Combining error-driven models of associative learning with evidence accumulation models of decision-making