Research map / Probabilistic and scientific modeling / Causal discovery
Causal Concept Explanations for Deep Neural Models
Joshua Rountree, Pulkit Verma, Oswin So, Chuchu Fan, Julie A Shah
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Introduces CCW-Net, a concept-based explanation method that applies mediation analysis to make concept explanations causal rather than merely correlational.
Problem
Concept-based explanations of neural models are typically correlational: a concept can fire alongside an output without causally driving it, producing misleading interpretability.
Approach
Uses mediation analysis to treat concepts as causal mediators between inputs and outputs, so explanations reflect whether a concept actually drives the output; implemented as a wrapper network (CCW-Net) around a deep model.
Closest papers by meaning
- FaCT: Faithful Concept Traces for Explaining Neural Network Decisions (NeurIPS 2025 · similarity 0.91)
- Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning (ICLR 2025 · similarity 0.89)
- Causally Reliable Concept Bottleneck Models (NeurIPS 2025 · similarity 0.88)
- You Can’t Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning (NeurIPS 2026 · similarity 0.87)
- Neural Causal Abstractions (AAAI 2024 · similarity 0.87)
- The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability? (NeurIPS 2025 · similarity 0.87)
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.
