Agents

1
Agents Agency

• What do we mean by agents?
• Science Fiction?
• Physical or Virtual?
• Are agents just a metaphor?
• By defining agents are we defining intelligence?
• Replacing people with agents
• Replacing systems with agents
• Agents force us to discover address tacit
  knowledge (experience) in knowledge work

2
Agents are here
3
Thats agents, not robots
4
What kinds of agents are there?

• Intelligent Agents
• Show intelligence in interacting with the
  environment
• Act based on previous facts or rules
• Make leaps of intuition
• Do things for you
• Autonomous Agents
• Interact with the environment
• Stimulus response, not always rules
• Can perform a wider range of activities over time
• Do things we do not or can not do

5
Agents Angels

• Agents as help to manage information
• Agents also are creating lots of new data
• Possible uses
• Managing our social life
• Replace conventional organizations
• Helping us find things (web crawlers)
• Performing repetitive information tasks
• A human face on information tasks?
• Jeeves, Watson, Sherlock, Eliza, the Wizard
• Pushes our interaction into something more
  natural

6
Agents or Angels?

• Who manages the agents?
• When is a decision made?
• Upon arrival of the information
• After information is compared
• Blink vs. DeepThought
• Is it possible to make our information use
  behavior transparent?
• Is it flattering?
• What if we find too much information?
• Enabling more knowledge work or just more work?
• Speed vs. Accuracy

7
Ranks of Agents

• Information Brokering
• Networks the Web
• Search Overload Management
• Product Brokering
• Shopping (Buying?)
• Recommendations / Collaborative Filtering
• Merchant Brokering
• Personalization, Customer service
• Logical, naïve decisions gaming
• Negotiating
• Multi-part decisions
• Conversation?

8
Autonomy Agents

• Delegating
• Do this every time
• More like this
• Buy when AUS to LAX 300
• How subtle is this kind of (tacit) work /
  decision making?
• Agents interacting with each other
• Negotiations between systems
• APIs, crawlers, auction bots, reminders
• Semantic Web
• Cancelbots (Nots)
• Easier to say no than yes

9
Autonomous Interface Agents

• Both showing doing things for users
• A combination of classical Artificial
  Intelligence Human Computer Interaction
• Interact with the interface the user
• Any program acting as an assistant with learning,
  inference, adaptibility, independence,
  creativity, etc. Lieberman p 1
• More examples
• Contextual help, tutoring systems
• Filtering, Highlighting systems
• Autonomous agents operate in parallel with the
  user

10
How agents should work

• Observe interface actions act
• Run in the background to
• Manipulate information ( the interface) for you
• Why this personalized view of agent interactions?
• Understanding one person is difficult enough
• Group interactions require making relationships
  explicit

11
Letizia - a surfing agent

• Watching what you surf
• Predicting where you might go
• Searching for you as links
• Another type of memory about what you do

12
Design principles for agents

• Suggest rather than act
• Take advantage of information the user gives the
  agent for free
• Take advantage of the users think time
• Problems with agents
• The users attention may be time shared
• Provide context for input
• Use the interface to show history state
• Deliberation vs. Action
• Sufficient data for a decision
• Waiting, watching doing
• Cognitive style differences
• Agents may not fit users mental models
• Distraction (over time?)

13
Social Filtering with Agents

• SOaP - a system of agents to mediate between
  people, groups the Web
• Privacy issues among people businesses
• Smart Groupware?
• Collaborative Filtering Agents?
• Coordinating knowledge work?
• Matching user interests
• Agents specialize in the system
• Supports groupwork automatically

14
Social Information Filtering

• Can a system replace word of mouth?
• Quality of information, not frequency
• Volume of information, broader perspectives
• Is the best information digital?
• Digital great for facts
• Not to great for intuitions or wisdom
• How do you code in organization culture?
• A focus on user (interest) profiles
• Helping you understand what youre already doing
• Large numbers help reveal quality
• Are we already used to recommendations?

15
Reducing Information Overload

• Most apt use of agents in the last 10 years
• Might be creating more information for us to
  manage
• Increases explicit knowledge,but tacit?
• We need agents to help do things in real time,
  without our intervention
• The Web agents are a good match
• Understanding behavior training agents
• Watching, examples, (others) profiles
• Not filtering as much as ordering
• Might be most interesting to see others agents,
  not your own

16
How might people use agents?

• Do we need more complexity to create simplicity?
• How long will it take before we dont worry about
  agent imperfection?
• Ensure control
• Meet privacy considerations
• Safeguards for actions
• Meet ( communicate) expectations
• Show actions, but hide complexity
• Build confidence slowly
• Social acceptance of agents (organizational
  expectation)

17
Blogs Social Dimensions

• Are blogs taking the place of newsgroups?
• RSS Readers
• Topic discovery methods
• Blog rolls
• Search engines
• Links
• Issues of Awareness
• Posting technologies s. Usenet