Science is a way of understanding the world that involves both:
1.    observing what happens in the world and
2.    making assumptions about what can't be observed directly.

Scientific ideas can come from either of these sources, but they never come from just one of them. The difference between what seems like two completely different types of scientific arguments can be seen as being different points on a line that shows how general or specific the ideas are. Ideas that are closer to the right-hand side of the line are more specific and based on close observation, while ideas closer to the left-hand side are more general and not based on direct observation.

The more specific ideas are always based on more general ideas, but that doesn't mean the general ideas are more important. In safety science, for example, it's important to not just theorize, but to do other work on the continuum as well. If we think about activities we suppose are good for operational health and safety, we'd better observe and describe real work and its constraints as well. And, if we see interesting correlations in data, we'd better do some theoretical work on the mechanisms behind it (as Rae et al. write: “For a field of research to move forward, each new project or paper must strive to change what has come before – adding, synthesising, testing, tearing down or making anew”).

If we look at the scientific continuum, making more general claims about a phenomenon (the left of the scale) is useful because it allows for greater generalization and better interpretation of results. But the degree of generalization should be based on the strength of the evidence. On the other hand, making more specific claims (the right of the scale) is more accurate and precise, but may not be as useful or practical in certain situations.

We could use a combination of tentative general and stronger specific claims, and aim to gather more evidence to support the general claims we make. By doing this, we can make evidence-based claims that are both accurate and useful for decision-makers in the field.

These are just some thoughts, influenced by my reading of Alexander (1982) and Rae et al. (2020). 

Ref.
Alexander, J. (1982), Theoretical Logic in Sociology – Volume One: Positivism, Presuppositions, and Current Controversies, Berkeley/Los Angeles: University of California Press
Rae, A.J., Provan, D.J., Aboelssaad, H., Alexander, R. (2020), A manifesto for Reality-based Safety Science, in: Safety Science Vol. 126, 104654.