Situated cognition and knowledge acquisition research
Highlights
► Real-time and post-drive systems were evaluated using the Technology Acceptance Model. ► Ease of use is the primary determinant of a system's value. ► Unobtrusiveness affects a system's value only through usefulness and ease of use. ► Real-time systems are less pleasant and less easy to use. ► Older adults found both systems more useful than younger adults.
Introduction
The difficulties in building expert systems led to a philosophical and cognitive science analysis of why it was so difficult obtaining knowledge from experts, largely from a situated cognition perspective (Winograd and Flores, 1987, Clancey, 1997). Situated cognition offers a broad-ranging perspective, applying particularly to education, but the key idea in relation to expert systems, can probably be summarised as simply that experts never describe how they reach a conclusion, rather the knowledge they express is a justification for their conclusion created for the particular context. In the late 80s and early 90s there was intense discussion about situated cognition at the knowledge acquisition workshops but this largely philosophical discussion soon faded. As Menzies comments:
“It could also be argued that a philosophical perspective on human reasoning has little relevance for tool builders such as pragmatic knowledge engineers (Menzies, 1998)”.
The consequent history of the knowledge acquisition workshops and conferences has been largely and appropriately a history of trying to develop engineering approaches. The intention of this paper is to briefly consider some of the major strands of this research and the impact or lack of impact of a situated cognition perspective on this research.
Section snippets
Situated cognition
I was involved in the development of one of the first medical expert systems to be in routine clinical use and had the task of maintaining it—and in four years the knowledge base doubled in size while the accuracy went from 96% to 99.7% (Compton et al., 1989). Encouraged by Mildred Shaw whom I met at the 1988 5th Generation Conference in Tokyo and Bill Clancey's keynote on situated cognition at the Australian AI conference that year, I submitted a paper to the 1989 EKAW which was an attempt at
Situated cognition as a positive statement
The ideas of situated cognition have largely been presented as stating what people cannot do, but it is probably more useful to re-express this as what situated cognition has to say about what people can do. Situated cognition suggests that people never explain the process of how they reach a conclusion; rather they create explanations to justify their decision in the particular context of the decision. Clancey went as far as saying that there is no knowledge in the mind, knowledge only comes
Software engineering and modelling
A central theme at the workshops through much of the 90s was the notion of modelling problem solving. Rather than simply asking the expert for their knowledge, research focussed on identifying the types of problems that occurred and the different problem-solving methods that could be applied to problems. The idea was that the knowledge engineer would come to a problem armed with a library of re-useable methods (whether fully specified and coded or not). Chandrasekaran initially suggested
Ontologies and the semantic web
The idea of a library of problem-solving methods went hand in hand with a more systematic approach to knowledge representation independent of any particular domain, and this became a dominant theme at the workshops. One concern has been the theoretical bases for ontologies and metadata and more generally for logically sound reasoning and representation and Gaines predicts advances here will be of major importance in the development of the semantic web (Gaines, 2012). Of more immediate concern
Knowledge creation and elicitation
There are a range of long-standing techniques that assist in eliciting knowledge but I will focus here on techniques based on Kelly's Personal Construct Psychology (Kelly, 1955). These ideas were introduced to the knowledge acquisition community by Gaines and Shaw, 1980 and Boose, 1984. The central technique in Personal Construct Psychology is to ask a person to select three objects in a domain of interest and then ask in which way two of these objects are alike and different from the third.
Ripple-Down Rules (RDR)
There has been a series of RDR papers presented at the knowledge acquisition workshops over the years, particularly at the Pacific Rim Knowledge Acquisition Workshops, covering a wide range of different research, but it does not qualify as a major theme, as the papers have been presented by a fairly small sub-community. A review of much of this work can be found in (Richards, 2009). The reason for presenting RDR here is because of the way it relates to situated cognition.
RDR was motivated by
Extracting knowledge from data
I include here both machine learning and the wide range of research presented over the years on areas such as information extraction from text. Such research avoids the problems of situated cognition and the disconnect between elicited knowledge and data by learning from data so that the resulting system can be applied to more examples of the same type of data. Such systems are not a total panaceas as their development is always limited by the range and appropriateness of training data.
Conclusions and future possibilities
Although situated cognition was a significant focus of discussion in the early years of the knowledge acquisition workshops, this interest largely faded as the focus moved primarily to comprehensive frameworks to deal with the problems of representation and reasoning and overall software engineering rather than actual acquisition.
When situated cognition is re-expressed as a minimum capability of what people can do, it emphasises that they can and do differentiate concrete situations and
Acknowledgements
I suspect that I must be near the record for participation in the knowledge acquisition conferences and workshops, as I have participated in 29 of the 50 events to date. My on-going involvement has been largely because of the welcome and encouragement I received from Mildred Shaw and Brian Gaines.
Pacific Knowledge Systems provided the data shown in Fig. 1. The Australian Research Council has supported research on Ripple-Down Rules over many years.
References (33)
Heuristic classification
Artificial Intelligence
(1985)- et al.
Brahms: simulating practice for work systems design
International Journal of Human Computer Studies
(1998) - et al.
New directions in the analysis and interactive elicitation of personal construct systems
International Journal of Man-Machine Studies
(1980) - et al.
The evolution of Protégé: an environment for knowledge-based systems development
International Journal of Human Computer Studies
(2003) Towards situated knowledge acquisition
International Journal of Human Computer Studies
(1998)- et al.
Interface terminologies: facilitating direct entry of clinical data into electronic health records
Journal of the American Medical Informatics Association
(2006) - et al.
Validation in a knowledge support system: construing and consistency with multiple experts
International Journal of Man-Machine studies
(1988) - Boose, J.H., 1984. Personal construct theory and the transfer of human expertise. Proceedings AAAI-84,...
Towards a taxonomy of problem solving types
AI Magazine Winter/Spring
(1983)Situated cognition: on human knowledge and computer representations
(1997)
Maintaining an expert system
A philosophical basis for knowledge acquisition
Knowledge Acquisition
Knowledge Acquisition: Past, Present and Future
International Journal of Human Computer Studies
Cited by (17)
What Shapes E-Learning Effectiveness among Tourism Education Students? An Empirical Assessment during COVID19
2022, Journal of Hospitality, Leisure, Sport and Tourism EducationCitation Excerpt :Virtual Social skills have a direct and significant impact on e-learning experience. Knowledge acquisition has been defined as obtaining knowledge from specialists on the condition of situated cognition (Compton, 2013). Knowledge acquisition includes the search for, identification of, and access to new relevant knowledge (Alavi & Leidner, 2001).
A chronic disease dietary consultation system using OWL-based ontologies and semantic rules
2015, Journal of Biomedical InformaticsCitation Excerpt :Fürst and Trichet [36] also proposed a scenario-based approach to integrate declarative knowledge into new domain ontologies. This modeling transition step is necessary because the general ontologies are too declarative to be directly used in a KBS for problem solving [26,38,39]. This study use ontological engineering for knowledge modeling.
Modeling for effective collaboration in telemedicine
2015, Telematics and InformaticsCitation Excerpt :In the Knowledge Acquisition (KA) community, there is the rise of situated cognition. There is growing insistence that much (perhaps all) of human decision making is critically dependent on the context in which it took place, suggesting that human problem solving could never be divorced from the myriad of environmental variables from every angle (Compton, 2012). The research emphasis here highlights the concern about constructing coherent, robust models of problem solving.
Effective Knowledge Dissemination Modeling and Regulation in Blended Learning Networks
2024, IEEE Transactions on Computational Social SystemsReflective mediation: Toward a sociocultural conception of situated reflection
2020, Frontline Learning ResearchGenres and situational appropriation of information: Explaining not-seeking of information
2019, Journal of Documentation