Research map / Learning theory and algorithms / Prediction and decision losses
On the Epistemic Limits of Personalized Prediction
Lucas Monteiro Paes, Carol Xuan Long, Berk Ustun, Flavio Calmon
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 group attributes to a logistic regression (personalization) overall | logistic regression · UCI Adult auditing data, Total | error rate: 19.5% -> 19.3% | Improved |
| personalize with Sex x Race x Immigration attributes | logistic regression · UCI Adult auditing data, group Male, NW, R | error rate: 21.6% -> 24.4% | Hurt |
| personalize with Sex x Race x Immigration attributes | logistic regression · UCI Adult auditing data, group Male, W, R | error rate: 22.1% -> 19.3% | Improved |
| personalize with group attributes | logistic regression · UCI Adult training data, group Male, NW, R | error rate: 23.0% -> 25.0% | Hurt |
| personalize with group attributes | logistic regression · UCI Adult training data, Total | error rate: 20.1% -> 20.0% | No clear effect |
| personalize with group attributes | logistic regression · UCI Adult auditing data, group Female, NW, R | gain from personalization: 1.1% -> -1.2% | Hurt |
| increase number of binary group attributes from 18 to 19 | BoP hypothesis test · n=8x10^9, epsilon=0.01 | minimax probability of error lower bound: 13% -> 97% | Hurt |
| increase number of samples per group beyond the number of groups | BoP hypothesis test · d=141, epsilon=0.01 | minimax probability of error | Improved |
| increase data points per group m | BoP estimator · semi-synthetic UCI Adult | MSE of BoP estimate | Improved |
| limit dataset to 50k samples with epsilon=0.01 | BoP hypothesis test · reliable auditing | max binary group attributes: at most 3 binary group attributes | Hurt |
Closest papers by meaning
- Providing Fair Recourse over Plausible Groups (AAAI 2024 · similarity 0.83)
- Test-Time Personalization: A Diagnostic Framework and Probabilistic Fix for Scaling Failures (NeurIPS 2026 · similarity 0.82)
- Scaling Laws for the Value of Individual Data Points in Machine Learning (ICML 2024 · similarity 0.82)
- Understanding the Statistical Accuracy-Communication Trade-off in Personalized Federated Learning with Minimax Guarantees (ICML 2025 · similarity 0.82)
- An Effective Theory of Bias Amplification (ICLR 2025 · similarity 0.82)
- Rethinking Personalization in Large Language Models at the Token Level (ICML 2026 · similarity 0.82)
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.
