It's legitimate to generalise from few samples: you are not a statistician*. It's legitimate to say "those may not be representative".
It's OK to need proof. It's OK to believe without: you are not a courtroom.
It's good to forgive people. It's good to hold people to account.
Who generalises what, and who demands proof from whom, and who forgives whom, is a good way to cut through bullshit and dog whistles to get to what people really think.
* At p=0.0005 confidence