Model Object Graph
Representing models, assumptions, dependencies, evidence, and governance as one navigable graph.
A model inventory is a graph pretending to be a spreadsheet. I wanted to see what becomes askable once you stop throwing away the edges.
An experiment in treating a model inventory as what it actually is — a graph of models, the assumptions they rest on, the data and other models they depend on, and the evidence and decisions attached to those relationships — rather than a list of rows.
The question it’s testing
Rows make you good at counting models and bad at understanding them. The graph is a bet that the questions that matter in model risk are relational: shared assumptions, dependency chains, blast radius of a bad data source. If the representation holds those as first-class edges, those questions become navigable instead of impossible.
Current state
Prototype. It grew out of the note Model Risk as a Knowledge Representation Problem and is currently more design than system — I’m working out the schema of nodes and edges before building the navigation on top.