Agent Definitions

Lashkari, Metral, and Maes, '94:

Interface agents are semi-intelligent systems which assist users with daily computer-based tasks.

Learning interface agents are computer programs that employ machine learning techniques in order to provide assistance to a user in dealing with a particular computer application.


Etzioni and Weld, '95:

By agent, we mean someone who acts on your behalf.

...desirable agent qualities:

• Autonomous: an agent is able to take initiative and exercise a non-trivial degree of control over its own actions:
  • Goal-oriented: an agent accepts high-level requests indicating what a human wants and is responsible for deciding how and where to satisfy the requests.
  • Collaborative: an agent does not blindly obey commands, but has the ability to modify requests, ask clarification questions, or even refuse to satisfy certain requests.
  • Flexible: the agent's actions are not scripted; it is able to dynamically choose which actions to invoke, and in what sequence, in response to the state of its external environment.
  • Self-starting: unlike standard programs which are directly invoked by the user, an agent can sense changes to its environment and decide when to act.
• Temporal continuity: an agent is a continuously running process, not a "one-shot" computation that maps a single input to a single output, then terminates.
• Character: an agent has a well-defined, believable "personality" and emotional state.
• Communicative: the agent is able to engage in complex communications with other agents, including people, in order to obtain information of enlist their help in accomplishing its goals.
• Adaptive: the agent automatically customizes itself to the preferences of its user based on previous experience. The agent also automatically adapts to changes in its environment.
• Mobile: an agent is able to transport itself from one machine to another and across different architectures and platforms.

By acting as an intelligent personal, the softbot supports a qualitatively different kind of human-computer interface.

• Goal oriented: the user specifies what; the softbot determines how and when
• Charitable: requests are incomplete, more of a hint
• Balanced: softbot evaluates cost of autonomy vs cost of pestering
• Integrated: softbot provides a single, expressive, and uniform interface

...intelligent software that performs the role of a human assistant.

To be truly helpful, an assistant must learn over time, through interacting with user and its environment--otherwise, it will only repeat its mistakes.


Terveen and Murray, '96:

Software agents are computer programs that act on behalf of users to perform routine, tedious, and time-consuming tasks.

One of the major promises of agents is personal assistance--each user can have an agent that serves his or her individual goals and preferences.

This means that the agent must acquire appropriate knowledge about the user.


Hedberg, '96:

...we will use the traditional definition of intelligent agents--autonomous software entities that can navigate heterogeneous computing environments and can, either alone or working with other agents, achieve some goal.


Hendler, '96:

Put simply, the goal of AI researchers working on intelligent agents is to create advanced AI programs that can function usefully in environmental niches of importance to humans.

...the real payoff comes from creating systems that can amplify human problem-solving abilities or can function autonomously in complex worlds.


Sycara, Pannu, Williamson, and Zeng, '96:

Although a precise definition of an intelligent agent is still forthcoming, the current working notion is that intelligent software agents are programs that act on behalf of their human users to perform laborious information-gathering tasks.

interface-agent: a single agent with simple knowledge and problem-solving capabilities whose main task is information filtering to alleviate the user's cognitive overload.

softbot: a single agent with general knowledge that performs a wide range of user-delegated information-finding tasks.


Kitano, '96:

I define an intelligent agent as a self-contained system that undertakes context-sensitive decision making and task enforcement in an open (or semi-open) environment.

Agent systems should behave reasonably in a broad range of situations--a feature that distinguishes agent systems from conventional software systems.


Maes, '96:

Agent programs differ from regular software mainly by what can best be described as a sense of themselves as independent entities.

An agent should also be robust and adaptive, capable of learning form experience and responding to unforeseen situations with a repertoire of different methods. Finally, it should be autonomous so that it an sense the current state of the environment and act independently to make progress toward its goal.


Maes, '94:

Interface agents are computer programs that employ Artificial Intelligent techniques to provide active assistance to a user with computer-based tasks.

The agent acquires its competence by learning from the user as well as from agents assisting other users.

Notice that the agent is not necessarily and interface between the computer and the user. In fact, the most successful interface agents are those that do not prohibit the user from taking actions and fulfilling tasks personally.


Minsky, The Society of Mind, '87:

We want to explain intelligence as a combination of simpler things. This means that we be sure to check, at every step, that none of our agents is, itself, intelligent.

Accordingly, whenever we find that an agent is has to do anything complicated, we'll replace it with a subsociety of agents that do simpler things.

• Agency: actions are performed on user's behalf
• Autonomy: agents are at least occasionally unsupervised
• Integration: agents or groups of agents utilize and integrate heterogeneous information sources
• Learning: performance improves over time via interaction