Why does economics primarily use a predictive modeling approach rather than an observational and descriptive approach?

Apologies that this is more of a second order and historical question but I’m not sure where else to ask it.

I often see economics compared to physics in that it’s less about trying to establish discrete facts and more about trying to model how things work and predict what that entails. Sometimes I see this conflated with the approach of science in general, ie “economics judges claims based on predictive power because it’s a science and that’s what sciences do.”

But there are obviously sciences that don’t primarily use predictive power to judge their claims. Linguistics, or Botany, or Archaeology for example. They rarely make predictive claims and are mostly concerned with observing, describing, categorizing, and interpreting the things they study. They’re still sciences and they’re still capable of generating nontrivial data. They also have the advantage of not really having to make many assumptions or generalizable claims, you can establish that something is true just by reproducing the empirical observation.

My understanding is that historically economics also looked more like this. The classical economists were doing far more description and interpretation and far less mathematical modeling and prediction.

Why is predictive modeling the right approach to studying economic activity? What advantages does it have? Is there something about the nature of “the economy” as a subject that necessitates an approach that looks more like physics than linguistics?