Prediction and explanation in social systems.

Area of Science:

• Social Science
• Computational Social Science
• Predictive Modeling

Background:

• Traditional social science research historically prioritized interpretable causal mechanisms over predictive accuracy.
• The rise of computational methods in social science is challenging this long-standing bias.
• This shift necessitates addressing key issues to advance the field.

Purpose of the Study:

• To argue for the increasing importance of predictive accuracy in social science.
• To highlight three critical issues that need resolution in the era of computational social science.
• To propose a path towards more accurate, replicable, and useful social science research.

Main Methods:

• Conceptual analysis of traditional versus computational approaches in social science.
• Identification and discussion of challenges in evaluating predictive accuracy.
• Theoretical framing of the relationship between predictive accuracy and interpretability.

Main Results:

• Current practices for evaluating predictions in social science lack standardization.
• Theoretical limits to predictive accuracy in complex social systems are not well-characterized.
• Predictive accuracy and interpretability are often treated as substitutes, rather than complements.

Conclusions:

• Standardizing prediction evaluation is crucial for reliable social science.
• Characterizing theoretical limits will set realistic expectations for prediction and explanation.
• Integrating predictive accuracy and interpretability will enhance the quality and utility of social science findings.