Research map / Probabilistic and scientific modeling / Treatment effects
Anticipating Performativity by Predicting from Predictions
Celestine Mendler-Dünner, Frances Ding, Yixin Wang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Include the deployed prediction as an input feature | Supervised learning meta-model · ACSIncome semi-synthetic test distribution; identifiable settings | extrapolation error | Improved |
| Include the deployed prediction as an input feature when predictions are deterministic and linear (ablation) | Supervised learning meta-model · ACSIncome semi-synthetic test distribution; non-identifiable setting | extrapolation error | Hurt |
| Add Gaussian noise to the deployed predictions (ablation) | Supervised learning meta-model · ACSIncome semi-synthetic test distribution | extrapolation error | Improved |
| Increase the complexity of the deployed predictor beyond the outcome function class (ablation) | Supervised learning meta-model · ACSIncome semi-synthetic test distribution | extrapolation error | Improved |
| Discretize the deployed predictions (ablation) | Supervised learning meta-model · ACSIncome semi-synthetic test distribution | accuracy | Improved |
| Increase the distance between training and test predictors (ablation) | Supervised learning meta-model with predictions as a feature · ACSIncome semi-synthetic test distribution; finite-data experiment | extrapolation error | Hurt |
| Increase training dataset size (ablation) | Supervised learning meta-model · ACSIncome semi-synthetic test distribution; identifiable settings | extrapolation error | Improved |
| Include predictions as a feature despite misspecification of the outcome model | Supervised learning meta-model · Census income, Census travel time, and Kaggle credit score | accuracy | Improved |
| Include predictions as a feature when performativity is very weak and the outcome model is misspecified (ablation) | Supervised learning meta-model · Census income, Census travel time, and Kaggle credit score | accuracy | Mixed |
| Use a less accurate training predictor by fitting it to noisy labels (ablation) | Supervised learning meta-model with predictions as a feature · Census income, Census travel time, and Kaggle credit score | accuracy | Improved |
Closest papers by meaning
- On the Actionability of Outcome Prediction (AAAI 2024 · similarity 0.87)
- Causal Inference out of Control: Estimating Performativity without Treatment Randomization (ICML 2024 · similarity 0.85)
- An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation (NeurIPS 2025 · similarity 0.85)
- Data Fusion for Partial Identification of Causal Effects (NeurIPS 2025 · similarity 0.85)
- Predictive Performance Comparison of Decision Policies Under Confounding (ICML 2024 · similarity 0.85)
- Do causal predictors generalize better to new domains? (NeurIPS 2024 · similarity 0.85)
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.
