The challenge: Verification and interoperability
Computational economics lacks a widely adopted specification for expressing dynamic models consistently across implementations, including their economic objects, conventions, and solution or estimation methods. Moving a model between implementations often requires manual reconstruction, so it is difficult to establish whether two implementations encode the same model and target the same solution. The rapid adoption of large language models compounds the problem: AI output is probabilistic by construction, and code an AI writes or translates can compute a model different from the one intended. Without an independent specification of model meaning — a semantics — the difference cannot be detected. Verification and interoperability alike require rigorous specifications of dynamic models and of their solution and estimation methods, separate from any implementation.
Vision
We want to enable economists of different backgrounds, and machines, to easily interpret, compare, modify, and implement dynamic economic models across toolkits. To that end, the group develops formal, community-written languages and semantic specifications for toolkits and domains of dynamic economic modeling. While we do not prescribe a single formalization or language, we aim for interoperability across different approaches.