Model Science: getting serious about verification, explanation and control of AI systems

2025-08-27

AI Summary

The widespread deployment of foundation models raises critical challenges concerning trustworthiness, safety, and human alignment. Method: This paper introduces “Model Science” as a novel paradigm that places post-trained models at the analytical center, systematically investigating verification, explanation, control, and interaction across diverse environments. It establishes a conceptual framework built upon four pillars—verification, explanation, control, and interface—and proposes context-aware evaluation protocols, internal mechanism probing methods, model alignment techniques, and interactive visualization tools. Contribution/Results: The resulting integrated analysis framework enables rigorous assessment, transparent interpretation, and reliable control of foundation models. It provides both theoretical foundations and practical pathways for AI system governance and engineering deployment, facilitating a paradigm shift from data-centric to model-centric science.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

The growing adoption of foundation models calls for a paradigm shift from Data Science to Model Science. Unlike data-centric approaches, Model Science places the trained model at the core of analysis, aiming to interact, verify, explain, and control its behavior across diverse operational contexts. This paper introduces a conceptual framework for a new discipline called Model Science, along with the proposal for its four key pillars: Verification, which requires strict, context-aware evaluation protocols; Explanation, which is understood as various approaches to explore of internal model operations; Control, which integrates alignment techniques to steer model behavior; and Interface, which develops interactive and visual explanation tools to improve human calibration and decision-making. The proposed framework aims to guide the development of credible, safe, and human-aligned AI systems.

Problem

• Shifting focus from data-centric to model-centric AI analysis
• Establishing rigorous protocols for verifying AI model behavior
• Developing tools for explaining and controlling AI systems

Innovation

• Model-centric analysis replacing data-centric approaches
• Four pillars: Verification, Explanation, Control, Interface
• Interactive visual tools for human calibration