# Medical AI Encodes a "Feeling of Error": Verifying Cancer Segmentation via Internal Concepts

arXiv: https://arxiv.org/abs/2609.08879 · HTML: https://mlsanity.com/paper/2609.08879/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Interpreting biomedical and science models, one of 13 directions in Mechanistic interpretability.
- Builds on 51 of its 73 references that are in the map.
- First posted on arXiv 8 Sep 2026.

## Placed in

- Mechanistic interpretability / [Interpreting biomedical and science models](https://mlsanity.com/mechinterp/interpreting-biomedical-and-science-models/)

## Builds on (most cited first, 30 of 51)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [XGBoost: A Scalable Tree Boosting System](https://mlsanity.com/paper/1603.02754/) (2016)
- [Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles](https://mlsanity.com/paper/1612.01474/) (2016)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks](https://mlsanity.com/paper/1610.02136/) (2016)
- [TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation](https://mlsanity.com/paper/2102.04306/) (2021)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)](https://mlsanity.com/paper/1711.11279/) (2017)
- [Energy-based Out-of-distribution Detection](https://mlsanity.com/paper/2010.03759/) (2020)
- [Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift](https://mlsanity.com/paper/1906.02530/) (2019)
- [Concept Bottleneck Models](https://mlsanity.com/paper/2007.04612/) (2020)
- [Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images](https://mlsanity.com/paper/2201.01266/) (2022)
- [The Medical Segmentation Decathlon](https://mlsanity.com/paper/2106.05735/) (2021)
- [Scaling and evaluating sparse autoencoders](https://mlsanity.com/paper/2406.04093/) (2024)
- [Causal Abstractions of Neural Networks](https://mlsanity.com/paper/2106.02997/) (2021)
- [Mechanistic Interpretability for AI Safety -- A Review](https://mlsanity.com/paper/2404.14082/) (2024)
- [Post-hoc Concept Bottleneck Models](https://mlsanity.com/paper/2205.15480/) (2022)
- [Aleatoric uncertainty estimation with test-time augmentation for medical image segmentation with convolutional neural networks](https://mlsanity.com/paper/1807.07356/) (2018)
- [SMIL: Multimodal Learning with Severely Missing Modality](https://mlsanity.com/paper/2103.05677/) (2021)
- [Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability](https://mlsanity.com/paper/2301.04709/) (2023)
- [Causality matters in medical imaging](https://mlsanity.com/paper/1912.08142/) (2019)
- [Are Multimodal Transformers Robust to Missing Modality?](https://mlsanity.com/paper/2204.05454/) (2022)
- [Open Problems in Mechanistic Interpretability](https://mlsanity.com/paper/2501.16496/) (2025)
- [Multi-modal Learning with Missing Modality via Shared-Specific Feature Modelling](https://mlsanity.com/paper/2307.14126/) (2023)
- [Disparities in Dermatology AI: Assessments Using Diverse Clinical Images](https://mlsanity.com/paper/2111.08006/) (2021)
- [BatchTopK Sparse Autoencoders](https://mlsanity.com/paper/2412.06410/) (2024)

## Abstract (arXiv)

Cancer segmentation models can fail silently, generating plausible but incorrect masks that risk missed findings or unnecessary biopsies. A critical question arises: Do AI models "know" when they are wrong, and if so, can we use the signal to predict their own failures? Humans do have a "Feeling of Error" (FOE): a spontaneous sense of unease that flags a potential error during thinking. We investigate whether cancer segmentation models exhibit an analogous internal signal. Unlike output-level cues (e.g., prediction confidence or uncertainty), which offer no insight into why a failure occurs and suffer from a sensitivity-quality tradeoff where high detection sensitivity could degrade overall segmentation quality. We instead propose to capture the model's FOE from its inner workings. Using mechanistic interpretability tools, specifically Sparse Autoencoders, we decompose internal neural activations into a dictionary of human-interpretable concepts and show that failure cases exhibit a distinct latent signature: fewer active concepts with lower activation magnitudes compared to successful segmentation. By training a classifier on these concept activations, we achieve accurate failure detection along with explanations for the model's mistakes. Experiments on prostate, pancreatic, and brain cancer segmentation demonstrate that our approach outperforms output-based methods in failure detection while preserving segmentation quality.