Bayesian Mixer London

Please register for the event here: https://buytickets.at/bayesianmixer/2507848

Abstract: First rule of statistics: correlation is not causation. Although sometimes, just sometimes, it is. Judea Pearl thinks most of us stop at the first rule too soon. We’ll start with a festive puzzle: children who believe in Santa get more presents. Is that just a correlation, or could it be causal? And how would we ever know? In The Book of Why, Pearl argues that the answer isn’t in the data alone but in the assumptions, we draw around it. With the right causal diagram, some correlations really can be read as causes. We’ll look at confounders that fool us, colliders that create correlations out of nothing, adjustments that let a correlation tell the causal truth, and the dangers of adjusting for too much. Then we’ll climb Pearl’s Ladder of Causation all the way to counterfactuals, and end with some hope for where causal thinking might take us next.

About Laura: Laura Illingworth believes data points tell stories, but not always the ones they seem to. An epidemiologist by training, she is a Senior Data Scientist at TRI – The RBQM Experts. There she builds the risk indicators and central monitoring methods that help clinical trials spot problems early and leads her team’s adoption of AI-assisted development. Outside work she builds her own projects, including Fridge Door, an AI platform that organises family life by coordinating a team of agents arranged, naturally, in a DAG.