The Challenges of Hyperparameter Tuning for Accurate Causal Effect Estimation
Damian Machlanski, Spyridon Samothrakis, Paul Clarke · 2026
Read The Challenges of Hyperparameter Tuning for Accurate Causal Effect Estimation by Damian Machlanski, Spyridon Samothrakis, Paul Clarke (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13675359.txt.
Description
Damian Machlanski and two co-authors ask how much the tuning of settings matters when machine learning estimates cause and effect from observational data. Many causal estimators exist, and all need their settings tuned. For plain prediction, agreed scores like mean squared error make tuning easy. For causal tasks no agreed score exists, so workers tune by proxy measures. Model choice also stacks parts: the estimator, the learner, the settings, and the score.
The team ran a broad study combining common estimators, learners, and scores across four benchmark datasets. Tuning raised the chance of reaching top performance for average and individual effects using only common estimators. Scores also behaved unevenly across scenarios, working well in one and poorly in another. The authors conclude that tuning matters more than often assumed. They call for work on scores that judge causal models evenly everywhere.
Catalog record
| CREATOR | Damian Machlanski, Spyridon Samothrakis, Paul Clarke |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 76 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |