Demographic statistical evidence with a humane face

This thesis shows that suspicion about the existence of bias in demographic statistical studies which align with social stereotypes and potentially support one side of a major political controversy can provide sufficient grounds for rationally suspending judgment about their content and refusing to use them in decision-making. Using the framework of bounded Bayesianism, it explores the perspectives of both statistically sophisticated agents and statistically novice yet socially-aware agents towards these studies. It concludes that both types of agents can have adequate reasons to rationally reject believing and using such studies. The thesis also examines the language of statistical reports about social groups and shows how, in different contexts, they can imply harmful essentialist claims about those groups.