This is something I genuinely wonder about. If intelligence is just whatever tasks a psychologist decides to include, two different psychologists with different assumptions should produce tests measuring completely different things. But scores from different IQ batteries correlate strongly with each other. How do we explain that if the tests are just reflecting designer bias?
The convergence across independently developed batteries is the strongest argument against the circularity criticism. Researchers in different countries and different decades kept finding the same general factor emerging regardless of which tasks they included. That is not what you would expect if tests were just measuring designer bias. Uncorrelated inputs producing correlated outputs means something real is being detected.
If intelligence were just whatever a psychologist put on a test, Binet and Wechsler working decades apart in different countries should have produced completely unrelated measures 
The fact that they correlate strongly is the whole argument 