can model not only decision outcomes but also the underlying cognitive processes.
• Understanding the heuristics of similarity judgments offers a pathway to improving decision-making.
• This approach provides a novel method for dissecting complex cognitive judgments.