\documentstyle [a4wide,11pt,epsf] {article} \parskip 0.05in \parindent 0in \topmargin -1.3cm \textheight 9.5in %\oddsidemargin -0.3in %\evensidemargin -0.3in %\marginparsep 0.2in \begin{document} %\parskip 0in \bibliographystyle{apalike} \begin{sloppypar} \title{Report on SIKS course on interactive and multi-agent systems, 30 november -- 4 december 1998} \author{B.W. van Schooten} \date{Version 1, 25 january 1999} \maketitle \tableofcontents \section{Introduction} The 5-day session was split into two separate courses: \begin{itemize} \item {\bf Interactive systems} is an introduction to human-computer interaction. Lecturers: Gerrit van der Veer, Charles van der Mast. \item {\bf Multi-agent systems} is about the various theories and applications related to `agent systems' and `agent technology'. Lecturers: Catholijn Jonker, Peter J. Braspenning, Floris Wiesman, Jules Meyer, La Poutre, Weigand. \end{itemize} The sheets can be found in \cite{siksdec98} and \cite{siksdec98:mast}. \section{Interactive systems} \begin{tabular}{|l|p{5in}|} \hline Lecturer & Material \\ \hline Van der Mast & 3 sets of sheets + appendices (bundled), thesis, supplemental sheets (SIKS9818) \\ Van der Veer & 2 sets of sheets + appendix: list of recommendations (SIKS9819)\\ \hline \end{tabular} \subsection{Miscellaneous} \subsubsection{Overview development methods for interactive systems} Lecturer: Gerrit van der Veer Material: sheets 1. General issues of cognitive ergonomics. Design usually starts with user and task analysis. Schools. Literature. \subsubsection{Prototyping} Lecturer: Charles van der Mast. Material: Supplemental sheets: prototyping (4 sheets), sheet of Digital Ink video (Digital Ink was also featured in CHI97 conference), appendix: "prototyping supermarket designs using virtual reality". Prototyping is needed to: enable more emphasis on synthesis rather than analysis, allow for testing/designing of human behaviour and behaviour in general. One prototyping model mentioned is the star model. The choices to be made with prototyping are: how to evaluate, what kind of prototype, throw away the prototype or not, degree of representativeness and completeness. Dangers of prototyping are: generally not all parties participate, less software discipline, unknown actual accuracy, expectations of client become too high, too much may be spent on making it look neat, aspects prototyped may be wrong. Tools are for example: VB etc, pen/paper, videos. Example of video prototype: Digital Ink prototype using scenarios. \subsubsection{Techniques, tools, development environment, and management} Lecturer: Charles van der Mast. Material: sheets 3, appendix: "creating the multimedia project experience database". Modelling of educational system design: what agents exist in the use of educational systems? Where lies the emphasis of the application design? Three modelling levels: macro (why: goal, users, and context), meso (what: instructional strategies, conceptual design), micro (how: semantic, syntactic, lexical design). A methodology accounting for each of these levels is discussed. \subsection{Techniques per development stage} \subsubsection{Task analysis and global design specification} Lecturer: Gerrit van der Veer. Material: sheets 1, appendix: list of recommendations. Methodological issues: evaluate early, design UVM separately, identify task model 1 (current system) and 2 (future system). Possible meanings of `task analysis' are: describing task model 1, task model 2, or UVM. Concepts: task=activity with specific goal, goal=state to be reached, primary task=task inherent to the task domain, secondary task=task that follow from use of specific tools, unit task=atomic primary task, basic task=atomic secondary task, action=subtask. Task model 1. Sources of knowledge: explicit or implicit, expert or group knowledge. Techniques for analysis: conceptual frameworks (+formal representations), planned processes, cognitive psychology. Examples of conceptual frameworks: TKS, ATOM, MAD. Means for obtaining knowledge. Examples: analysing history documents, ethnography, interaction analysis, GTA. Task model 2. Obtain questions, options, criteria (QOC), then resolve conflicts between options and criteria. \subsubsection{Detailed-design stage} Lecturer: Charles van der Mast. Material: sheets 1, appendix: student report on the application of the Foley-Delft method for designing a graphical editor, "Two-handed input in a compound task", "An evaluation of earcons for use in auditory human-computer interfaces", "Hoogtevrees in cyber-space". The book `the media equation' is reviewed, including some of the experiments that show that interaction with computer has a social aspect. This means that politeness, anthropomorphism, and Gricean principles are important for usability. Example of substages of design of a graphics editor: conceptual design (analysis of users and task, overall UI design), semantic design (list of functions that user can do and what program can report), syntactical design (state diagrams), lexical design (screen layouts). \subsubsection{Scenario-based design} Lecturer: Gerrit van der Veer. Properties of scenarios. Ways in which scenario may be used. Using claims analysis, which is stating claims about: task, functionality, human-machine interaction, technology, user opinion. \subsubsection{Evaluation criteria} Lecturer: Charles van der Mast. Material: sheets 1, supplemental sheets: evaluation criteria, matrix of Debora J. Mayhew. Classes of evaluation heuristics: standards, design guidelines, commercial style guides, modified style guides for own use. Two levels of design evaluation are addressed: interface style design and layout design. Some examples of each are given. Interface style design: ISO standards, some sets of design guidelines. Layout design: common errors, some techniques. \subsubsection{Evaluation techniques} Lecturer: Gerrit van der Veer. Material: sheets 2. Evaluation may be evaluation of: task model 2, UVM, full working system. Representations of evaluation results: formal, for designers (GOMS-lookalikes), exemplary, for everyone (scenario, mock-up, simulation, prototype). Evaluation techniques: claims analysis, cognitive walkthrough, objective observation, subjective evaluation (SUMI, SMEQ), hermeneutic techniques, standards, performance measures. %\subsubsection{Case study: analysis} \section{Multi-agent systems} \begin{tabular}{|l|p{5in}|} \hline Lecturer & Material \\ \hline Jonker & course program + 9 sets of sheets (SIKS9801), 9 articles (SIKS9802-SIKS9810), exercise.\\ Braspenning & 3 sets of sheets (SIKS9811-SIKS9813).\\ Weigand & 2 sets of sheets (SIKS9814).\\ Meyer & 1 set of sheets (SIKS9815), 2 articles (SIKS9816-SIKS9817).\\ Wiesman & 1 set of sheets (online) \\ La Poutre & 1 set of sheets (haven't got a copy!)\\ \hline \end{tabular} \subsection{Agent systems and models} %\subsubsection{Introduction to agent technology} % %Lecturer: Catholijn Jonker, Peter J. Braspenning. % %Material: ?????????? \subsubsection{Introduction to software agent technology} Lecturer: Peter J. Braspenning. Material: sheets 2. History and application areas of agents. What is an agent? Typology of agents: level of mobility, deliberative or reactive, level of autonomy, learning, and co-operation, roles (for example, interface and information agents), hybrid agents. Heterogeneous agent systems: how would different kinds of agents interoperate? \subsubsection{Agent concepts and agent behaviour} Lecturer: Catholijn Jonker. Material: sheets 1-2, article 1 \cite{jonker98:acb}, 2 \cite{jonker98:abs}, Why agents? Weak versus strong agency. Forms of intentionality. Classification of agents: life-like, robots, or software. Types of agent: collaborative, interface, mobile, information, reactive, hybrid, smart, collaborative interface. External (interaction, communication) versus internal (world model, self model, agents model, knowledge, reasoning, own control, beliefs, co-operation, adaptation/learning) agent concepts. Applications: industrial, commerce, medical, entertainment, sociological/biological. Practical session: choose an agent, fill in a form, classifying the agent according to the above concepts. This form may be seen as part of the DESIRE methodology. Animal behaviour. Three example agents are described using DESIRE notations: reactive, reactive with delayed response, pro-active. Suggestion: fourth type, a social agent. \subsubsection{Agent models} Lecturer: Catholijn Jonker. Material: sheets 3-6, article 3 \cite{brazier98:pcm}, 4 \cite{brazier96:fcm}, 5 \cite{brazier98:drg}, 6 \cite{brazier98:anl}. Compositional design in DESIRE: processes are modelled as a hierarchy of interacting components. Two kinds of compositionality: process and knowledge. Design is split into 3 stages: conceptual, detailed, operational. During each stage the aspects design rationale and problem description have to be maintained. Each of the 5 aspects are addressed in more detail. The software environment consists of: graphical editor, prototype generator and execution environment. Generic agents. two types of genericity: process and knowledge. Processes, information types (like world information, self information, agents information, etc), knowledge types (like classification, world-interaction, communication, own process control) of the generic agent model are described in more detail. Generic cooperation model for determining and executing a plan with multiple agents. Needed are: joint intention model, shared responsibility, shared schedule, monitoring of progress by all agents, dynamic adaptation of schedule. Process composition of a cooperative agent is described in detail. \subsubsection{Agent theories and agent languages} Lecturer: Jules Meyer. Material: sheets, articles \cite{rao91:mra}, \cite{meyer98:ia}. Informational attitudes and motivational attitudes. Agent theories: general and specific (BDI, KARO). The formal theories BDI and KARO are explained in more detail. Agent programming languages: AGENT-0, 3APL, (CON)GOLOG, Concurrent Metatem, ACL. Transition systems. \subsubsection{Theories of rational agents} Lecturer: Peter J. Braspenning. Material: sheets 1. The main topic is commitments. Why model commitments? Commitments versus intention. Pre-commitments. Conative policies. A commitment model. Social commitments. \subsubsection{Communication and cooperation} Lecturer: Weigand. Material: sheets 2. Cooperative Information Agents (CIA) are argued to be the next generation of databases. The model distinguishes 3 levels: task, contract and transaction. Formalisms and architecture. Methodology. Formal Language for Business Communication (FLBC). Meta patterns: general communication patterns. MEdiating and MOnitoring electronic commerce (MEMO). \subsection{Agent tools and languages} \subsubsection{FIPA: a generic agent toolkit} Lecturer: Peter J. Braspenning. Material: sheets 3. FIPA is an attempt at providing a standard for agent interoperability. The agent language used is ACL, which includes negotiation protocols and may include any domain-specific information. The specification also includes agent life cycle model (the agents report their state globally so they can be found), security model. \subsubsection{Java and agent technology} Lecturer: Floris Wiesman. Material: sheets. History of Java. Why Java (advantages/disadvantages)? Java and internet. Java security. Some software for implementing agents: Aglets (IBM), Voyager (Objectsspace), Concordia (Mitsubishi), JATLite (Stanford), JAFMAS, Java Intelligent Agent Library, Gadget (UM). \subsection{Agent applications} \subsubsection{Adaptivity, evolution, and emergent behaviour} Lecturer: La Poutre. Material: sheets. \subsubsection{Electronic commerce and information brokering} Lecturer: Catholijn Jonker. Material: sheets 7-9, article 7 \cite{jonker98:cdm}, 8 \cite{jonker98:iba}, 9 \cite{brazier98:dss}. Negotiating: a form of cooperation with both common and opposed interests. Example: bidding. Bidding strategies. More complex example: bidding for electricity prices in load balancing negotiation. Another form of negotiation is information brokering: mine for information according to implicit and explicit interests of user. Requirements and model of personal-assistant agent, including its environment. The agent is specified in more detail according to the aspects as listed in `generic agent modelling'. Call centre using a multi-agent system. Requirements: handle simple requests, delegate complex requests to people, standard procedure, maintaining an agenda. The system is described in detail. Practical session: design an electronic department store. Methodology: first, start with compositional model, then fill in agent forms (as was done in previous practical session). \subsubsection{Future of electronic commerce and information brokering} Lecturer: Weigand. Material: sheets 1. %\subsubsection{Case study: EC and IB} \bibliography{inf-2061-schooten} \end{sloppypar} \end{document}