**What to traffic lights, cryptoeconomic models and recommendation engines have in common?**
1) **They are subjective**: ultimately a wide range of assumptions are encoded as mathematics, these choices are subjective even if the conclusions are objective relative to those assumptions. _In some areas yellow comes before red, in others, yellow comes before green._ Why; _Local Laws, Norms, Culture, merely different design sensibilities?_
2) **They are performative**: the resulting model steers the behavior of agents in a manner consistent with the models subjective premise*. The inherent feedback makes it possible for the model to be both the reason for and the result of the observed behaviors. _A machine learned model matches a subspace of features to a particular subspace of products, and so the relation is perpetuated, through the redistribution of the users' attention. Change the feature space, change the mappings, do you really expect the same buying patterns?_
**_Design is the art of navigating the subjective elements of technology!_**
_*Footnote: when successful; there are countless examples models in life from ML to traffic management that failed to live up to their design goals. _