Abstract: A critical aspect of open systems such as the Internet is the
interactions amongst the component agents of the system. Often this interaction
is organised around social principles, in that one agent may request the help
of another, and in turn may make a commitment to assist another when requested.
In this paper we investigate two measures of the social responsibility of an
agent known as reliability and helpfulness. Intuitively,
reliability measures how good an agent is at keeping its commitments, and
helpfulness measures how willing an agent is to make a commitment, when
requested for help. We discuss these notions in the context of FIPA protocols.
It is important to note that these measures are dependent only on the messages
exchanged between the agents, and do not make any assumptions about the
internal organisation of the agents. This means that these measures are both
applicable to any variety of software agent, and externally verifiable, i.e.
able to be calculated by anyone with access to the messages exchanged.
Keywords: Social commitment, reliability, helpfulness, agent protocols, measurement