Article

A layer-resolved diagnostic identifies bias-driven decisions in deep neural networks

2025-09-22

Abstract excerpt

Modern AI systems can be accurate and confident, but this alone does not reveal whether a decision is well supported by the input. This creates a trust problem because confidence reports how decisive a model is, but not what supports that decisiveness. Here we show that neural-network confidence can be decomposed into input-dependent feature support and input-independent offset support. We formalize this decomposi...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: 93c6bde7-f24f-566d-ab20-b710ea568e2f
DOI: 10.1101/2025.09.16.676625

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

A layer-resolved diagnostic identifies bias-driven decisions in deep neural networksDOI 10.1101/2025.09.16.676625Read

Neurosymbolic AI for Safe and Trustworthy High-Stakes ApplicationsSemantic proximity Read

Learning Representations by Humans, for HumansSemantic proximity Read

Scalable and Interpretable Mixture of Experts Models in Machine Learning: Foundations, Applications, and ChallengesSemantic proximity Read

Predictive and Interpretable: Combining Artificial Neural Networks and Classic Cognitive Models to Understand Human Learning and Decision MakingSemantic proximity Read

Flow-by-Flow: Content-Judgment Bypass for Governing AI Output in High-Loss DomainsSemantic proximity Read

Using Deep Learning to Predict Human Decisions, and Cognitive Models to Explain Deep Learning ModelsSemantic proximity Read

Computation noise promotes cognitive resilience to adverse conditions during decision-makingSemantic proximity Read

What Makes Neural Networks Trainable? Invexity as a Structural Design Principle in AISemantic proximity Read

Abstraction, Validation, and Generalization for Explainable Artificial IntelligenceSemantic proximity Read

Sparse Autoencoders as an Interpretable Interface for LLM Development and ControlSemantic proximity Read

Neural networks reliability for feedback processing.Semantic proximity Read

Deceptive Bias Measurement in Deep Learning: Assessing Shortcut Reliance in TCGA Cancer ModelsSemantic proximity Read