Recommendation as generalization: Using big data to evaluate cognitive models.

Area of Science:

• Cognitive Psychology
• Machine Learning
• Computational Social Science

Background:

• The increasing volume of online interaction data provides novel opportunities for psychological research.
• Large-scale datasets enable the development of more complex cognitive models than traditional laboratory settings allow.

Purpose of the Study:

• To evaluate the scalability and efficacy of established psychological generalization models using web-scale datasets.
• To demonstrate how cognitive models can enhance machine learning recommendation systems by capturing human judgment nuances.

Main Methods:

• Implementation of three popular psychological generalization models.
• Application of these models to two large-scale online datasets commonly used in recommendation system development.
• Comparison of psychological model performance against standard machine learning approaches.

Main Results:

• Psychological models are efficiently scalable to web-scale data.
• These models successfully captured trends in human judgments, outperforming standard recommendation systems in specific instances.
• Cognitive modeling insights can supplement predictive analytics in machine learning.

Conclusions:

• Internet-scale datasets present a valuable frontier for psychological research and cognitive modeling.
• Integrating cognitive modeling with machine learning offers a more comprehensive approach to understanding human behavior and improving predictive systems.