Balancing the criticisms: Validating multi-agent models of social systems