How Researchers Expanded Model Quality Frameworks for Man and Machine

When people talk about quality, they are often having a conversation about what matters to them and whether something meets their expectations. In the realm of computer science and systems modeling, defining what makes a model high quality becomes infinitely more complex as artificial intelligence steps in to build and interpret systems alongside humans.

To untangle this complexity, a comprehensive theoretical framework known as SEQUAL has been updated to evaluate how well both symbolic human-made models and mathematically driven machine learning models align with human goals. Published in the book Quality of Models for Man and Machine (2027), the expanded framework provides a structured lens to assess models across multiple tiers of abstraction, from physical storage to emotional resonance.

The Multi-Tiered Ladder of Model Quality

The revised SEQUAL framework organizes model evaluation into eight distinct quality dimensions. Physical quality examines whether a model or dataset is persistent, accessible, and current, ensuring it resists data loss and remains available to the audience when needed. Empirical quality focuses on the visual and statistical patterns of the data, drawing on Gestalt psychology and graph aesthetics to ensure node-edge diagrams are legible and balanced. Syntactic quality checks whether every statement inside a model obeys the strict vocabulary and grammar rules of the chosen modeling language, such as Unified Modeling Language or Business Process Model and Notation.

Semantic quality measures how accurately and completely a model reflects the real-world domain it is intended to represent. Because real-world domains are rarely fully formalised, researchers often rely on perceived semantic validity, comparing what human stakeholders interpret from a model against their own explicit knowledge. Pragmatic quality shifts the focus to human and machine comprehension, evaluating whether the audience can successfully understand and trace the model’s dynamics to take appropriate action.

Balancing Human Emotion and Machine Alignment

One of the notable additions in the updated framework is affective quality, which acknowledges that pleasing and aesthetically harmonious models are more frequently adopted and easier to learn. While machines remain indifferent to beauty, human stakeholders respond positively to intuitive layouts and familiar visual cues, while reacting negatively to poor design or deceptive patterns.

Social quality addresses the degree of agreement and consensus among different stakeholders, particularly when merging disparate mental models into a unified enterprise view. Finally, deontic quality ties everything back to the overarching goals of the modeling task, weighing the economic and operational value against the resources expended. It incorporates critical concepts such as AI alignment, ensuring that autonomous systems steer toward human intentions while avoiding inner and outer misalignments.

The research was published in Quality of Models for Man and Machine on January 1, 2027.