The connection between causal analytics and cybernetics
This article is the entry point to a series of articles revolving around cybernetics and data science, and their logical connection. Concerning that connection causal analytics will be particularly important. Since the purpose of these articles is neither an introduction to cybernetics nor an introduction to data science and causal statistics, the contents from those topics are mereley rearranged and put into the context of the central question of this collection: How can an agent interact with its environment in order to achieve its goals?
Here are two variants of the truism that answers the question:
- If you want to reach a goal, you have to intervene in an informed and intelligent way
- If you want to reach a quantifiable goal, then in order to optimize your interventions you need an optimal statistical model that connects your interventions with your goal-KPI
If you find those two propositions obvious, that’s great and we can immediately procede to two consequences:
- The creation of an optimal model for interventions requires “causal analytics” – i.e., not just predictive models, but models that remain intact after interventions
- In doing this we have to face the problem of extracting the required information mostly from time series data, since as such the data that underly our model are often given.
The first of the consequences derives from one central word in the two initital propositions, that is “intervene”. Whatever we know about a system within which our goal is situated, as soon as we intervene into that system, we change it, and therefore we lose the certainty of knowledge about it that has been acquired before the intervention. This central predicament is what requires causal analytics, a statistical technique this article series will try to elucidate.