And yet the idea of what someone considered similar to some other thing is in fact a conflation with the actual reality of what differentiates things.
The approach people take to taxonomy is just an arbitrary grouping based on what dimensions we arbitrarily decide are similar based on arbitrary measures.
One thing having things in common with another thing is not a guarantee that any other similarities will be observed or demanded within the system, which means that the rules were always made up in the first place, even if the results were apparently repeatable.
There's still no actual reality to these statistical imaginaries. They do not bind us in ways that allow prejudicial thinking to be apt.
You might be able to find that knowing statistical facts about groups let's you gamble, but gambling on people that way has a name, and that name is "prejudice".