Naive Learning with Uninformed Agents