Intelligent User Interfaces (IUIs) rely on user metrics to infer user characteristics and adapt interaction accordingly, yet these metrics are often specified in an implementation-dependent manner and become tightly coupled to a particular application context and technology stack. This paper proposes a model-driven abstraction process to specify user metrics independently of the underlying programming language, enabling their integration into Model-Driven Development (MDD) workflows. Building on a curated set of automatic user metrics for knowledge, skills, and goals, we describe how metric parameters can be expressed using conceptual user interface primitives and context constructs. We instantiate the approach with UsiXML and illustrate, through a practical scenario, how abstract metric values can drive systematic UI adaptations at the model level. To evaluate the feasibility of the proposal, we conducted a preliminary expert-based validation with three MDD specialists who evaluated the operationalization of the proposed metrics in UsiXML, WebRatio, and Integranova languages. Results indicate that most metrics can be supported across tools, while time-dependent metrics reveal limitations in environments lacking explicit temporal primitives.