In #complexsystems, we learn that #interactions between autonomous agents lead to #adaptation and #learning, which then often lead to #nonlinear #dynamic behavior. In contemporary #machinelearning, we try to model real-world behaviors by capturing #bigdata about entities, e.g. “features” that could lead us to the predictors. However, we rather naively assume that those agents all interact in a linear and deterministic fashion and hence those #correlations could indeed explain the #causality, or at least lead us to a satisfactory explanation of the underlying real-world phenomena. Unfortunately, bigdata limits our understanding to the static datapoints and doesn’t quite capture the dynamic #context of these interrelationships between agents, especially in #wickedproblems, and we probably need something else?
I stumbled upon the work on so-called #warmdata by nora bateson, and found the idea very interesting. As opposed to the bigdata that captures “cold data”, i.e. the static attributes of a dataset or a real-world system, warm data is all about the qualitative interactions between entities. It sounds similar to the ideas like #thickdata, and I am looking to learn more about them both. Looking forward to connect with folks working on this, and learn more.
