On the value of information in multi-agent decision theory