CHAPTER 35
Could the Use of a Knowledge-Based System
Lead to Implicit Learning?
University of Kentucky, Lexington, KY, USA
The primary objective of a knowledge-based system (KBS) is to use stored knowledge to provide support for decision-making activities. Empirical studies identify improvements in decision processes and outcomes with the use of such knowledge-based systems. This research suggests that though a KBS is primarilydeveloped to help users in their decisionmaking activities, as an unintentional consequence it may induce them to implicitly learn more about a problem. Implicit learning occurswhen a person learns unconsciously or unintentionally, without being explicitly instructed or tutored. To test these ideas, a laboratory-based experiment was conducted with a KBS that could provide support for datamodeling activities. Results indicated support for implicit learning because subjects who interacted with the KBS exhibited better nowledge on data-modeling concepts. Two versions of the KBS were tested, one with a restrictive interface and the other with a guidance interface, and both versions ofthe interface supported implicitlearning. Implications for
future research on the design and development of KBSs are proposed.
Keywords:Learning; Interface; Knowledge-based tool; Data modeling
1 Introduction
Knowledge based systems (KBSs) are developed to improve their users’ decisionmaking and problem-solving capabilities. A KBS is defined as a systemthat uses stored knowledge of a specific problem to assist and provide support for decisionmaking activities related to the specific problem context (Holsapple and Whinston 1996, Keen and Scott-Morton 1978). KBSs have been developed and used for a variety of applications, including database design activities, with controlled experiments showing that KBSs, as decision-aiding tools, can alter the decision outcomes, processes, and strategies of users as they engage in tasks (Dhaliwal and
Benbasat 1996, Santhanam and Elam 1998, Storey and Goldstein 1993). Conse-1Reprinted from Decision Support Systems, Volume 43 Issue 1, Solomon Antony and Radhika Santhanam, with permission from lsevier.
Solomon Antony and Radhika Santhanam quently, the primary emphasis on KBS research has focused on the role of a KBS as a decision-aiding tool and the intended consequences of improvements in users’ decision processes and outcomes. In this study, it is proposed that a KBS can play yet another role; it can be an agent of change to improve the user’s knowledge. When a user interacts with a KBS and obtains help in solving a problem, the user may learn more aboutthe problem and thus implicitly acquire knowledge. Implicit learning occurs when the user applies no deliberate or intentional effort to learn, but learning still occurs
unconsciously (Berry and BroadBent 1984, Berry and Dienes 1993, Prabhu and Prabhu 1997). Implicit learning differs from conscious and directed learning that might occur with knowledge repositories or with tutoring systems that arespecifically developed to teach students (Alavi and Leidner 1999, Anderson et al. 1985, Holsapple 2003). The objective of tutoring systems is to teach the user how to acquire knowledge about the problem area. These systems typically test the user’s initial knowledge level and then teach in an approach similar to methods an instructor would use. Knowledge repositories discussed in the context ofknowledge management systems store vast amounts of knowledge and allow users to consciously access this storehouse whenever it is needed.
As opposed to the above systems, it is proposed that decision support/aiding systems, such as computer-aided software engineering tools that are primarily designed to assist a decision maker, when embedded with knowledge, may induce users to learn more about problems as they interact with the system. To test these ideas a laboratory-based experiment, using theoretical perspectives from implicit
learning and a KBS designed to support database design activities, was conducted. Database design is a complex task, and many knowledge-based tools have been proposed to support this activity (Batini et al. 1992, Lo and Choobineh 1999, Storey and Goldstein 1993). This study used a KBS that had embedded knowledge on data modeling and could assist novice users to complete data-modeling tasks (Antony and Batra 2002). Two versions of this KBS were test, each of which interacted differently with the user. One version had a restrictive interface, as it forced
the user to follow a specific decision strategy in developing a data model,while the other version had a guidance interface, because it offered suggestions to help users complete their data-modeling tasks (Schneiderman 1992). When users interacted with these versions of a KBS, the level of their learning was determined and compared with the learning of users who interacted with a control system that had
no embedded knowledge on data modeling. Results suggest that KBSsmay indeed induce users to implicitly acquire more knowledge.
2 Knowledge-Based Systems and Implicit Learning
In one of the earliest works on decision-aiding tools, decision support systems (DSSs) were defined as coherent sets of computer-based technology thatmanagers can interact with and use as aids for their decision-making activities (Keen and Morton 1978). This definition has spurred a tremendous amount of research in improving the functionalities of DSSs and understanding their impact on users’ decision-making activities (Elam et al. 1996). Research studies and reports from practice
indicate that, indeed, the use of a DSS can lead to substantive improvements in decision-making outcomes and processes (Sharada et al. 1998, Todd and Benbasat 1991). Many research avenues are pursued in DSS research, with one path of inquiry focusing on improving the functional capability of a DSS by embedding it with knowledge of the problem area. These systems are often referred as KBSs or intelligent DSSs (Goul et al. 1992, Holsapple and Whinston 1996). Decision making is a knowledge-intensive activity where knowledge of a particular problem area is used to understand and make choices during the decision process. Hence, including a knowledge base in a DSS can be very advantageous in that the system can interject and provide necessary knowledge at the appropriate points in the decision process (Holsapple and Whinston 1996). In addition, researchers in the humancomputer interface field recommend that users of systems be less burdened with
cognitive load, wherever that knowledge inputs can come from within thesystem (Norman 1998, Schneiderman 1992). For example, a KBS for supporting manufacturing planning activities may suggest to the user a method to reduce the setup time for manufacturing when the user is engaged in developing a manufacturing plan. The decision maker is provided this information showing how to reduce setup time based on the expertise that is embedded in the knowledge base as rules (Kimand
Arinze 1992). KBSs are used in many diverse applications such as financial planning, manufacturing, tax planning, equipment design, etc., and are more useful, in fact, than expert systems that attempt tototally replace the decision makers (Goul et al. 1992, Santhanam and Elam 1998, Wong and Monaco 1995). Another research avenue has used changeagency perspectives to investigate how design attributes could influence users’ decision choices. A DSS could be designed to restrict users’ choices and lead them through a specific decision strategy,
or it could suggest possible decision choices and thus allow the user to follow a certain strategy (Schneiderman 1992, Silver 1990, Silver 1991a, Silver 1991b). A DSS interface designed with a restrictiveness approach limits a user to a subset of all possible decision-making options, while a system with a decisional guidance approach guides its users by advising and assisting them in choosing decision options. These design principles could also be used to develop a KBS in that the system could use the
embedded knowledge to provide guidance on a topic or restrict the user from making certain choices. For example, a restrictive interface in a KBS for strategic planning may restrictthe user from utilizing multi-objective decision modeling options but allow the use of uniobjective decision programming modeling options. To achieve the same objective, a guidance interface will make available all the modeling options,
both multiobjective and uniobjective, but suggest to users thatthey use a uniobjective modeling approach. Thus, with a guidance interface, the system recommends design choices, but does not restrict choices. These design principles have been researched with findings indicating that attention to these design principles help inbuilding more focused and effective DSSs (Limayen and DeSanctis 2000, Montazemi et al.
1996, Norman 1998, Silver 2006, Singh 1998, Wheeler and Valacich 1996). Thus, considerable research is being conducted to identify ways to enhance the functional capabilities and design attributesof a DSS/KBS (Holsapple and Whinston 1996, Santhanam and Elam 1998). But a less-researched aspect also deserves attention. When users interact with the KBS and are focused on task completion, they may be implicitly learning about concepts, rules, and principles in the problem area that improve their knowledge structures. By the very definition of a KBS, knowledge about the specific problem area is embedded within the system. When
the system intervenes, it uses this embedded knowledge to provide advice and may even state it in the form of knowledge rules. For example, while a user is using a tax-planning KBS, the system may advise using a taxation rule that could be applied to prepare a better plan. Or, based on its knowledge rules, the KBS may identify and intervene to reveal an error in the tax plan. The user may not be specifically foc
đang được dịch, vui lòng đợi..
