Research map / Learning theory and algorithms / Prediction and decision losses
The Utility of “Even if” Semifactual Explanation to Optimise Positive Outcomes
Eoin M. Kenny, Weipeng Fuzzy Huang
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 |
|---|---|---|---|
| Account for causal dependencies when computing gain (ablation) | S-GEN · Adult | gain: $0.055 \pm 0.001$ v. $0.063 \pm 0.001$ | Improved |
| S-GEN vs Karimi et al. on diversity | S-GEN · causal experiment | diversity: $0.84 \pm 0.09$ v. $0.43 \pm 0.03$ | Improved |
| S-GEN vs Domineguez et al. on diversity | S-GEN · causal experiment | diversity: $0.84 \pm 0.09$ v. $0.34 \pm 0.03$ | Improved |
| S-GEN vs Domineguez et al. on robustness | S-GEN · causal experiment | robustness success: $87\%$ success v. $54\% $ success | Improved |
| S-GEN vs Karimi et al. on robustness | S-GEN · causal experiment | robustness success: $7.2\%$ | Improved |
| Semifactual (S-GEN) instead of counterfactual (DiCE) explanation | S-GEN · User study, loan accepted | usefulness rating: 3.60$\pm$0.27 v. DiCE=2.33$\pm$0.34 | Improved |
| Semifactual (S-GEN) instead of counterfactual (DiCE) explanation | S-GEN · User study, loan rejected | usefulness rating: 2.6$\pm$0.32 v. DiCE=4.53$\pm$0.17 | Hurt |
| S-GEN vs DiCE* and PIECE* baselines on non-causal tests | S-GEN · non-causal, averaged datasets | gain, diversity, robustness etc. | Improved |
| Traditional counterfactual approach DiCE* for gain maximization | DiCE* · non-causal | gain | Hurt |
| S-GEN vs baselines on individual datasets | S-GEN · 16 individual dataset tests | multiple: significantly better on 11/16 tests | Mixed |
Closest papers by meaning
- Even-if Explanations: Formal Foundations, Priorities and Complexity (AAAI 2025 · similarity 0.87)
- Semi-factual Explanations in AI (AAAI 2024 · similarity 0.85)
- Generating Likely Counterfactuals Using Sum-Product Networks (ICLR 2025 · similarity 0.85)
- Counterfactual Identification Under Monotonicity Constraints (AAAI 2025 · similarity 0.85)
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching (NeurIPS 2025 · similarity 0.85)
- Collective Counterfactual Explanations: Balancing Individual Goals and Collective Dynamics (NeurIPS 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.
