4. Implications of time design for agents and autonomous systems

The above research on the design of trust in human collaboration in merging realities indicates the need to design time explicitly. This holds for all distributed systems. Individual systems have their own clocks. As a result, their interaction is asynchronous unless otherwise designed. Different methods for synchronicity between systems have been designed and are deployed. A typical domain for which this is needed is that of auctions: rounds need to be well-defined in time. This paper argues that new forms of time design are needed for systems in which human beings play a role: systems in which human beings participate. Existing solutions do not address this aspect. If human beings are to trust multi-agent systems, multi-agent systems need to be designed to this purpose.

Human participants need to be able to assess the capabilities, integrity, reputation and intentions of individual agents, and groups of agents: they need to decide/judge which agents they trust, which they don’t trust depending on the situations. They need to know how responsibilities are distributed. Knowledge of the owner of an agent can help, depending on the situation in which the agents are encountered, the tasks involved.

Systems designed to include human participants, albeit owners of agents involved or other human participant, need to be include time in their interface design. Time design must consider the time human beings need to reason and to react. All participants must be identifiable and addressable.

Time design must include explicit consideration of the three components defined above: durations of engagement, synchronisation and rhythm.

Duration of Engagement: agents should be designed to inform their owners and other human participants of their state either periodically, on request or in specific situations, depending on the application. Human owners cannot take responsibility for their own agents if they are not informed/capable of being informed. In interaction, over time, agents may adapt to the characteristics of other agents or human participants. Design must include the implications over time.

Synchronisation: Multi-agent systems need to tune/synchronise their activities with human participants, they need to decide which information is needed/relevant when. Timeliness needs to be defined and meeting synchronization procedures designed. This holds for both individual agents and groups of agents.

Rhythm: is the result of duration and synchronization: it defines the expectations of all participants with respect to interaction over time, structuring time, determining the degrees of trust users place in their collaborative environment.

Frances Brazier , CN