Research map / Probabilistic and scientific modeling / Causal discovery
In the Eye of the Beholder: Robust Prediction with Causal User Modeling
Amir Feder, Guy Horowitz, Yoav Wald, Roi Reichart, Nir Rosenfeld
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 |
|---|---|---|---|
| Add conditional regularization and train a separate skeptic predictor rather than using no regularization (ablation) | Anti-causal user subclass model · Synthetic user subclass experiment, skeptic | OOD accuracy: 50.0 -> 75.6 | Improved |
| Add conditional regularization and train a separate believer predictor rather than using no regularization (ablation) | Anti-causal user subclass model · Synthetic user subclass experiment, believer | OOD accuracy: 75.1 -> 85.2 | Improved |
| Train separate predictors by user subclass instead of one pooled predictor (ablation) | Conditionally regularized anti-causal model · Synthetic user subclass experiment, OOD | accuracy: 64.57 -> 75.6 (skeptic); 78.13 -> 85.2 (believer) | Mixed |
| Train separate predictors by user subclass instead of one pooled predictor (ablation) | Conditionally regularized anti-causal model · Synthetic user subclass experiment, ID | accuracy: 70.03 -> 71.1 (skeptic); 78.88 -> 88.5 (believer) | Mixed |
| Use conditional regularization rather than no regularization (ablation) | Joint predictor trained on both anti-causal subclasses · Synthetic user subclass experiment, OOD | accuracy: 50.0 -> 64.57 (skeptic); 75.1 -> 78.13 (believer) | Improved |
| Partition mixed user populations and train distinct predictors using estimated types (ablation) | Anti-causal user population · Synthetic mixed-subpopulation experiment, OOD | accuracy | Improved |
| Use regularization matching the anti-causal user model instead of causal or non-causal regularization | RecFashion · Fashion product images, shifted color-choice correlations | accuracy | Improved |
| Use causal regularization for causal users instead of anti-causal or non-causal regularization | RecBeer · RateBeer, years with significant taste shifts | accuracy | Improved |
Closest papers by meaning
- Anytime User Engagement Prediction in Information Cascades for Arbitrary Observation Periods (AAAI 2023 · similarity 0.86)
- Causal Representation Learning for Generalisable Recommendation (NeurIPS 2026 · similarity 0.84)
- Robust Graph Based Social Recommendation Through Contrastive Multi-View Learning (AAAI 2025 · similarity 0.84)
- Identifiability Matters: Revealing the Hidden Recoverable Condition in Unbiased Learning to Rank (ICML 2024 · similarity 0.84)
- A Recipe for Causal Graph Regression: Confounding Effects Revisited (ICML 2025 · similarity 0.84)
- BayesCNS: A Unified Bayesian Approach to Address Cold Start and Non-Stationarity in Search Systems at Scale (AAAI 2025 · similarity 0.84)
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.
