# Non-identifiability of Explanations from Model Behavior in Deep Networks of Image Authenticity Judgments

arXiv: https://arxiv.org/abs/2604.07254 · HTML: https://mlsanity.com/paper/2604.07254/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 43 references that are in the map.
- First posted on arXiv 8 Apr 2026.

## Builds on (most cited first, 22 of 22)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [Densely Connected Convolutional Networks](https://mlsanity.com/paper/1608.06993/) (2016)
- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- ["Why Should I Trust You?": Explaining the Predictions of Any Classifier](https://mlsanity.com/paper/1602.04938/) (2016)
- [Visualizing and Understanding Convolutional Networks](https://mlsanity.com/paper/1311.2901/) (2013)
- [Barlow Twins: Self-Supervised Learning via Redundancy Reduction](https://mlsanity.com/paper/2103.03230/) (2021)
- [ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness](https://mlsanity.com/paper/1811.12231/) (2018)
- [Sanity Checks for Saliency Maps](https://mlsanity.com/paper/1810.03292/) (2018)
- [Getting aligned on representational alignment](https://mlsanity.com/paper/2310.13018/) (2023)
- [AIGCIQA2023: A Large-scale Image Quality Assessment Database for AI Generated Images: from the Perspectives of Quality, Authenticity and Correspondence](https://mlsanity.com/paper/2307.00211/) (2023)
- [Decoding Brain Representations by Multimodal Learning of Neural Activity and Visual Features](https://mlsanity.com/paper/1810.10974/) (2018)
- [Deepfake Detection: A Comprehensive Survey from the Reliability Perspective](https://mlsanity.com/paper/2211.10881/) (2022)
- [Neither hype nor gloom do DNNs justice](https://mlsanity.com/paper/2312.05355/) (2023)
- [Quality Assessment in the Era of Large Models: A Survey](https://mlsanity.com/paper/2409.00031/) (2024)
- [An Empirical Evaluation of the Rashomon Effect in Explainable Machine Learning](https://mlsanity.com/paper/2306.15786/) (2023)
- [Agree to Disagree: When Deep Learning Models With Identical Architectures Produce Distinct Explanations](https://mlsanity.com/paper/2105.06791/) (2021)
- [CLIP-AGIQA: Boosting the Performance of AI-Generated Image Quality Assessment with CLIP](https://mlsanity.com/paper/2408.15098/) (2024)
- [ImageNet-trained CNNs are not biased towards texture: Revisiting feature reliance through controlled suppression](https://mlsanity.com/paper/2509.20234/) (2025)
- [The Value of AI Guidance in Human Examination of Synthetically-Generated Faces](https://mlsanity.com/paper/2208.10544/) (2022)
- [Cross-Model Consensus of Explanations and Beyond for Image Classification Models: An Empirical Study](https://mlsanity.com/paper/2109.00707/) (2021)

## Abstract (arXiv)

Deep neural networks can predict human judgments, but this does not imply that they rely on human-like information or reveal the cues underlying those judgments. Prior work has addressed this issue using attribution heatmaps, but their explanatory value in itself depends on robustness. Here we tested the robustness of such explanations by evaluating whether models that predict human authenticity ratings also produce consistent explanations within and across architectures. We fit lightweight regression heads to multiple frozen pretrained vision models and generated attribution maps using Grad-CAM, LIME, and multiscale pixel masking. Several architectures predicted ratings well, reaching about 80% of the noise ceiling. VGG models achieved this by tracking image quality rather than authenticity-specific variance, limiting the relevance of their attributions. Among the remaining models, attribution maps were generally stable across random seeds within an architecture, especially for EfficientNetB3 and Barlow Twins, and consistency was higher for images judged as more authentic. Crucially, agreement in attribution across architectures was weak even when predictive performance was similar. To address this, we combined models in ensembles, which improved prediction of human authenticity judgments and enabled image-level attribution via pixel masking. We conclude that while deep networks can predict human authenticity judgments well, they do not produce identifiable explanations for those judgments. More broadly, our findings suggest that post hoc explanations from successful models of behavior should be treated as weak evidence for cognitive mechanism.