One obvious disadvantage of testing a given claim with scientific constraints is that one may never know the number of possible constraints to try, in which combinations, and in which order to modify them. The tester may be incapable of demonstrating a claim's validity due to the sheer number of possible influences.
For example, consider the following political/economic claim:
Increasing the minimum wage for companies not operated by philanthropists, will increase societal wealth by giving more money to consumers who will spend this money in the economy and generate more demand and therefore more jobs, profits, etc.
If this claim were to be empirically tested across many types of businesses, mixed results could show up due to seasonality, geography, climate, employee characteristics, outside global trade factors, regulation, etc. So, in theory, someone could end up testing this claim until the end of time, with no clear way to control the "proper" factors and get a consistent, reproducible result. They could perpetually claim that their results are flawed because of flaws in the experiment itself, or they could end up with a flawed result while a flaw in the experiment is still present but claim to have proven the claim to be true or false.
The logical alternative to this approach could look something like this rough sketch:
- Humans act in the world
- Humans act out of self interest, pursuing what they believe will benefit them.
- Increasing the cost of operating a given business, by providing higher wages, will lower the profit margins of a given enterprise.
- If a individual (who is not pursuing philanthropic goals) decides to increase the wages of his employees, he can only do so by A) taking less profit himself B) lowering the quality of his goods/services or C) firing employees.
Therefore, increases to the minimum wage (in businesses that do not exist for philanthropic reasons) cannot make a society wealthier, according to this logic.
...but how can one know when logic, empirical study, both, or neither will provide the most efficient access to the truth value of a given claim? Is there even a suggested heuristic for such a choice? Some recommended reading would be appreciated.