Computational social science studies behavior through digital traces such as text, clicks, transactions, mobility patterns, and networks. These data enable new forms of measurement and high-frequency analysis, but they also create severe challenges related to selection, platform bias, hidden algorithms, and privacy.

Prediction and explanation are related but not identical. A model may forecast behavior accurately while remaining theoretically opaque, or it may be strongly explanatory yet weak in out-of-sample prediction. The frontier challenge is to build models that are both empirically useful and conceptually interpretable.

Active research problems include causal inference with text and images, fairness and bias in algorithmic decision systems, synthetic data generation, privacy-preserving methods, and integrating generative AI into research workflows without compromising validity or reproducibility.