The Enduring Value of Conceptual Modelling — Fundamentally Understanding What We Are Reasoning, Explaining, and Coding About

Prof. Henderik Proper

Abstract

Conceptual modelling, as a discipline, originated primarily from the field of database design. Its early development was driven by a need to clarify what a database system must “know” about the world, independent of any technology-specific implementation details. We argue, however, that the value of conceptual modelling extends far beyond database design. Conceptual models (including domain ontologies) enable us to explicitly articulate and discuss our understanding of a domain, whether existing, desired, or envisioned, in terms of the involved “things” and their relationships. More specifically, we maintain that, just as logic has given us discipline in how we reason, conceptual modelling can give us more discipline in clarifying and agreeing on what we are reasoning about (and likewise what we are talking, AI-explaining or coding about).

At the same time, within information systems engineering (including database & software engineering), the role of conceptual modelling appears increasingly “under threat”. The rise of agile software development, with its emphasis on “coding over documenting over modelling,” seems to suggest a reduced need for modelling. Furthermore, large language models are now frequently used to help generate models themselves. This raises an important question: Does conceptual modelling still have a role to play?

We strongly argue that it does. In particular, when it comes to clarifying and aligning our understanding of what we are reasoning about. We will argue the point “ViA RoME”. RoME refers to the concept of Return on Modelling Effort. Since creating conceptual models requires effort, it is reasonable to ensure that this effort aligns with the expected benefits. The ViA idea emphasises that models derive value only in action; i.e. when they are created and/or used. Together, these concepts offer a cost-benefit perspective on the creation and use of conceptual models.

Using the ViA RoME perspective as a guiding framework, we illustrate several situations in which conceptual models continue to play a vital role, including:

• Enabling ownership — Helping stakeholders take ownership of the concepts and rules embedded in code or in trained sub-symbolic AI systems.
• Explaining beyond explainable — Providing meaningful explanations of the behaviour or outputs of AI-driven systems (or any complex software-intensive system) demands a shared terminology and mutual understanding between the explainer and the explainee.
• Do you mean interoperability? — We live in an interconnected world, and both human and digital (including AI-based) actors require a shared (and thus explicit) understanding in order to interoperate effectively.