From Attribution to Action: A Human-Centered Application of Activation Steering
On the map: Mechanistic interpretability → Activation steering methods
Abstract (arXiv)
Explainable AI (XAI) methods reveal which features influence model predictions, yet provide limited means for practitioners to act on these explanations. Activation steering of components identified via XAI offers a path toward actionable explanations, although its practical utility remains understudied. We introduce an interactive workflow combining SAE-based attribution with activation steering for instance-level analysis of concept usage in vision models, implemented as a web-based tool. Based on this workflow, we conduct semi-structured expert interviews (N=8) with debugging tasks on CLIP to investigate how practitioners reason about, trust, and apply activation steering. We find that steering enables a shift from inspection to intervention-based hypothesis testing (8/8 participants), with most grounding trust in observed model responses rather than explanation plausibility alone (6/8). Participants adopted systematic debugging strategies dominated by component suppression (7/8) and highlighted risks including ripple effects and limited generalization of instance-level corrections. Overall, activation steering renders interpretability more actionable while raising important considerations for safe and effective use.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Activation steering methods, one of 13 directions in Mechanistic interpretability. Builds on 16 of its 61 references that are in the map. First posted on arXiv 13 Apr 2026.
Builds on
16 of 61 references.
Closest ancestors · 1
Same direction, newest first
Other references · 15
- 2017Towards A Rigorous Science of Interpretable Machine Learningcited by 1,416
- 2022Toy Models of Superpositioncited by 971
- 2023Eliciting Latent Predictions from Transformers with the Tuned Lenscited by 623
- 2023Towards Best Practices of Activation Patching in Language Models: Metrics and Methodscited by 359
- 2020Opportunities and Challenges in Explainable Artificial Intelligence (XAI): A Surveycited by 254
- 2014Qualitatively characterizing neural network optimization problemscited by 188
- 2019Ablation Studies in Artificial Neural Networkscited by 60
- 2025Steering CLIP's vision transformer with sparse autoencoderscited by 38
- 2024Towards Unifying Interpretability and Control: Evaluation via Interventioncited by 32
- 2026Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Modelscited by 27
- 2025SCAM: A Real-World Typographic Robustness Evaluation for Multimodal Foundation Modelscited by 13
- 2025RelP: Faithful and Efficient Circuit Discovery in Language Models via Relevance Patchingcited by 13
3 more
- 2023Mind the Gap! Bridging Explainable Artificial Intelligence and Human Understanding with Luhmann's Functional Theory of Communicationcited by 9
- 2025From What to How: Attributing CLIP's Latent Components Reveals Unexpected Semantic Reliancecited by 6
- 2026Evaluating Actionability in Explainable AIcited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Embodied Interpretability: Linking Causal Understanding to Generalization in Vision-Language-Action Models
- 2025Towards Unified Attribution in Explainable AI, Data-Centric AI, and Mechanistic Interpretability
- 2026Steered LLM Activations are Non-Surjective
- 2026Position: Explainability Research Must Prioritize Foundations over Ad-hoc Methods
- 2025Steering Awareness: Detecting Activation Steering from Within
- 2026The Metagame of Interpretability and Meta-Attributions
- 2026ConceptSMILE: Auditing the Trustworthiness of Concept-Based Explainable AI
- 2025Toward universal steering and monitoring of AI models