...social theories will be judged on their merits and not by a false analogy with natural science. I propose this as a convention for the protection of scientific method, not as a demotion or devaluation of social science. The convention sets no limits on what social science may be able to accomplishI don't quite agree. If we abandon the idea of testable social theories, we do limit what social science can accomplish.
I'd propose an alternative approach. Allow and expect that some social theories will not be testable. In many cases a subjective interpretation of the facts can provide value; whether it's Michael Porter's five forces model of markets, or the Jungian model of storytelling interpreted in 'The Seven Basic Plots'.
Other theories, however, will be testable. An economic relationship between growth, inflation and unemployment should absolutely be stated in falsifiable terms. Being testable - and tested - will strengthen its credibility immensely.
We simply need to make sure, when working with an idea, that we know the difference. Interpretive models can illuminate how we see the world, but we can't rely on them to make universal distinctions that predict anything. On the other hand we can use a falsifiable, scientific model - as long as someone has made an actual attempt to falsify it! - to objectively guide our behaviour.
This, after all, is why it's called social science. Sometimes - if not always - it is a science.
One final thought from Soros:
...social theories are reflexive. Heisenberg’s discovery of the uncertainty principle did not alter the behavior of quantum particles one iota, but social theories, whether Marxism, market fundamentalism or the theory of reflexivity, can affect the subject matter to which it refersThe assertion about Heisenberg is not exactly true - physicists have changed the behaviour of a lot of particles in lots of laboratories since Heisenberg. But more importantly: just because something is reflexive doesn't mean it is not amenable to theory. Mathematics (Godel), computer science (Turing) and cognitive science (Hofstadter) all contain self-reference; but all three have many strong, falsifiable theories making clear, repeatable predictions.
Reflexivity is a useful insight but it certainly doesn't overthrow the logical foundations of knowledge.