Guest Talk: Causal AI — Core Ideas and Intuition Behind Causal Data Analysis (RUS)
Date:
Guest lecture: an introduction to causal inference and causal data analysis.
The familiar mantra “correlation does not imply causation” is often repeated, but the follow-up question — what can we actually infer? — is rarely discussed.
This question has two sides. First, what does causation require — what conditions must hold for data to support causal conclusions? Second, what does correlation actually tell us — what inferences can (and cannot) be drawn from statistical associations, and is it really true that correlation “means nothing”?
The lecture explores why statistical dependence alone is insufficient for causal conclusions, and what typical misconceptions arise when interpreting data. We look at how different causal structures can produce identical observed correlations, and why without additional assumptions or interventions it is impossible to distinguish one story from another.
Through simple examples we cover the fundamental structures of causal relationships and the core intuition behind modern causal methods and models.