We propose a new estimator to deal with confounders missing at random, in observational causal inference settings.

We decompose bias in synthetic control methods with missing control-unit data and develop a sensitivity tool for bias calibration.

We conduct a large-scale experiment to measure ideological biases in major commercial large language models.

Using an experimental design that mimics the YouTube interface, we demonstrate that presenting people with more partisan video recommendations has no detectable polarizing effects on users’ attitudes in the short term.

We propose a new method for measuring public opinion in sensitive survey contexts and validate it using data from mainland China and Hong Kong.