“Data driven campaigns are killing the Democratic Party,” argues this article on Politico. It argues Democrats need to return to story-telling, and not put so much trust in analytics and modeling.
It is absolutely true that far too many people naively trust models and big data. But I don’t think Democrat defeats are the fault of the models, however. Instead, it’s a matter of recognizing what such models are good for. Analytics are much better at optimizing within a particular set of rules, of squeezing out inefficiencies and inconsistencies. They can maximize a given set of variables.
But models are much less good at telling you what isn’t in the model, or recognizing new features in the environment. So they can optimize, but they can’t help you much at adaptation, recognizing new opportunities or threats. It’s not to say that computers can’t do this, but it’s a much more difficult problem than running statistical tests on a set of data.
Storytelling may be a more effective means of communication than a scatterplot, to be sure. But figuring out how you need to change your story or recognize something you were not seeing is a very different matter. What if the Democratic Party needs to adapt or change the mix somehow? Just changing the communication technique doesn’t help with that. You also need to think about the message.