causal inference
The process of determining whether and how much one factor (e.g., a drug) actually causes a given effect, as opposed to mere correlation.
As defined in Jacob Stegenga on Medical Nihilism
Where it comes up · 5
- Susan Athey on Machine Learning, Big Data, and Causation7 claims
- Jacob Stegenga on Medical Nihilism2 claims
- Tim Cook’s Dashboard1 claim
- 164 - Cause Effect1 claim
- 232 | Amy Finkelstein on Adverse Selection and Hidden Information1 claim
Most important claims · 10 of 12Search all claims for “causal inference” →
- Empirical economics has undergone a 'credibility revolution' (Angrist's term) emphasizing isolating causal effects from spurious correlation, with the randomized experiment as one of the most valuable tools — applying medicine's randomized-trial logic to policy, as in the 2008 Oregon Medicaid lottery, where the state randomly allocated Medicaid to some low-income uninsured adults and not others to measure true causal effects.232 | Amy Finkelstein on Adverse Selection and Hidden Information
- For causal inference, you must distinguish the treatment variable (the minimum wage, the drug) from the control covariates and model them differently: apply machine learning to flexibly select controls, but give a causal interpretation only to the treatment variable, since data-mined controls are uninterpretable proxies that capture correlated information parsimoniously.Susan Athey on Machine Learning, Big Data, and Causation
- To give machines the ability to reason about cause and effect, we need an algebra that is asymmetric, because the equations of physics are symmetric (X causes Y to the same degree Y causes X), and only the human-like notion of causal asymmetry filters those equations to prevent absurd inferences like changing the weather by moving a barometer.164 - Cause Effect
- Machine learning does not solve any of the fundamental problems of distinguishing correlation from causation, because econometric identification theorems already assume infinite data; without assumptions like parallel time trends, you cannot identify the minimum wage effect, and changing the estimation technique does not change these fundamentals of causal identification.Susan Athey on Machine Learning, Big Data, and Causation
- First-person reports that a drug 'worked for me' should be approached with huge cautionary skepticism for three reasons that work together: the natural course of disease (symptoms fluctuate or naturally decline, and people seek treatment at their worst, so regression to the mean alone produces apparent improvement), the placebo effect, and confirmation bias.Jacob Stegenga on Medical Nihilism
- In a predictive model, the marginal effect of any component should never be given a causal interpretation; political sentiment correlating with employment does not mean it causes employment, and drawing causal conclusions requires a designed model justifying the necessary assumptions.Susan Athey on Machine Learning, Big Data, and Causation
- Medicine has historically been imperialistic in neglecting patient reports, and physicians should listen carefully to and respect patients—but when it comes to causal inference about whether a drug had a given effect, strict evidential standards must be maintained.Jacob Stegenga on Medical Nihilism
- For the sake of scientific progress, smoother organizations, and rapid innovation, every group—academic or business—tends to narrow its focus and get really good at a few things while ignoring others: the machine learning community systematically ignored causal inference and got excellent at prediction, while economists got good at small-data causal inference but weaker prediction.Susan Athey on Machine Learning, Big Data, and Causation
- For the sake of progress and smoother functioning, every group—academic or business—systematically narrows its focus, getting very good at a few things while ignoring others; the machine learning community ignored causal inference to become excellent at prediction, while economists got good at small-data causal inference, leaving the gap of big-data causal inference where Athey works.Susan Athey on Machine Learning, Big Data, and Causation
- Athey envisions a future of applied work where researchers are more honest by delegating the model-selection component they were embarrassed about to a machine and disclosing how it works—since it is easier to correct for a transparent algorithm than for arbitrary human selection among spreadsheet columns—freeing human energy for the genuinely hard causal-inference and robustness work.Susan Athey on Machine Learning, Big Data, and Causation