Regulating Autonomous Agents Facing Conflicting Objectives: A Command and Control Example

Military commanders in the United Kingdom have a degree of devolved decision authority delegated from command and control (C2) regulators and are trained and expected to act rationally and accountably. Recent experimental results suggest that experienced commanders usually appear to act as if they are subjective expected utility maximizers. The only scenarios where this appears not to be so are when the immediate mission objectives conflict with broader campaign objectives. Then the apparent rationality of even experienced commanders often evaporates, in this paper we show that if the C2 regulator assumes her commander is expected utility maximizing and that he uses a suitable multiattribute utility function, then even when she is remote from the field of action and her information is sparse, this regulator can nevertheless predict when scenarios might lead her commanders into making irrational decisions.