Interdisciplinary Advances in Adaptive and Intelligent Assistant Systems: Concepts, Techniques, Applications, and Use

Interdisciplinary Advances in Adaptive and Intelligent Assistant Systems: Concepts, Techniques, Applications, and Use

Gunther Kreuzberger (Ilmenau Technical University, Germany), Aran Lunzer (Hokkaido University, Japan) and Roland Kaschek (Gymnasium Gerresheim, Germany)
Indexed In: SCOPUS
Release Date: August, 2010|Copyright: © 2011 |Pages: 300
DOI: 10.4018/978-1-61520-851-7
ISBN13: 9781615208517|ISBN10: 1615208518|EISBN13: 9781615208524
Hardcover:
Available
$180.00
TOTAL SAVINGS: $180.00
Benefits
  • Printed-On-Demand (POD)
  • Usually ships one day from order
E-Book:
(Multi-User License)
Available
$162.00
List Price: $180.00
10% Discount:-$18.00
TOTAL SAVINGS: $18.00
Benefits
  • Multi-user license (no added fee)
  • Immediate access after purchase
  • No DRM
  • ePub with PDF download
Hardcover +
E-Book:
(Multi-User License)
Available
$215.00
TOTAL SAVINGS: $215.00
Benefits
  • Printed-On-Demand (POD)
  • Usually ships one day from order
  • Multi-user license (no added fee)
  • Immediate access after purchase
  • No DRM
  • ePub with PDF download
OnDemand:
(Individual Chapters)
Available
$37.50
TOTAL SAVINGS: $37.50
Benefits
  • Purchase individual chapters from this book
  • Immediate PDF download after purchase or access through your personal library
Description & Coverage
Description:

Information is becoming the raw material of modern society; it is the driving force of modern service industries. Our information spaces have been pervaded by technology and are characterized by their increasing size and complexity. Furthermore, access to information spaces and the ability to use them effectively and efficiently have become key economic success factors.

Interdisciplinary Advances in Adaptive and Intelligent Assistant Systems: Concepts, Techniques, Applications, and Use encourages knowledge on effective and efficient approaches to accessing information spaces. It fosters an emerging key competence: accessing and processing large, highly complex corpora of information by applying collaborative, intelligent technical systems. It is the mission of this book to trigger interdisciplinary research and cooperation at the intersection between information sciences, information technologies and communication sciences. This publication also raises awareness of the field’s importance in business and management communities, thus contributing to the dissemination of scientific ideas and insights.

Coverage:

The many academic areas covered in this publication include, but are not limited to:

  • 3D geographic simulation framework
  • Combinatorial keyword query processing
  • Describing and automating Web information access processes
  • Expert online retrieval tools
  • Knowledge web
  • Mobile intelligent information technology
  • Software Agents
  • Video viewing experience
  • Virtual librarian
Indices
Reviews and Testimonials

With this book we try to highlight the role of interdisciplinarity and adaptivity. There is good reason for that. First, from our definition as given in the preface of the earlier book, intelligent assistant systems inherently are adaptive; it turns out that this adaptivity is a key issue, that we try to explore in more detail here.

– Roland Kaschek, KIMEP, Kazkhstan; Gunther Kreuzberger, TU Ilmenau, Germany; & Aran Lunzer, Hokkaido University, Japan
Table of Contents
Search this Book:
Reset
Editor Biographies
Gunther Kreuzberger works at the Ilmenau Technical University. Having received a masters degree in computer science in 1996 for the implementation of a neuro-fuzzy system, his research then focussed on multi-agent systems for therapy planning. In 1998 he joined the Institute of Media and Communication Science and became a senior lecturer and executive assistant to the collegiate administrative committee. As a lecturer he is nowadays in charge of lectures and seminars in the fields of digital communication, electronic documents and interactive media such as digital games or iTV. His research interests cover higher education courses on interactive media/ digital games, collaborative IT-enhanced life-long learning, and media applications for children as well as elderly people.
Aran Lunzer is a British researcher in Human-Computer Interaction, currently employed at Hokkaido University as an Associate Professor with special responsibilities for overseas research liaison. After undergraduate studies in Engineering at Cambridge University, he worked from 1986 to 1991 on software technology for IBM UK Laboratories, then returned to full-time study to obtain his PhD in Computing Science from the University of Glasgow. From 2002 to 2004, between spells in Japan, he was an Assistant Research Professor at the University of Copenhagen. Aran’s research is focused on what he calls “subjunctive interfaces”, which are interfaces that support users in examining and comparing alternative results during trial-and-error use of software applications. He believes this is a necessary but generally lacking form of support in domains such as simulation, design, and data retrieval. In 2008/9 he has been building and refining a subjunctive interface for a cancer treatment simulator, as part of a large EU Integrated Project.
Roland Kaschek studied mathematics at the university of Oldenburg (Germany). He received a joint Soviet-German PhD grant for study in Novosibirsk and Moscow in 1986-1987, and obtained his PhD in mathematics from the University of Oldenburg in 1990. After that he was an Assistant Professor at the University of Klagenfurt (Austria); at that time he worked on various aspects of information systems design, database design and business process design. From 1999 to 2002 he was an informatics consultant with UBS AG in Zurich (Switzerland), working on software architecture, software quality and data warehousing. Then until 2008 he was Associate Professor with Massey University in Palmerston North (New Zealand), where he continued to deal with information systems design issues and in particular became involved in Web information systems design and eLearning. He was then appointed Full Professor with the KIMEP in Almaty (Kazakhstan) until 2009, and was additionally a guest lecturer or professor with universities in Austria, Brazil, Germany, Thailand, and the Ukraine. Currently he is a mathematics and informatics teacher at Gymnasium Gerresheim in Düsseldorf (Germany).
Editorial Review Board
  • Klaus Peter Jantke, Fraunhofer IDMT, Germany
  • Paul Klimsa, Technische Universität Ilmenau, Germany
  • Masahiko Sato, Kyoto University, Japan
  • Nicolas Spyratos, Université Paris Sud, France
  • Yuzuru Tanaka, Hokkaido University, Japan
Peer Review Process
The peer review process is the driving force behind all IGI Global books and journals. All IGI Global reviewers maintain the highest ethical standards and each manuscript undergoes a rigorous double-blind peer review process, which is backed by our full membership to the Committee on Publication Ethics (COPE). Learn More >
Ethics & Malpractice
IGI Global book and journal editors and authors are provided written guidelines and checklists that must be followed to maintain the high value that IGI Global places on the work it publishes. As a full member of the Committee on Publication Ethics (COPE), all editors, authors and reviewers must adhere to specific ethical and quality standards, which includes IGI Global’s full ethics and malpractice guidelines and editorial policies. These apply to all books, journals, chapters, and articles submitted and accepted for publication. To review our full policies, conflict of interest statement, and post-publication corrections, view IGI Global’s Full Ethics and Malpractice Statement.

Preface

This book is the second publication of the loosely organized PISA (Perspectives of Intelligent Systems' Assistance) project on assistant systems. In our first book (Kaschek, 2006) we tried to set the stage for a general debate of assistance and tried to push the development and use of assistant systems to the level of what was possible back then. With this book we try to highlight the role of interdisciplinarity and adaptivity. There is good reason for that. First, from our definition as given in the preface of the earlier book, intelligent assistant systems inherently are adaptive; it turns out that this adaptivity is a key issue, that we try to explore in more detail here. Second, and going beyond what we saw back in 2006, interdisciplinarity is very important too. Maybe not for each and every assistant system, but certainly many instances of requiring computer-based assistance will have to take into account a number of different disciplines at the same time. As is customary for a preface we first introduce our subject, discuss some of the related work, and then briefly summarise the chapters of this book.

As we did in 2006, we still feel that the most fundamental concept we have to deal with is that of information. Our related view is one in which we take into account structural aspects of modern life and do not necessarily restrict ourselves to the individual human using a computer. We consider an information society as a society in which information is the overwhelmingly dominant item in production, management, support and consumption. What we did not really understand at the time of the first PISA book is that information is becoming key even in consumption processes. That process is driven by the following causes: (1) many products or services are complex, and using them appropriately is not necessarily a trivial task; (2) providing the product or service meta-information required for a conscious choice about whether and how to buy them (i.e., what business model to use for purchasing or, more generally, for acquiring access or consumption rights), as well as how to consume or use them, is expensive. One way of getting an edge over competitors is to offer that information in computer-readable electronic format only; that way it can be kept up to date much more easily and effectively. Also, that way the information can be made much more meaningful to consumers, as it can be computer processed and turned into a shape useful for the consumer. That processing could even involve the consumption context and boost meta-information usage. Obviously, the act of product or service purchase can also be subject to computerised support. On the horizon we see a stage of society in which, without one’s personal digital assistant (PDA), one will not reasonably be able to go shopping in a supermarket, to buy products or services at standardised points of sale, to navigate within the city (including getting timetables for buses or trams) or on the highway. The technology of informational infrastructure is going to change fundamentally. PDAs enable that process, and require it as a precondition of general usage in modern society. The ability to produce powerful PDAs is creating new needs that rely on their availability. Since information technology is a major modern industry it is very likely to find ways of convincing citizens to use these devices on a large scale.

Information is becoming the raw material of modern society. According to a definition that goes back to Bateson (1979), information is the “difference that makes a difference”. It is the driving force of modern service industry. It is customary to use the metaphor of information space to characterise the effects of the ongoing change process. Our information spaces have been technologised, and their size as well as their complexity increased. Access to information spaces and the capability to use them effectively and efficiently have become key economical success factors. Kuhlen (1999) diagnosed an information dilemma. The alternatives he identifies are (1) searching the information spaces oneself, spending an increasing amount of time for satisfying one’s information needs; and (2) delegating that searching task to an information assistant, whose performance is time-consuming to control and whose effectiveness is hard to assess. Obviously, the only way out of that dilemma is the emergence of information brands, in other words, information quality, as certified by a trusted and trustworthy agency. While many of the contributors to Wikipedia, for example, are not necessarily aware of that issue, it was foreseen in philosophy of science when Popper discussed the fundamental role of authority for truth. Also Fleck (1935) discussed this when he commented on the emergence of scientific fact, the role of this emergence, and the role of the related community that has established a fact and keeps maintaining it.

Currently the first of Kuhlen's choices dominates. However, using such an information assistant must be the ultimate goal, as one cannot reasonably expect that technologisation will stop or even be undone. In fact the technologising of the world has a long history and must be considered as one of the most fundamental driving forces of human civilisation. It is inextricably connected with technologising the word (Ong, 1996, p. 101), as “(s)poken words are always modifications of a total situation which is more than verbal. They never occur alone, in a context simply of words.” According to Ong, technologisation has significantly restructured human consciousness. What we are witnessing right now should be understood as the latest step in the process of technologising the word. That process began with the invention of script (about 3200 – 3100 BCE), and was followed by the invention of letters (about 1500 BCE), and print (1540 CE), (Ifrah, 2000, pp. 3 – 25) – although some recent research suggests somewhat different dates. The latest step of technologisation is the automation of uttering and understanding words and text, in which we here include calculation and computation. This sums up the capability of verbally controlling processes of all kinds. Enzensberger (1970) reports the analogy that El Lissitsky drew in 1923 between language use and modes of transportation and, in particular, vehicles. He related articulated language to upright gait; writing to the wheel; and the printing press to animal-powered carts. El Lissitzky did not suggest analogies in language use for automobiles and aeroplanes. One is tempted to suggest that computers correspond to automobiles, and the Web to aeroplanes. Would it be venturing too far to relate intelligent assistants to spacecraft?

The idea of assistant systems is not a new one. Rather, it has been around for 30 years. Recently the interest in assistant systems has been refreshed by various authors because of a need to address a deep-rooted worrying issue. That issue is the role that is played by computers and humans respectively during their interaction. A number of advanced general approaches to computer-based technologisation of words and text were proposed in the past but were not heavily used in application systems, a point made by Sowa (2002). We expect that this will not happen to intelligent assistant systems. They will make their way to mass systems. That they did not do so already is, on the one hand, due to aspects of the genesis and evolution of computers. Computers were only invented in the fifth decade of the 20th century, and as they came into widespread use in the sixth and seventh decades they continued to become ever more powerful, so that solutions of practical problems became feasible that were not feasible only a short time earlier. Miniaturisation of computers along with ongoing increases of computational power and storage capacity has now led to a situation in which the devices required for general usage of assistant systems are starting to emerge. The economic need identified above to completely reorganize the informational infrastructure towards a situation in which the human individual without a PDA is lost in information spaces is going to enforce the mass usage of assistant systems.

It was as early as at the end of the eighth decade of the 20th century that Winograd (1979) (relying on a study of the US Department of Defense) wrote that computers “are not primarily used for solving well-structured mathematical problems or data processing, but instead are components in complex systems. … Many computer scientists spend the majority of their time dealing with embedded computer systems such as message systems and text editing and formatting systems.” By about 1975 the task focus of computers had changed from computation to communication (Hevner & Berndt, 2000). With the dominant use of computing technology today being for business application (p. 13), a significant increase has taken place in the number of individuals who are directly affected by computers and are engaged in using them. They also use computers in an increasing number of domains and tasks. The complexity of problems for whose solutions computers are key components continues to increase. The traditional approaches of using computers turn out to be ineffective for these problems of higher complexity. On the one hand, these approaches require the human problem solver simultaneously to maintain a good understanding of both the overall solution architecture and the solution detail. Many humans experience difficulties meeting this requirement. On the other hand, the approach fails to let computers work out the problems and their solutions, as humans working under their own resource restrictions cannot understand and validate these solutions.

The Oxford English Dictionary Online defines ‘complex’ as “(c)onsisting of parts or elements not simply co-ordinated, but some of them involved in various degrees of subordination; complicated, involved, intricate; not easily analyzed or disentangled.” A related but more operationalised definition was given by Victor Basili, (see Banker et al., 1993). He defined the term ‘complexity of a system’ as the resources needed to successfully use that system. If we consider a problem as a system the main components of which are a current state C, a future state F, and an actor A who wants to transform C into F given the boundary conditions B, then problem complexity can be understood as a particular instance of system complexity, i.e., as the resources required to transform C into F. That definition can be looked at from a quantitative point of view. Then problem complexity can, for example, be understood in terms of the number of points being used to score a problem. This approach is similar to the one taken in the Function Point methodology (see Garmus & Herron, 1996). Alternatively to this standpoint, problem complexity can be considered qualitatively. Then problem complexity can be understood in terms of the kind of resources needed for solving the problem at hand. New levels of problem complexity then call for new kinds of resources or for new ways of using known ones.

In terms of computer applications, we think that new resources and new ways to use old resources can best be understood as what could be called “the interlocutor metaphor” or “role” that a computer system plays in an interaction with a human. We focus here on the various different ways of understanding the interlocutor role. We identify the roles from a dominance perspective and reuse what already has been suggested in the preface of (Kaschek, 2006), namely the roles tool, collaborator, peer, or master in the 1:1 human-computer-interaction. For a more elaborated list of such interlocutor metaphors we refer the reader to, for example, Buschmann et al. (1999). Our interlocutor metaphors for 1:1 human-computer interaction were taken from human cooperative labour organisation. The generalness of computers enables them to embody many different interlocutor metaphors and thus to appear as qualitatively different resources. Computers can therefore be used for solving problems of quite different complexity.

The mentioned interlocutor metaphors for one-on-one human-computer interaction are discussed in more detail below.
  • master – tool, i.e., in this relationship the master is superior to the tool and the latter has no or only little latitude in performing the tasks allocated to him / her. The success criterion applied to that relationship is that the tool exactly carries out what it is instructed to. Successful use of a tool depends on the master’s capabilities of, first, structuring problem-solving procedures into steps that can be carried out by the tools at hand and, second, handling the tools appropriately. A currently employed version of this relationship is the chain of command in the military all over the world.
  • master – collaborator, i.e., in this relationship the collaborator has a relatively large latitude regarding what tasks to perform and how to do them. However, the master is superior. The success criterion applied to this relationship is that the master achieves his / her goals. For that, the collaborator appropriately contributes his / her capabilities to the problem the master tries to solve. Successful use of a collaborator depends on his / her capability to learn about and to adapt to the master’s way of conceptualizing and solving problems, as well as the master’s ability to use the collaborator’s capabilities and to delegate tasks to him / her, i.e., let the collaborator do what he / she is good at. A currently employed version of this relationship is the relationship between an enlightened manager and his / her subordinates.
  • peer – peer, i.e., in this relationship partners interact with each other and neither of them has power over the other or dominates him/her. They cooperate according to their individual goals for tasks and in a way that appears advisable to them. Successfully employing this relationship means perpetuating it and depends on the capability of finding peers who are willing to achieve compatible goals and to engage in communication processes that stimulate the ongoing process of goal achievement. A currently employed version of this relationship is the relationship between collaborating colleagues.
These three relationships between interlocutors cause five different roles in 1:1 relationships to exist in which computerized systems can interact with humans. The role of collaborator has two obvious specialisations, namely agent (one might consider the notorious 007 as an archetype) and assistant (an archetype of which might be Sherlock Holmes’ well known companion Dr. Watson). In this book we mainly discuss aspects of the assistant role occupied by a computer in the master – collaborator relationship. While authors of chapters of this book were free to express their own definitions for agent and assistant, the editors believe that this distinction, as introduced in the preface of (Kaschek, 2006), is a reasonable one. If we would want to extend the range of our concepts then, among other domains, we would certainly have to deal with gaming applications. Under certain circumstances they can be considered as assistant systems, insofar as they aid humans in “pleasantly killing time” (recreation). This then gives rise to a different model of assistance. What we have sketched above can be regarded as interlocutor assistance. In addition we see now that an assistant might enhance the reality of the human it is assisting, and that as a consequence that human gains the power to accomplish things they could not accomplish before. A simple example of this is a parking assistant that issues signals to the driver indicating the distance to the next physical obstacle; this clearly simplifies parking cars in overcrowded cities. Hence we feel that assistant systems should not be limited in purpose to avoiding human mistakes; clearly there is another ever-present goal of improving the quality of human life. Therefore user-centric perspectives matter. This is the ultimate reason why Intelligent Assistant Systems as a research topic is necessarily interdisciplinary.

We forecast that the trend will continue towards more (ubiquituous) computing power pervading everyday life and enabling mankind to deal with ever more complex applications. We therefore anticipate that “computers in the role of human-like peers” will continue to attract attention. For example, the ongoing demographic process of an aging population in industrialised countries will create a need to provide companions for elderly people whose loved ones have passed away, but who do not need the constant attention of a human nurse. It is not hard to forecast that for the younger members of society, too – children of all age groups – specific electronic companions will come into existence. In fact with the tamagotchi (see for example the related Wikipedia article) and the various kinds of artificial dogs we have seen a number of such attempts. Obviously, the performance characteristics of such companions may differ considerably from one to another. A few examples of what might be the focus of such devices are (1) performing helpful jobs in the household, including nursing the elderly, babies or toddlers; (2) aiding human users with pleasant ways to pass the time; or (3) serving as a general purpose interlocutor. It might not even be too far-fetched to assume that such an interlocutor device will one day come with a placeholder for a mental and physical fingerprint, in such a way that after adding the related parameter instances in most cases (perhaps in 95 % of all cases) one cannot really tell the device apart from the human being whose prints were taken. That way mankind really and for the first time would defy death. It is, of course, clear that there are immense cultural obstacles to overcome for a vision of this magnitude to become reality. Also, the cultural consequences of systems of that kind becoming used on a large scale would be enormous. And it is certain that for all of this there are not only civilian applications; rather, due to the large amount of resources that will have to be spent it is likely that the first creators and users of such advanced assistant systems will be the military.

The analysis regarding the increasing complexity of problems that have to be solved is not entirely novel. It is actually quite old. Engelbart (1962) writes, for example, that “Man’s population and gross product are increasing at a considerable rate, but the complexity of his problems grows still faster, and the urgency with which solutions must be found becomes steadily greater in response to the increased rate of activity and the increasingly global nature of that activity. Augmenting man’s intellect … would warrant full pursuit by an enlightened society if there could be shown a reasonable approach and some plausible benefits.” Engelbart even used a particular instance of the interlocutor metaphor. He called the resource that he was asking for a “clerk”. There are at least two respects, however, in which we differ in our approach from Engelbart’s and others that are comparable. First, we do not aim at creating a particular artifact that would be capable of aiding a human user in all contexts, situations, and conditions. Rather, we aim at domain dependent aids. We furthermore concede that “agent”, “assistant”, and “clerk” are unlikely to be the only reasonable instantiations of the interlocutor metaphor that can be employed in solving complex problems.

We follow Winograd (1972) in conceiving of assistants as systems that understand what they do and that are capable of answering relevant questions. However, we regard assistants of this kind as intelligent assistants, and distinguish these from reality-enhancing assistants that do not necessarily have to understand anything they do. In Winograd's spirit, Robertson, Newell, and Ramakrishna, cited after (Sondheimer, Relles, 1982, p. 106) write, “One does not wield an intelligent assistant, one tells it what one wants. Intelligent assistants figure out what is necessary and do it themselves, without bothering you about it. They tell you the results and explain to you what you need to know.” From an architectural view, that spirit suggests that assistants contain a domain model, a user model, and a situation model that combines the state of affairs of these two models as far as they are relevant for the task at hand. We do not aim at conceiving all sorts of assistant systems as a particular kind of information system, as the latter are understood from a functional view as, according to Hirschheim et al. (1995, p. 11), “technologically implemented [media] for the purpose of recording, storing, and disseminating linguistic expressions as well as for the supporting of inference making.”. For several modern assistant systems this would be too narrow a definition, since it presupposes a human as one of the communicators, and the communication as employing a language. To cover these modern systems one must conceive assistant systems as embedded systems, i.e., systems that are embedded in other systems and store, record, exchange, and process signals. If one matches the master-assistant communication against Shannon’s communication model (Shannon, 1993) then one finds that certain encoding or decoding steps may be superfluous. Additional to this more technical point one finds that the assumption of a deliberate choice of messages does not necessarily apply to the input that is fed into an assistant system. Rather, such input may result from a master activity that is independent of the assistant. Consider for example a power-steering or braking assistant. The master’s activity (steering or braking respectively) occurs as a natural problem-solving behaviour, and sensors derive the assistant system’s input from this behaviour. Furthermore, the technical problem of Shannon’s communication theory (i.e., the reconstruction of a message that was chosen at a particular point in space at a different point in space at a later time) is not the technical problem of an assistance theory. The technical problems of assistance theory are to identify (1) the master’s state of affairs, intent, and activities; (2) those responses to a master’s activity that are likely to aid the master in their current course of action; and (3) the most suitable way of telling the master what they need to know.

With respect to those cases in which a human is supposed to communicate verbally with the assistant, we can, however, stick to understanding the assistant as an information system. The kind of linguistic expressions that the assistant system is then supposed to handle successfully depends on the purpose of that system. According to our understanding of assistant systems as highly capable, adaptive, cooperative, domain-specific problem-solving aids, the mentioned linguistic expressions must enable an assistant to autonomously find solutions for problems for which the master needs a solution, but is unwilling or unable to find one alone. The assistant’s capabilities must be such that they can be tailored towards as smooth as possible an interaction with the master. A promising approach to this was suggested by Maes (1994), based on a three-valued logic (“yes”, “maybe”, and “no” being the truth values). The assistant would obtain a confidence value with respect to its current suggestion to the master. The assistant would then match this confidence value with one of the truth values, and would use for that the corresponding master-definable confidence-threshold values. Further action to be taken would then depend in the obvious way on the resulting truth value.

Intelligent assistant systems are knowledge media, which, according to Tanaka (2003, p. 11), are defined as media to “externalize some of our knowledge as intellectual resources and to distribute them among people”. Until now the archetype of knowledge media is the book. Tanaka (p. 30–31) provides interesting details regarding the history of books. While books are very effective in disseminating knowledge, they are less effective in terms of operationalising (i.e., putting knowledge to action) and reusing it, as the book is a passive medium. Tanaka points out that each knowledge medium, in addition to the knowledge represented in it, is equipped with an access method. The computer is an active knowledge medium. The access methods that can be used with respect to the knowledge stored in a computer can thus be more effective and efficient than the ones that can be used with respect to a book. Also, parts of the knowledge stored in a computer can be stored in “active form”, i.e., as a program that can actually be executed. Consequently the computer has the potential to substitute the book as the number-one knowledge medium. For that to be achieved, however, the computer’s powers must be harnessed and end-user interfaces provided. These interfaces would enable modification and reuse of the stored knowledge with high-level end-user-proof operations. This is what Tanaka and his collaborators have aimed at and achieved with their implementation of meme media. These media are highly relevant for assistance theory, as they seem to have the potential for effective reuse of existing knowledge.

Assuming two systems interacting with each other, Liebermann and Selker (2000, p. 618) define the context of an interaction step as any aspect of that system interaction which is not explicit in that step. With this definition, obviously adaptive knowledge media such as assistant systems have to be context sensitive because adaptation means changes carried out without the issue of an explicit command but only an incomplete or syntactically invalid one. As Liebermann and Selker note (p. 623), computer applications always include an application model, a user model, and a task model. Traditionally these models have been only implicit in the application’s code. They further argue that better suited for contextual computing are “systems that represent a system model explicitly and try to detect and correct differences between the user’s system model and what the system actually can do.” Winograd’s 1972 request that intelligent assistant systems should understand what they are doing seems to suggest that an intelligent assistant has an explicit and dynamic model of itself.

If one tries to identify computerized systems that implement the interlocutor metaphors mentioned above then one easily identifies tools such as case-tools, text-processors, compilers, etc as implementing the master – slave relationship. Regarding assistant systems the situation appears to be more difficult, however various works (e.g., Winograd, 1972; Marcus, 1982; Kaiser et al., 1988; Boy, 1991; Maes, 1994; Burke et al., 1997, O’Connor, 2000) show assistants that were constructed as examples for programming / software development and software project planning; email processing; meeting scheduling; retrieving bibliographic information, piloting aircraft, and browsing (selecting new cars, choosing a rental video, finding an apartment, selecting a restaurant, and configuring home audio systems). There are several assistant systems in modern cars, for example for steering, braking and accelerating. As an aid in the rather well understood area of software component installation, intelligent assistants have found quite widespread use. In this book we bring some new examples. Of course there are not only success stories. Realising intelligent automated assistance is quite difficult for domains that are not well understood, such as everyday office work. To illustrate this, Liebermann and Selker (2000) use the example of Microsoft Word’s capability of forcing the first non-blank character after a full stop to be in upper case. MS Word often, even after a user has for some reason undone that correction, re-“corrects” to the capital letter. It wouldn’t be too hard to stop MS Word from doing that. However, the vendor would have to see this as a sensible feature. Microsoft, Inc, for example, has acknowledged difficulties with their office assistant by making it easier to switch on or off. They also have changed the default for this assistant from switched-on to switched-off in the latest versions of MS Word.

According to the Oxford English Dictionary Online, an assistant is “(o)ne who gives help to a person, or aids in the execution of a purpose; a helper, an auxiliary; a promoter; also, a means of help, an aid.” This book is supposed to promote the genesis of assistant systems as gear for everyday life. According to our analysis so far, assistant systems are interactive, user-adaptive problem solving aids that understand what they do, accept input in terms of goals rather than instructions, or deduce such goals and, once these goals are identified, aim at solving them independently from their user. Assistant systems that interact with humans will often incorporate a question-answering component as well as an explanation component. We assume that intelligence can improve effectiveness and efficiency of assistants. We thus anticipate assistants to be intelligent. For us that means that available information is exploited as effectively as possible for aiding the master in achieving his / her / its goals. It is a characteristic of assistance (by information systems) that (1) the responsibility remains with the master; (2) the master may ask or cause the assistant to execute a task whenever the master wishes so or it becomes necessary; and (3) the assistant may be pro-active such that it is ready for upcoming tasks, provided the associated preparation does not interfere with the current activities (Boy, 1991).

Agent systems are similar to assistant systems. We distinguish these from each other by noting that agents are systems that have an operational mode in which they are not necessarily interactive, adaptive, or even accessible. Again querying the Oxford English Dictionary Online one finds that “agent” may be applied for referring to persons or things. The most important of the given definitions are (1) “One who (or that which) acts or exerts power, as distinguished from the patient, and also from the instrument”; (2) “He who operates in a particular direction, who produces an effect. … The efficient cause.”; (3) “Any natural force acting upon matter, any substance the presence of which produces phenomena, whether physical as electricity, chemical as actinism, oxygen, medicinal as chloroform, etc.”; and (4) “One who does the actual work of anything, as distinguished from the instigator or employer; hence, one who acts for another, a deputy, steward, factor, substitute, representative, or emissary. (In this sense the word has numerous specific applications in Commerce, Politics, Law, etc., flowing directly from the general meaning.)”. These explanations seem to be consistent with our belief that agents are different from tools, as agents act on their own in achieving complex user goals while tools require successive instructions to carry out even relatively simple tasks. It also appears to be justified to distinguish agents from assistants. Agents often do their job without much or any interaction with the user. Assistants would typically do their job in close interaction with the user, observe what he or she is doing, and propose aid that he or she is likely to find acceptable.

With this book, we continue our efforts as part of a group of researchers working to popularise assistant systems. Many of the chapters are extended and quality-assured versions of the most relevant papers presented at the PISA workshop held in August 2007 in Sapporo, Japan. These submissions were put through two further reviewing stages before being accepted here.

To widen the book’s scope beyond the horizon of the PISA-related group, we advertised on mailing lists and also directly contacted researchers who had previously published or shown an interest in this area. We obtained both new contributions, which went through our full review process, and enhanced versions of existing published work.

We now briefly summarise the book’s chapters, in alphabetical order of the first authors’ names.

Akaltun et al.

Akaltun et al start out with the observation that “[the] more unspecified content [is] published, the more general usability of the World Wide Web is lost.” They then claim that the “(t)he next generation's Web information services will have to be more adaptive to keep the web usable.” While it might seem a little questionable whether in fact computer support may enable an uneducated web user to perform highly in their searches, it certainly is worth trying to find out the boundaries of what can be achieved with computerised aid. Akaltun et al “present a new technology that matches content against particular life cases, user models and contexts.” They start out with a discussion of the term “knowledge” and then discuss an example application that matches different types of content against information available to it about various kinds of users.

Akaltun et al. address the problem that in the web, due to lack of authoritative knowledge brands, a user is frequently unable to determine the value of a number of accessible documents. From an epistemological point of view the indispensable role of authority for knowledge has been addressed by Popper decades ago. To some extent the problem may be solvable by automated means that evaluate a large part of what is available. Akaltun et al. implicitly use the aphorism, mentioned above, that information is a difference that makes a difference, by employing a five-layer model that they retrieve from the literature, and according to which information is suitably organised and connected data. Akaltun et al. then make a very important point, namely that users are imperfect and that computerised aid to some extent needs to take into account these imperfections.

For computerised web retrieval aid to be able to take into account user imperfections it is necessary to have access to a user model. According to Akaltun et al. the next generation web will be forced to focus on matching these user models and web content. Clearly, the specifics of web content as weakly structured data is going to impact any matching of web content with user models.

As Akaltun et al. do not believe in all-purpose solutions they propose an approach that focuses on ease of use, high-quality content delivery in respect of user life cases that also corrects the defects of adaptability by taking into account the construction of the web story. It is clear that, to some extent, combinatorial difficulties of matching content and user models can be solved by incorporating into that comparison the usage context. It will, however, be a challenge to cope with the complexities of combining three items reasonably rather than only two.

Erfurth & Schau

One of the assumptions of the earlier, traditional systems analysis was that the systems themselves as well as their users were stationary. In modern life that assumption is no longer necessarily true. Accordingly Erfurth and Schau drop that assumption, and consider the task of achieving “a seamless integration of users in on-line communities”. Their chapter discusses “the potential of software agents for sophisticated interaction in ubiquitous, mobile applications by utilizing the example of social networks as a practical scenario.”

Erfurth and Schau claim that humans “are on the starting blocks to integrate and share our daily life within the context of web communication channels. That is the new understanding of mobile access and modern life. Thereby, mobile communication devices and applications are primarily designed to increase efficiency and productivity, as well as to manage our rapid way of life. For many people, particularly younger users, interaction functions of cell phones, smart phones, and other handhelds are paramount (Eagle & Pentland, 2005).” Obviously this new style of life is at the same time the cause and effect of technological innovation and business opportunity. Consequently informatics cannot refrain from discussing the related phenomena.

According to Erfurth and Schau, “humans are no more able to conceive the possibilities and to utilize available services efficiently.” They thus conclude that a kind of intelligent broker is needed that would translate a human’s requests into suitable service calls. These would then result in what they call “purified information”. Obviously, heavy use of computerised media may make humans accommodate more to virtual reality. This creates the need to provide computerised aid to bridge the gap between realities. Due to their autonomous nature, “mobile software agents are in addition well suited to provide essential support in this matter”

Erfurth and Schau look into possible improvements of human communication in the web utilizing social virtual worlds by means of mobile software agents. They provide interaction schemas applying to communication between machines and humans, between machines, and between humans (through machines). Thereby communication is supported by personalized agents acting as intelligent messengers or even as humans’ substitutes within the virtual world. Humans can focus on the interaction with friends, regardless of the social network to which they belong.

Fujima

The chapter by Fujima et al. is one that deals with the challenge of providing users a simplified way to access and make use of potentially rich information resources, in this case by supporting the combination of resources offered by Web applications. Despite the recent emergence of numerous technologies that enable the “mashing up” of Web-accessible data sources (such as use of the Google Maps API to map announcements of house sales, or of crime reports), there is a tendency for the technologies either to require sophisticated programming skills (as is the case with Google Maps) or to be constrained to information delivered in a form explicitly designed for merging with other sources, such as RSS feeds. For the kind of ad hoc combinations of resources that users call on every day – such as taking a price from a shopping site and feeding it to a currency-conversion application – it is still difficult for users without advanced programming skills to set up even the simple form of assistance involved in automating that data transfer, so as to save time and effort in subsequent use of the same resource combination. Fujima et al.’s proposal is to enable the use of a popular spreadsheet environment – Microsoft Excel – as a platform for such automation, on the basis that a large body of computer users, though not considering themselves to be programmers, find the simple declarative semantics of spreadsheet calculations to be unintimidating and useful. The result is a set of custom extensions to Excel, demonstrating a basic range of facilities that enable users to set up “orchestrations” of Web applications that not only automate the data transfers between applications but also provide a way to generate sets of related results (such as overall purchase costs from a range of competing shopping sites) to assist the user in making comparisons.

Although for simplicity of explanation the examples in this chapter address only straightforward lookups on shopping sites and bioinformatics repositories, the potential power of this platform becomes apparent when one considers that it can be applied equally to Web applications that incorporate intelligent, adaptive techniques. It is now common for sites to keep long-term records of the activities of individual users, enabling the sites both to adapt the information displayed according to an individual user’s history of use, and to pool these histories in support of techniques such as collaborative filtering, thereby driving suggestions and recommendations as to what each user might like to try or buy next. Users who add such applications into the spreadsheet-driven mix stand to benefit from a dramatic new level of customised access to the Web resources that interest them.

Ghiassi & Spera

In their chapter, Ghiassi and Spera deal with the challenges of increasing product customisation and customer requirements, demanding new strategies for “Mass Customisation”. Providing an architecture for a web-enabled, intelligent agent-based information system to support on-demand and mass customised markets such as car manufacturing, their approach is based on the assumption of a customer-centric production system involving many businesses each of which focuses on its core competencies, resulting in products that are outcome of a collaborative effort. In fact, they envision the idea of users assisted by IT applications, namely distributed and specialized software agents. Based on a taxonomy that accounts for the core functionalities (monitoring, analyzing, acting, and coordinating) of assisting entities and sufficient roles in the value chain, these agents systematically “address the needs of end users, manage product demands, and coordinate supply and resource management continuously”. Accordingly, the authors introduce an architecture for the dynamic instantiation and interaction of the agents and give examples of their approach through a number of case studies. With an interdisciplinary combination of economic demands and computer scientific solutions, they encourage efforts to consider user-centric issues in production chains. Moreover, the case studies, although relying to some extent on simulation results, clearly show how to assist users such as end-customers in IT supported environments.

Guo

Similar to Fujima et al., Guo and Tanaka are reporting work designed to enable non-sophisticated programmers to build or configure their own specialised information-access applications, in this case in the domain of Geographic Information Systems. The form of assistance being demonstrated here is that while the end user performs only simple operations such as visually repositioning and/or reconnecting knowledge-media components, these components assemble and transfer rich data structures to drive complex simulation tools and display advanced 3D simulation results. The key to enabling this is that the components conform to local standards, established by the framework developer, for accessing a limited but useful subset of the capabilities of the underlying simulation tools. It is the framework developer’s careful choice of the capabilities to be made available, and a simplified inter-component communication model based on custom-designed data formats and a flow-graph-style protocol for update propagation, that is crucial in obtaining a good balance between ease of use and breadth of capabilities. Guo and Tanaka explain in detail two application examples that have been built on their framework, demonstrating how it is possible to replace the key simulation elements and thereby obtain, on the same three-dimensional GIS-enabled substrate, the modelling of quite different phenomena. This form of framework can be seen as offering at least two levels for applying intelligence-supported assistance: one is above the level of the components, for example augmenting the user’s simple construction operations with automatic sensing and resolution of mismatches in data formats and protocols; another is below the components – in other words, building components that offer simplified access to assistant services. Both can be seen as potentially profitable extensions of the principles demonstrated here.

Krumpholz et al


Krumpholz et al. point out the crucial role of librarians in days gone by, when “Librarians … fetched the needed materials and passed them out to the reader. … Later, librarians got access to computers and translated readers’ needs into queries to retrieve the meta-information about the requested literature. Beyond their retrieval tasks, experienced librarians also helped students to identify popular books or drew their scientists’ attention to new publications in their field.”, and ask whether, now that most researchers have access to search engines through their own computers, the literature retrieval services available to them match up to the job being done by librarians.

They describe a number of currently existing search aid systems and research efforts to improve literature retrieval, focusing on literature retrieval for computer scientists, and on medical literature retrieval for clinicians or medical researchers.

The authors note that the need for literature retrieval depends on the role and the field of the researcher as well as the task they wish to achieve. In order to better understand search tasks, Krumpholz et al. identify a number of search purposes that might warrant a specialised approach to literature search. These are: (1) Finding a specific publication, (2) Finding related publications, (3) Staying up to date in the field, and (4) Entering a new field. Krumpholz et al. additionally address medical research as a particular domain for study – a focus that makes sense, as medical literature search is a particularly frequently occurring and important task in a problem area with a highly standardized terminology. As a guiding metaphor for understanding search processes, Krumpholz et al. use an ad hoc interpretation of Web retrieval processes. They assume that a query provided by a user contains a number of search terms and that as a response a ranked list of matching documents is returned. A commonly used strategy of dealing with that response is skimming the first few results and, based on the acquired information, altering the query and searching again. This cycle of output creation and limited-scope evaluation of it can then be repeated for as long as the results seem to be improving, or until the strategy clearly fails. The key query change goals are (1) improving output quality; and (2) adapting output size. Employing literature retrieval systems differs from Web search in that it frequently exploits specific metadata for improving output quality. In particular this shows in using links to further publications of a given author.

Krumpholz et al. point out that in Web search one usually gets access to a large number of documents, which creates the need to check them for relevance; that in itself implies further potential for aiding search.

Minato

The work introduced in the chapter by Minato and Spyratos stands as an enabling technology for letting users express, in simple ways, their information interests as handled in a large digital library organised by a rich, hierarchical taxonomy of terms. The key to the technique is Binary Decision Diagrams, a way of encoding complex logic formulas into a compact representation by taking advantage both of the sparseness of the formulas relative to the total number of possible terms, and of the terms shared between formulas that are being calculated in parallel. By automatically taking into account the structure of the taxonomy of keywords used to tag the documents in the library, the system proposed by Minato and Spyratos allows users to build subscriptions based on a free mix of highly specific and more general terms, depending on the specificity of their interests. In one of the examples that they give, a user who is interested in documents dealing with a specific sorting algorithm could subscribe using that specific term, whereas a user who is interested in sorting algorithms in general could subscribe at that higher level without needing to worry about the sub-categories that this encompasses; not only will the subscription ensure that the latter user receives notification of every document dealing with any such sub-category, but if the taxonomy is later changed all subscriptions can be automatically recompiled so that they continue to deliver all documents that the users were expecting.

The need to be able to expand simple queries given by users into the complex combinations of keywords that will deliver what the users had in mind is clearly demonstrated in the medical examples cited in the chapter by Krumpholz et al.; this chapter gives insight into how such expansions can be handled without overwhelming the servers on which the repositories are held.

Sturm & Igel

The chapter provided by Sturm and Igel presents an empirical study regarding the utilisation of a knowledge management system in the tertiary education sector. The KMS was designed to meet the perspectives of both learners and teachers. It therefore provides so called teaching-learning modules to be experienced by the learners, as well as a multimedia database storing the digital material to support lecturers during their apprenticeship. By means of the KMS, “digital information and knowledge objects can be retrieved and selected, collected in a separate download domain and, where required, archived and downloaded. Alternatively, users can make digital information and knowledge objects …created [by] themselves available to the (scientific) community”. Thus the two functionalities “search” and “submit” essentially characterize the use of the KMS. From an evaluative perspective, the authors examine the impact of the KMS, i.e., its utilisation and potential barriers preventing such utilisation, and present result from two studies. In a first, exploratory study they identified typical behaviour patterns and search strategies, looked for difficulties in use and attributed them to a set of personal and non-personal barriers. A second, quasi-experimental study was conducted to determine the impact of utilisation barriers on usage patterns. Sturm and Igel were able to confirm their assumption that a variety of aspects prevent existing as well as potential users from using all capabilities offered by the KMS. According to this finding the authors also show that having complete technical equipment, network-connected computers or Internet access does not automatically guarantee a new level of quality in teaching and learning. Interestingly, their studies show that system training, i.e. retrieval training, “helps system novices with a low level of task-solving knowledge when completing specific, detailed set tasks more than it does system novices with a high system competence”. Going into more detail they found that “course participants with a low system expertise tend to be more motivated and focused, and therefore probably handle tasks more conscientiously. They ask more questions in the scope of the system training and assure themselves of the fact that their user identification is still active at the end of the test phase”. Hence, their studies clearly give empirical evidence that even untrained users can benefit from assistance offered by training courses and KMS. However, these findings do not conflict with but clarify (at least some of) the assumptions made by Akaltun et al., in their chapter on the idea of a knowledge web, regarding limited user capabilities.

Takashima

Takashima’s chapter presents a system for producing customised presentations of videos, in which some parts of the video are emphasised relative to others. The form of emphasis supported by the current version is the automatic use of the playback controls (including slow-motion and fast forward) to slow down for sections that are deemed to be especially interesting, and to skip through the sections of least interest. Takashima’s view is that this can be seen as an assistant-like functionality whereby the system recognises and reproduces the user’s habitual behavior, so that once the system has been appropriately trained it will automatically show newly arrived videos in the way the user would probably have wanted to watch them anyway. The main example used for his experiments is segments of football games, which form a convenient object of study because the play tends to be divided relatively cleanly into scenes of different types (goal shots, game pauses, etc), and camerawork conventions ensure that these differences are quite straightforward to detect. Takashima shows that because his system represents viewing styles as knowledge objects that can be communicated from one user to another, and can be manipulated so as to produce various forms of composition of alternative viewing styles, users are able to benefit from the use of viewing styles defined by others – such as, in the football example, a style that would be used by a football coach when demonstrating or analysing a team’s tactics.

Roland Kaschek, Gunther Kreuzberger, & Aran Lunzer
December 2009

REFERENCES

Ahituv, N., Neumann, S. (1990). Principles of information systems fore management. Dubuque: Wm. C. Brown Publishers, third printing.
Banker, R., Datar, S., Kemerer, C., & Zweig, D. (1993). Software complexity and maintenance costs. Communications of the ACM, 36(11), 81 – 94.
Barstow, D., Shrobe, H. & Sandewall, E. (Eds.) (1984). Interactive programming environments. New York et al.: McGraw – Hill.
Bateson, G. (1979). Mind and nature: a necessary unity. New York: Dutton.
Boy, G. (1991). Intelligent assistant systems. Academic Press.
Burke, D., Hammond, K. & Young, B. (1997). The FindMe approach to assisted browsing. IEEE Expert, 12(4), 32 – 40.
Buschmann, F., Meunier, R., Rohnert, H., Sommerlad, P. & Stal, M. (1999). Pattern – oriented software architecture: a system of patterns. Chichester et al.: John Wiley & Sons, 4th printing.
O’Connor, R. (2000). An architecture for an intelligent assistant system for use in software project planning. PhD thesis, Department of Computing, City University of London.
Engelbart, D. (1962). Augmenting human intellect: a conceptual framework. Excerpt from Summary report AF0SR-3223 under contract AF 49(638)-1024, SRI Project 3578 for Air Force Office of Scientific Research, Menlo Park, California. Reprinted as (Wardrip-Fruin, N., Montfort, N., 2003, pp. 95 – 108).
Engelbart, D., English, W. (1968). A research centre for augmenting human intellect. AFIPS conference proceedings 33, part 1, 395 – 410. Fall Joint Computer Conference. Reprinted as (Wardrip-Fruin, N., Montfort, N., 2003, pp. 233 – 246).
Enzensberger, H.(1970). Constituents of a theory of the media. New Left Review, 64(Nov / Dec), 13 – 36. Reprinted as (Wardrip-Fruin, N. and Montfort N., 2003, pp. 261 - 275).
Fleck, L. (1935) Entstehung und Entwicklung einer wissenschaftlichen Tatsache. Einführung in die Lehre vom Denkstil und Denkkollektiv. Schwabe und Co., Verlagsbuchhandlung, Basel.
Garmus, D. & Herron, D. (1996). Measuring the software process: a practical guide to functional measurements. Upper Saddle River: Prentice Hall.
Haase, K. (1997). Do agents need understanding? IEEE Expert, 12(4), 4 – 6.
Häuble, G., Trifts, V. (2000). Consumer decision making in online shopping environments: the effects of interactive decision aids. Marketing Science, 19(1), 4 – 21.
Hevner, A., Berndt, D. (2000). Eras of business computing. Advances in Computers, 52, chapter 1. Academic Press.
Hirschheim, R., Klein, H., & Lyytinen, K. (1995). Information systems and data modeling: conceptual and philosophical foundations. Cambridge, UK: Cambridge University Press.
Ifrah, G. (2000). The universal history of numbers III: the computer and the information revolution. London: The Harvill Press.
Kaiser, G., Feller, P., & Popovich, S. (1988). Intelligent assistance for software development and maintenance. IEEE Software, 5(3), 40 – 49.
Kaschek, R. (2006). Intelligent assistant systems: concepts, techniques and technologies; IDEA Group Publishing, Inc.: Hershey, PA. 2006.
Kuhlen, R. (1999) (In German). The consequences of information assistants. Frankfurt: Suhrkamp Verlag.
Liebermann, H., Selker, T. (2000). Out of context: computer systems that adapt to, and learn from context. IBM Systems Journal, 39(3/4), 617 – 632.
Maes, P. (1994). Agents that reduce work and information overload. Communications of the ACM, 37(7), 31 – 40, 146.
Marcus, R. (1982). User assistance in bibliographic retrieval networks through a computer intermediary. IEEE Transactions on Systems, Man, and Cybernetics, March/April, 116 – 133.
Ong, W. (1996). Orality and literacy: the technologizing of the word. London and New York: Routledge, 7th. printing.
Sandheimer, N., Relles, N. (1982). Human factors and user – assistance in interactive computing systems: an introduction. IEEE Transactions on Systems, Man, and Cybernetics, 12(2), 102 – 107.
Satzinger, J., Jackson, R., & Burd S. (2004). Systems analysis and design in a changing world. Thomson Course Technology, 3rd edition.
Shannon, C. E., (1993). A Mathematical Theory of Communication. In Sloane, N. J. A., Wyner, A. D. (Eds.), Claude Elwood Shannon, Collected Papers, (pp. 5–83). New York: IEEE Press.
Sowa, J. (2002). Architectures for intelligent systems. IBM Systems Journal, 41(3), 331 – 349.
Tanaka, Y. (2003). Meme media and meme media market architectures: knowledge media for editing, distributing, and managing intellectual resources. IEEE Press and John Wiley & Sons, Inc.
Wardrip-Fruin, N., Montfort N. (2003). The new media reader. Cambridge, London: The MIT Press.
Winograd, T. (1973). Breaking the complexity barrier (again). Proceedings of the ACM SIGPLAN – SIGIR interface meeting on programming languages – information retrieval, November 1973, 13 – 30. Reprinted as chapter 1 of (Barstow, Shrobe & Sandewall, 1984).
Winograd, T. (1979). Beyond programming languages. Communications of the ACM, 22(7), 391 – 401. Reprinted as chapter 25 of Barstow et al. (1984).