Research map / Probabilistic and scientific modeling / Causal discovery
Generative causal explanations of black-box classifiers
Matthew O'Shaughnessy, Gregory Canal, Marissa Connor, Christopher Rozell, Mark Davenport
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 |
|---|---|---|---|
| Use the learned causal factor α to vary generated inputs | CNN trained on MNIST 3 and 8 · MNIST 3 and 8 | classifier output | Improved |
| Vary noncausal factors β instead of causal factors | CNN trained on MNIST 3 and 8 · MNIST 3 and 8 | classifier output | No clear effect |
| Replace the joint unconditional causal objective with the independent unconditional objective C_iu (ablation) | Linear-Gaussian explainer for the and classifier · Synthetic and classifier | causal objective | Hurt |
| Increase λ to give data fidelity more weight (ablation) | Linear-Gaussian explainer for the and classifier · Synthetic and classifier | data representation | Improved |
| Increase λ to give data fidelity more weight (ablation) | Linear-Gaussian explainer for the and classifier · Synthetic and classifier | causal alignment | Hurt |
| Use the joint conditional objective C_jc instead of the joint unconditional objective (ablation) | Linear-Gaussian explainer for the and classifier · Synthetic and classifier | causal objective | Mixed |
Closest papers by meaning
- Learning Discrete Concepts in Latent Hierarchical Models (NeurIPS 2024 · similarity 0.86)
- Generative Classifiers Avoid Shortcut Solutions (ICLR 2025 · similarity 0.86)
- Object centric architectures enable efficient causal representation learning (ICLR 2024 · similarity 0.86)
- Neural Causal Abstractions (AAAI 2024 · similarity 0.85)
- Measure-Theoretic Anti-Causal Representation Learning (NeurIPS 2025 · similarity 0.85)
- Toward Causal Generative Modeling: From Representation to Generation (AAAI 2025 · 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.
