I wonder if it's like any technology and trade offs between helping/hurting people. If religions help 95% of people be more loving and 5% be more hateful, would it be a net benefit? Does it depend what those 5% do with the hate or the 95% with the love? What if the percentages were different?
Obviously an overly simplistic thought experiment, yet I wonder if we have any way, that isn't arbitrary, to say if a tool is net beneficial.
Side note...I've often played with the idea in the past of what I called the case for moral superposition: nothing is good or bad, everything is good AND bad, but thinking makes it good OR bad.