AI Paper Atlas

Research map / Probabilistic and scientific modeling / Causal discovery

Counterfactually Comparing Abstaining Classifiers

Yo Joong Choe, Aditya Gangrade, Aaditya Ramdas

NeurIPS 2023 · Probabilistic and scientific modeling › Causal discovery · approach: Causal approaches to fairness · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use doubly robust estimator instead of plug-in with super learnercounterfactual score CI · simulated MARCI miscoverage: 0.91 -> 0.05Improved
use DR instead of IPW with random forest nuisancecounterfactual score CI · simulated MARCI miscoverage: 0.14 -> 0.05Improved
use DR instead of IPW with super learnercounterfactual score CI · simulated MARCI width: 0.12 -> 0.06Improved
IPW vs DR with super learnercounterfactual score CI · simulated MARCI miscoverage: 0.03 vs 0.05No clear effect
use linear/logistic nuisance estimators for DR (ablation)DR estimator · simulated MARCI miscoverage: 0.05 -> 1.00Hurt
use random forest instead of linear nuisance estimators for DR (ablation)DR estimator · simulated MARCI miscoverage: 1.00 -> 0.05Improved
use random forest instead of super learner for plug-in (ablation)plug-in estimator · simulated MARCI miscoverage: 0.91 -> 0.64Improved
increase sample size or absolute score differenceDR-CI hypothesis test · simulated MARpower: power reaches 1Improved
compare classifiers differing only in abstention threshold (0.8 vs 0.5)DR CI · CIFAR-100 scenario Icounterfactual score difference: (-0.005, 0.018)No clear effect
compare classifiers with different stochastic abstention rules (SR vs Gini)DR CI · CIFAR-100 scenario IIcounterfactual score difference: (-0.014, 0.008)No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy