The Intelligent Interfaces and Knowledge Communication group studies how IT-based artifact design can be tuned to human user's cognitive capabilities and patterns of communication and how the introduction of such tools influence the scenario of use from a cognitive and communicative point of view. Another important aspect of the work is to let models and theories of human cognitive processes inspire design principles for such artifacts. Techniques from areas like artificial intelligence, human-machine interaction and natural language processing technology are combined.
_______________________________________________________________________________________________________
_______________________________________________________________________________________________________
A system may adapt its behaviour to the human users in different ways:
Common to all these forms of adaptation is the problem of what to adapt to and how to adapt to it. It is partly an issue of design methods and methods for analysing the domain or functionality and tasks of the system. It is also an issue of providing more knowledge on what aspects of human behaviour that are at all relevant for adaptation, and which kinds of adaptations will actually meet the perceived needs of the users. Much research is needed both to determine how adaptivity can be mastered technically and to determine how users accept, adapt to and adapt the adaptivity features. Some research questions related to these issues concern:
These questions relate to basic issues of technological methods as well as basic issues of work task analysis, general learning issues and individual differences. On a deeper level, the work on adaptivity requires a deeper understanding of human cognition, and learning in particular. Emerging views of cognition, such as situated cognition, constructive cognition, distributed cognition, etc., challenge the previously dominating symbolic view of cognition. The users may not be as goal-oriented and rationalistic as some of the adaptive features of proposed systems require. We need to further our knowledge on human cognition in order to find good adaptive features that will, in fact, correspond to human behaviour, increase the usability of systems, and be computationally feasible to implement.
The thesis focuses on the uses of plan recognition in different types of interface applications. Based on a set of application examples, the thesis first try to partition these into different classes of applications, and discuss the problems and challenges of each separately. In particular, this part of the thesis put forward the claim that the reactive nature of human behaviour in interfaces must be dealt with both by the development of appropriate algorithmic solutions, but also by the design of interfaces that allow more explicit interaction about the user tasks and goals. The second part of the thesis focuses around principles for knowledge representation and inference strategies geared to deal with reactive user behaviour. This is divided into three main topics:
When designing adaptive systems, it is of crucial importance that the adaptivity is given a form that is acceptable to users. Depending upon the task of the adaptive system, it might be more or less important that the user is given control over the adaptivity. This might be achieved through making the system transparent (the glass box) to any of its more complex parts (the black boxes), and through making the adapted behaviour predictable to the user. The thesis shall describe how a good level of transparency has been achieved for one particular domain, and how different more or less adaptive techniques has been designed, implemented and studied in a comparative study. _______________________________________________________________________________________________________
Human communicative behavior is influenced by what communicative mechanism, what perceptual mode, or what channel of communication it is performed in: people communicate differently over the phone, compared to face-to-face communication, and compared to pen and paper. The character of the communication is also influenced by temporal factors, such as how interactive or immediate a communicative situation is: writing is different in a newspaper from that in an electronic mail message; spoken messages are different over a voice mail system than over a live phone connection. Another major factor is what expectations a participant in a dialog has on the competence and commitment of the counterpart: communicating with a computer is different than communicating with a human friend. All of these factors influence the setup of a communicative situation, and their effects are more or less well documented from studies both on human-human communication and human-machine interaction. This research track is designed to study the interaction of such factors, and to construct and examine interactive situations which utilize new combinations of communicative and perceptual modes such as pictures, graphs, and gestural interaction; new characteristics as regards degree of interactivity and other temporal and synchronal characteristics; new attempts to influence participant expectations of the communicative competence and commitment of the counterpart or counterparts.
What mechanisms do people use to show and track focus in discourse, specifically in different multimodal settings? The object of study is variation in and varieties of language, as determined by individual, situational, and social factors, and as evidenced by surface cues perceptible in language use. The assumption is that linguistic and communicative behavior is strongly determined by local and situational factors - turn taking mechanisms, characteristics of available communicative channels, extralinguistic cues -- and that the higher level regularities observable in language to a large extent follow from local processes. In the present set of studies, language use is investigated in a multimodal setting: in a text with accompanying picture frames (Tintin), in an interactive computer system which combines direct manipulation with text input (PUSH), and an interactive computer system which combines speech input and output, text input and output, gestures and 3-D graphics (DIVERSE). The data examined are cross-modal references in collected corpora - how are depicted or analogically represented objects referred to, and how can they be referred to, or reference to them inferred, in text or speech.
The approaches to multimodality detailed in this thesis are based on the idea of anchoring the interpretation of natural language utterances in model world environments. The model world considered take the shape of two-dimensional abstract visual languages an a three-dimensional concrete rendering of an artificial reality. This thesis presents three interface architectures for the integration of natural language interaction into model world interfaces. «V«, a visual query language provides a separate model world for natural language database querying which in effect is just a visual language for logico-linguistic representations. VINST, a specification environment, combines natural language and a visual language describing conditional transitions within finite state automata in a more integrated fashion than in V. DIVERSE, finally, took the idea of visually supported linguistic interaction into three dimensions, and made it ossible to act in an artificial rendering of a physical and metaphorical «real«world both through direct manipulation and language.
The overall research goals are to study the issues that arise when trying to build a state-of-the-art natural-language system in a logic-programming environment, and using it for serious applications. These involve exploring methods and tools for developing NL systems, as well as the design of the system itself, taking the applications at hand into consideration. A general-purpose NL system may be used for a wide range of applications and must therefore have a broad functionality, including the ability to transform between utterances in NL and database queries, transform between NL and machine-translation transfer format, maintain records of discourse, and to infer and reason about the user's goals. From a development point of view, we consider using a standard high-quality kernel system as the central natural-language processing unit and adapting it to the various applications to be the best strategy. This particular thesis focus on the adaption of the kernel system to achieve at a robust «swedish«version. The emphasis lies on issues of swedish grammar and morhology.
This thesis is concerned with techniques for aggregation in generation of natural language discourse from formal models such as conceptual models or knowledge bases. On one hand technical contributions is made to aggregation techniques such as the assembling of an appropriate set of aggregation rules, heuristics for ordering of the application of generation rules and disambiguation of generated text. The thesis also describes several applications where sentence and discourse generation have been applied, such as a system for validation of conceptual schemas and the VINST specification tool. _______________________________________________________________________________________________________
The study of different conceptual models held by humans and how these models are developed over time. That is, knowledge regarding how individuals understand, learn and develop their explanatory models regarding a certain domain or system. Such knowledge is fundamental to system design regardless of the type of HMI-system, e.g. CSCW systems, knowledge based systems, and tutoring systems.
This thesis focuses on the problem of designing for organisational control, routiniza tion, and effectiveness, on the one hand, and for flexibility, local, autonomy and indi vidual empowerment, on the other hand. With a focus on ordering procedures for exter nal customers, a version of the generic schema of conversation-for-action [Winograd and Flores] is extended, e.g. with ideas from conversation analysis. The basic perspec tive is to view the communication partner as an autonomous and competent agent, guided by organisational rules and constraints in various ways. The core part of the work suggests a set of metaphors for describing the functional role of information tech nology in this context.
The thesis is focused upon the study of how individuals' explanatory models are devel oped and refined in the course of learning a specific domain. Knowledge regarding how individuals understand, learn and develop explanatory models regarding a certain do main or system, is crucial in a context of providing intelligent help and good explana tions that fulfil specific needs of users in a context of computer supported education, for instance. A related research topic focuses upon how combinations of modalities (explanation alternatives) can / should, or can - / should not be combined in a specific learning environment, for instance. That is, how combinations of different modalities affect individuals perceptual and cognitive performance.
This work focuses on conceptual change in learning and understanding, on how certain concepts in certain domains are viewed in different ways by different individuals, or groups of individuals. Knowledge of users' understanding and learning of key concepts is crucial to efficient design of usable and adaptable systems. The knowledge a novice possesses tends to be of a fragmented nature and different in dividuals may categorize concepts into incompatible ontological categories rendering learning more difficult. Language, or different language games, may also strongly in fluence the reasoning around a specific problem. In a project (PUSH) different under standings of a development process are currently being studied. Knowledge of these understandings will affect the design of a help system for the development process. _______________________________________________________________________________________________________
This research theme is concerned with the design of personal software assistants (PSAs) who is collaborating with the user in the
same work environment. A personal software assistant (PSA) is a software realisation of an intellectual partner of a users for a specific task. The personal assistant metaphor provides a rich and powerful connection between the user interface and the capabilities of an underlying agent architecture. It can support various forms of user interaction: search through navigation and direct manipulation, the formulation of specific search requests, and more advanced
user - agent collaboration for interacting with user-specific functionalitites such as information filters and proxy functionalities in agents. The integration of vivid and active personal assistants in a stable, non-intelligent and non-adaptive environment such as a direct manipulation or Virtual Reality interface allows a rich multimodal interaction scheme. A number of issues have to be addressed:
_______________________________________________________________________________________________________
_______________________________________________________________________________________________________
This page is maintained by Carl Gustaf Jansson _______________________________________________________________________________________________________