The seamless messaging agents include coarse-grained agents such as user agents, message transfer agents, server agents (for resources such as speech recognition and generation), application agents (interfacing to mail applications such as Microsoft and Unix mail) and device agents (for wireless phones, laptops, etc.). It also includes fine-grained agents, namely, user surrogate agents. Seamless messaging requires real-time performance which necessitates a tradeoff between agent communication, adaptability and intelligence to achieve a variety of tasks. In addition, seamless messaging requires user interface mediation between users and device agents to allow, for example, messages sent by users from a powerful graphic workstation to be received transparently by a cellular phone.
This extended abstract describes our work on the interactivity of interface agents. In particular, it focuses on the development of a model for designing agents which are more or less autonomous, but which recognize opportunities for initiating communication with their users in order to garner further useful information. This design can be applied to many different domains. The workshop call for papers indicates two main themes around which the workshop discussions will be organized - the work below fits into the individual agent (rather than multi-agent) theme, and describes how future requirement of making agents less autonomous and more trustworthy may be addressed. In recent years, the area of intelligent agents has been one of the most prevalent fields of research in the AI community.
The concept of agency has been used widely in computer science. Agent-like autonomous processes appear as modeling tools in several areas, such as cooperative problem solving (Gasser & Huhns 1989), distributed systems (Agha 1986), object-oriented simulation (Zeigler 1990), software engineering (Nigay & Coutaz 1992), and teleconferencing systems (Lee, .Mansfield, & Sheth 1993). In most of these cases, however, agents axe not directly providing a service to the user, or the user is unaware of the tasks agents are performing in order for the overall system to provide its functionality. Recently, user agents (or interface agents) have been receiving increasing attention from researchers (ACM 1994). User agents are autonomous or semi-autonomous processes that carry out missions delegated by the user.
As computer-based technology continues to progress and the increase of information-based paradigms in modern work force operations continues, solutions for these operations must be capable of delivering increased amounts of information for the human user. Unfortunately, as the amount of information increases, the user's task to process this information has become overwhelming. Furthermore, for human operators to just simply use the interface to these highly complex computer systems currently requires highly specialized training and education that is both lengthy and costly to businesses. The need exists for systems that can help solve this problem by providing abstractions and intelligent assistance in a self-contained software agent that communicates with the user through the user interface. Since each user operates in a different fashion, processing information in their own way, a complete intelligent interface agent must be capable of adapting to user needs and behaviors. We desire to address the two following difficulties with developing interface agents: (1) The extensive number of existing computer systems makes it impractical to build these agents from scratch for each system; (2) The agent must be compliant with existing user interface standards and business practices. Thus, an environment for constructing intelligent interface agents that satisfies standards compliance is necessary.
Rapid changes have taken place in our world over the ability to retrieve and manipulate data quickly between the different software applications. Information technology is revolutionizing the way we live. Knowledge is power and the ease layer of hidden agency consists of the five supportbased agents: user interface, database, unity, prior-of access to this knowledge makes the manipulation of the data critical. Yet, if one can harness the vast integration of data which flavors the users' knowledge the task that assists in the integration of data and proity, and timephase. Each agent represents a specific end result could somehow satisfy the elusive quest for vides recommendations, based on the users objectives.