Research map / Learning theory and algorithms / Prediction and decision losses
When Machine Learning Gets Personal: Evaluating Prediction and Explanation
Louisa Cornelis, Guillermo Bernardez, Haewon Jeong, Nina Miolane
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 |
|---|---|---|---|
| personalize with age and race (all population) | generic LOS classifier · MIMIC-III LOS classification | prediction vs incomprehensiveness benefit: 0.0026 prediction, -0.0077 incomprehensiveness | Mixed |
| personalize for NonWhite 18-45 group | generic LOS classifier · MIMIC-III LOS classification | prediction vs incomprehensiveness benefit: -0.0216 prediction, 0.0560 incomprehensiveness | Mixed |
| personalize for White 45+ group | generic LOS classifier · MIMIC-III LOS classification | prediction vs incomprehensiveness benefit: 0.0063 prediction, -0.0226 incomprehensiveness | Mixed |
| personalize for NonWhite 45+ group | generic LOS classifier · MIMIC-III LOS classification | prediction benefit: -0.0026 | Hurt |
| personalize with age and race (all population) | generic LOS regressor · MIMIC-III LOS regression | MSE benefit, prediction: 0.0051 | Improved |
| personalize with age and race (all population) | generic LOS regressor · MIMIC-III LOS regression | incomprehensiveness benefit: -0.0550 | Hurt |
| personalize with age and race (minimal group benefit) | generic LOS regressor · MIMIC-III LOS regression | sufficiency benefit: 0.1914 | Improved |
| raise improvement threshold epsilon | hypothesis test · BoP test validity | lower bound on error probability: lower bound decreases (null harder to reject) | Improved |
| use regression instead of classification at epsilon=0.002 | BoP hypothesis test · MIMIC-III LOS | lower bound on P_e: 40% -> below 5% for prediction | Improved |
Closest papers by meaning
- On the Epistemic Limits of Personalized Prediction (NeurIPS 2022 · similarity 0.91)
- CohEx: A Generalized Framework for Cohort Explanation (AAAI 2025 · similarity 0.83)
- Providing Fair Recourse over Plausible Groups (AAAI 2024 · similarity 0.82)
- Rethinking Personalization in Large Language Models at the Token Level (ICML 2026 · similarity 0.82)
- Position: Amazing Things Come From Having Many Good Models (ICML 2024 · similarity 0.82)
- DISCRET: Synthesizing Faithful Explanations For Treatment Effect Estimation (ICML 2024 · similarity 0.81)
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.
