Year: 2025
Type: article
Abstract: Explainability in generative AI -- the ability to provide meaningful accounts of why a language model, image generator, ormultimodal system produced a particular output -- is substantially harder than explainability in discriminative AI.Discriminative models map inputs to class labels, and existing explainability methods (SHAP, LIME, GradCAM) attributethe prediction to input features. Generative models map inputs to complex, high-dimensional outputs where attribution toinput features is only one of several explanation needs: users also need to understand what training data influenced theoutput, why alternative outputs were not generated, and how the model's uncertainty manifests in the output space. Thisstudy evaluates nine explainability approaches for generative AI across four explanation target categories: inputattribution (token importance, attention), training data influence (data attribution, TracIn), output characterisation(uncertainty decomposition, counterfactual generation), and process explanation (chain-of-thought faithfulness, activationsteering). Evaluation covers six generative AI applications: scientific paper generation, legal document drafting, codegeneration, creative writing, image generation, and clinical report generation. A proposed Unified GenerativeExplainability Framework (UGEF) combines all four explanation target categories into a layered explanation system.TracIn data attribution achieves the highest ground-truth faithfulness for training influence (0.784). Chain-of-thoughtfaithfulness evaluation reveals that 38.4% of GPT-4 stated reasoning steps are post-hoc rather than causal.Counterfactual generation (GRACE) achieves the highest user comprehension for output explanation (3.84/5). UGEFachieves the best overall explanation quality (composite 7.84/10). A regulatory compliance framework for generative AIexplainability under EU AI Act Article 13 is proposed
Source: Bio-QI Journal
Authors: Hugo Klein, Daniel Muller, Laura Dubois
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
Award: G993971353
OpenAlex ID: W7203660075