Measuring Complex Systems: A New Approach

A new study introduces a framework for measurement in contexts that are difficult to analyze, where direct access to data is limited or impossible. This innovative approach combines indirect data from multiple sources, interpretable machine learning models, and triangulation techniques to overcome the limitations of traditional methods.

Multi-source Triangulation and Interpretable Machine Learning

The proposed framework relies on the use of interpretable machine learning models and the triangulation of data from multiple sources. Instead of relying on accuracy against unavailable ideal data, the framework seeks consistency across separate, partially informative models. This allows for drawing solid conclusions about the state of the analyzed system, based on the convergence or divergence of signals.

Applications and Inferential Limits

The framework offers an analytical workflow suitable for quantitative characterization in the absence of data sufficient for conventional statistical or causal inference. The researchers demonstrated the effectiveness of the method through an empirical analysis of organizational growth and internal pressure dynamics in a clandestine militant organization, using multiple observational signals that individually provide incomplete and biased views of the underlying process. The results show how triangulation and interpretable machine learning can recover substantively meaningful variation.