Meeting scheduling with preferences with limited comparability: Effects of agent knowledge ∗

In scheduling meetings, agents generally have preferences regarding possible times and sites. In addition, they may have privacy concerns along with the desire to solve collective problems efficiently, which necessarily involves a degree of communication. The present work is a study of meeting scheduling by independent agents, where preference scales are not assumed to be identical and where actual preferences are not communicated directly. The purpose was to study the means of reducing the effort (number of communications) required to find a solution that is in some sense optimal with respect to all agent preferences, as well as the relations between efficiency, solution quality (based on various measures) and privacy loss. Agents propose meeting times and places consistent with their schedules while responding to other proposals by accepting or rejecting them. Agents also store information about other agents that is gleaned from the communications, together with general assumptions about travel constraints and possible meeting sites. This provides a way of improving efficiency but also exacerbates the problem of privacy loss, since agents can deduce personal information regarding meetings as well as relative preferences from limited forms of communication. We develop strategies for finding optimal solutions, given minimal assumptions about the comparability of agent preferences, and we show that efficiency can be improved either by making deductions based on communications received or by revealing a limited amount of information about one’s own schedule in communications. We find that in some respects the problem of privacy loss can be finessed by this additional information.