Spreadsheets as
System Dynamics Language: Unlocking Model Translation with AI
Spreadsheets are everywhere. Yet we rarely think of them as modeling languages.
In this presentation, I share a core idea that has been shaping my recent work:
👉 When rigorously structured, spreadsheets already encode stocks, flows, variables, parameters, and causal relationships.
In that sense, they can be understood as a Domain-Specific Language (DSL) for System Dynamics.
Building on this perspective, I explore how Large Language Models (LLMs) can act not as black boxes, but as reliable semantic translators, converting structured spreadsheets into formal System Dynamics models - while preserving structure, meaning, and traceability.
This is not just a technical contribution. It has broader implications for:
-
communication among system modelers,
-
model reproducibility and auditability,
-
education,
-
and dialogue between research, policy, and practice.
📎 Here is the presentation e here is the spreadsheet with original data.
I would be very interested in hearing feedback, critiques, and related experiences.
This feels like an early step toward a future in which models can truly “talk” to each other, regardless of their original language.
Prof. Dr. Paulo Villela
villela.paulo@gmail.com
linkedin.com/in/paulovillela/