XMind

Software Agents

1. 1. Definiton and Differentiation

   1. Intelligent Agents

      1. Compared to standard software

         1. Autonomic

         2. Social/interactive

         3. Reactive

         4. Proactive

         5. possible requirements

            1. Adaptive/learning

            2. Persistent

            3. Cooperation, coordination, bargaining

            4. Flexible

            5. Mobile

      2. Interface Agent

         1. Continuous

         2. User interface

         3. Suggestions and reactions on user behaviour

         4. Integration in larger software applications

         5. Problems

            1. Isolated development

            2. Missing user testing

         6. Embodied/ animated Agent

            1. Cultural influence

            2. Motivation and expectations

      3. Mobile Agent

         1. Capable to migrate

         2. Life-Cycle

   2. Multi-Agent-Systems

   3. Component-based software / modular programming

      1. Standards

         1. OSGi (Java)

         2. EJB (Part of J2EE)

2. 2. Advantages

   1. Reduction of complex tasks

   2. Task delegation

      1. until reaching complete automation

      2. cost reduction

      3. security

   3. reducing software developing costs

   4. electronic transactions and bargaining

   5. Acquisition, processing, preparation and distribution of information

   6. less manual switching between programs

   7. social approval by using innovative products

   8. Perception of higher productivity (comp. Multitasking)

   9. better usage of computing resources

   10. mobile agents

       1. mobile code

       2. secured transport (signed, encrypted)

       3. situation-based protocols

       4. compensate constraints of mobile work (e.g. processing power)

       5. asynchronous communication

       6. own algorithms on external data

       7. reduced network traffic

   11. >especially suited for ubiquitous+mobile systems (vs. traditionell client-server)

   12. higher user satisfaction caused by context-sensible services

3. 3 Fields of application

   1. Assisting information workers

   2. Controlling/tracking

      1. Production

      2. Traffic

         1. autopilots

      3. Usage

   3. Ambient Intelligence (see Ubicomp)

   4. E-Commerce

      1. Comparison and autonomous buying

   5. Information Retrieval

      1. Data Mining

   6. Distributed information processing

      1. Grid-Networks

   7. Personalized services

   8. Remote maintenance

   9. Workflow Assitence

   10. by Kuhlen (1999)

       1. Support Assistants (Interface Agents)

       2. Development Assistants (Wizards)

       3. Information Assistants

          1. For reference (orientation)

          2. For validity (quality)

       4. Entertainment Assitants

          1. RPGs

   11. Many more... (e.g. by Wooldridge 2009)

4. 4. Problems

   1. Possible impacts of delegation

      1. Intrusion into privacy

      2. Relieve vs. Incapacitate

      3. maintain control

         1. competency of information management

      4. Dependence

      5. Overestimation of performance

      6. amount of delegation

         1. Utopia of a common level of information

   2. Trust

      1. Forms of trust

      2. Role of the state

   3. Data protection

   4. Legal practice

   5. User Perception

      1. Usability

      2. Usefulness

      3. Joy of Use

      4. Intrusiveness

      5. Transparency

   6. Operability

      1. Compatibility

         1. see standards and languages

      2. Reliability

      3. Security

         1. Espionage

         2. Attacks between agents and servers

         3. Manipulation

   7. for mobile agents

      1. missing interoperability

      2. disadvantages compared to desktop workplaces

   8. core problem: adjustment of following a target vs. adaption to the environment

   9. core problem: agent cooperation each with different targets

   10. reaction on unexpected events

5. 5. Optimization

   1. communication languages

      1. KQML (Knowledge Query and Manipulation Language)

      2. successor: FIPA ACL (Agent Comm. Language)

   2. Programming-by-demonstration

      1. form-building (tables)

   3. care for legal validity

   4. means of security

      1. transport encryption (SSL)

      2. not available on servers

      3. defeat against enemy agents

   5. higher intelligence

      1. machine learning

      2. textual analysis

      3. dynamical adjustments according to user behaviour

   6. Interface design

      1. multimodale Interaction

   7. management of trust

   8. open for established standards of modular programming

6. 6. Examples of use

   1. Interface Agents

      1. E-Mail-Function agents

         1. filtering/prioritization

         2. deliver additional information

         3. Pre-processing

         4. delegate tasks

         5. status tracking

         6. automatic reply

      2. calendar assistants

         1. Pattern Mining

         2. suggestion of meeting times

      3. collaborative filters for E-Commerce

      4. intelligent to-do lists

      5. internet search (e.g. Alexa)

   2. popular agent platforms

      1. JADE

      2. others

   3. Mobile Agents

      1. send an agent with a task into other networks/systems

      2. webbased planning of events

      3. distribution control of documents

   4. Multi-Agent Systems

      1. against E-Mail-Overload

         1. classification tasks

         2. observe who is suitable for which tasks

         3. suggestions: time frame, steps, warnings, ...

1. Freitext