You are right to point out the moral hazard of a scientific institution normalizing this kind of behavior (not unlike a Milgram scenario).
“We work too exclusively on establishing airtight proof of cause and effect relationships between variables, with no input from the populations actually being studied”. Psychologists should be informed by other social sciences (like anthropology & sociology), but to shift focus from establishing causation among objective variables to actually doing the work of anthropologists with do both fields a huge disservice.
However, what this experiment measured is something completely different to the experience of racism. This measured the existence of dehumanizing racism. The former measures the recipient group, the latter the aggressing group. Both are important.
Consider for example only measuring the impact of racism on the oppressed. It could be argued to be the result of a few explicit racists. Experiments like this can quantify the degree to which racist attitudes exist implicitly, and what other traits or states they correlate with.