6
© 2007 Tom Beckman Features : Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating these tasks Perform tasks based on knowledge of the user, task domain, and environment Constantly monitor their environment to trigger actions Perceive their environment through sensors, receive stimuli that trigger action, operate under constraints, and then perform actions on their environment Use knowledge about the interests and priorities of people to perform routine tasks such as automatically screening, directing, revising, and responding to information Basic structure : Consists of goal, environment, sensors, and actuators Agent performance/behavior depends on : The goal or performance measure that defines the criterion of success The agent’s prior knowledge of the environment The actions that the agent can perform Agent Basics

© 2007 Tom Beckman Features: Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating

Embed Size (px)

Citation preview

Page 1: © 2007 Tom Beckman Features:  Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating

© 2007 Tom Beckman

• Features: Are autonomous software entities that act as a user’s assistant to perform discrete

tasks, simplifying or completely automating these tasks Perform tasks based on knowledge of the user, task domain, and environment Constantly monitor their environment to trigger actions Perceive their environment through sensors, receive stimuli that trigger action,

operate under constraints, and then perform actions on their environment Use knowledge about the interests and priorities of people to perform routine tasks

such as automatically screening, directing, revising, and responding to information• Basic structure: Consists of goal, environment, sensors, and actuators• Agent performance/behavior depends on:

The goal or performance measure that defines the criterion of success The agent’s prior knowledge of the environment The actions that the agent can perform The agent’s perception sequence to date The agent’s ability to learn what it can to compensate for partial or incorrect prior

knowledge

Agent Basics

Page 2: © 2007 Tom Beckman Features:  Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating

© 2007 Tom Beckman

After FSU Center for Performance Technology• Communication:

Understand user goals, preferences, skills, interests, and constraints Communicate knowledge to user about the task being performed Alert user as to progress and status of task

• Autonomy: Has a purpose and task Acts to achieve task until it is fulfilled Work and launch actions independent of user or other actors Appropriate level of autonomy – doesn’t overstep

• Adaptive: Learn from experience about its tasks and about user preferences Adapt its behavior based upon a combination of user feedback (both passive—usage

and active) and environmental factors (stimuli patterns)• Dynamic:

Detect changes in the environment and react in a timely manner Operate as a continuously running process without starting and stopping for each

process or task

Agents – Ideal Characteristics

Page 3: © 2007 Tom Beckman Features:  Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating

© 2007 Tom Beckman

After FSU Center for Performance Technology• Advisory Agents – Assist in complex help or diagnostic systems• Filtering Agents:

Remove data that does not match the user’s profile Helps to reduce information overload

• Navigation Agents – Remember shortcuts, site bookmarks, and pre-load caching information

• Monitoring Agents: Alert user to certain events Provide information when data are created, retrieved, updated, or deleted

• Recommender Agents – Take information from previous user behaviors and make recommendations based on these behaviors

• Workflow Management Agents• Retrieval Agents – Are more sophisticated search engines such as semantic

search• Interface Agents – Add presentation, speech, and natural language capabilities• System Agents – Assist with the management of the computing environment

Types of Agents

Page 4: © 2007 Tom Beckman Features:  Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating

© 2007 Tom Beckman

• Collects content across all media types (TV, radio, newspaper, magazines, blogs, business conferences, and podcasts)

• Agents can prioritize content based on user’s profile• User Profile Attributes:

TopicsPeopleOrganizationsPrior usage and alert history

• Rules are developed about content that are most authoritative: AuthorsPublishersTypes

• Agents continually remind user of great content

Features of Intelligent Alerts (After Paul Allen)

Page 5: © 2007 Tom Beckman Features:  Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating

© 2007 Tom Beckman

• Greater intelligence in agents:More AI – machine learning (NN, GA, FL) and expert systems (RBS & CBR)Autonomous and proactive learningProvide context, but not just contentJust-in-time, and just the right amount of information provided

• Agents easier to develop and interact with: User developed, improved, and maintained Interact via speech and gesture recognition

• Improved anthropomorphic features:Understanding user speech, gestures, animation, facial expression, and

non-verbal communication Expressing gestures, animation, facial expression, and non-verbal communication

• Agent collaboration:Agents talking to other agentsAgents interacting in a coordinated manner to perform larger tasksUse of the Semantic Web to provide context

Future of Agents (Steve Knode)

Page 6: © 2007 Tom Beckman Features:  Are autonomous software entities that act as a user’s assistant to perform discrete tasks, simplifying or completely automating

Advantages of Agents (after Peter Knode)

© 2007 Tom Beckman

• Accuracy: Adaptive ability ensures that intended result is produced despite changes in the

environment Provide context-sensitive help Step-by-step instruction leads and guides users through complex tasks and

reduces the potential for error• Convenience:

More effective use of user’s time Situates user in a computational work environment Provides relevant information and executes actions as soon as possible Remembers user history and preferences

• Personification: More timely and consistent application of LL Enhanced knowledge capture More permanent problem resolution (as opposed to “point solutions”) More “stickiness” in enforcing behavioral change