Causal Models in Requirement Specifications for Machine Learning: A vision

2025-02-17

AI Summary

In machine learning (ML) software systems, domain-specific prior knowledge is often difficult to systematically translate into concrete data and model requirements. Method: This paper introduces causal modeling into requirements engineering for the first time, proposing a requirements traceability workflow grounded in structural causal models (SCMs) and causal graphs. The approach enables a traceable, interpretable mapping from high-level domain knowledge to low-level ML requirements, explicitly encoding distributional assumptions, feature-level causal relationships, and intervention boundaries to bridge the semantic gap. Contribution/Results: Evaluated on an industrial fault detection system, the method achieves high-fidelity requirement translation, significantly enhancing the interpretability and verifiability of ML requirements specifications. This work establishes a novel methodological foundation for deep integration between requirements engineering and machine learning.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Action, Change, and Causality

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: ML for personalized search and recommendations

Abstract

Specifying data requirements for machine learning (ML) software systems remains a challenge in requirements engineering (RE). This vision paper explores causal modelling as an RE activity that allows the systematic integration of prior domain knowledge into the design of ML software systems. We propose a workflow to elicit low-level model and data requirements from high-level prior knowledge using causal models. The approach is demonstrated on an industrial fault detection system. This paper outlines future research needed to establish causal modelling as an RE practice.

Problem

• Integrate prior domain knowledge into ML system design.
• Elicit detailed model and data requirements systematically.
• Establish causal modelling as a requirements engineering practice.

Innovation

• Causal modelling integrates domain knowledge
• Workflow elicits requirements from prior knowledge
• Demonstrated on industrial fault detection system