Murray, W. (1991). An endorsement-based approach to student modeling for planner-controlled tutors. In Proc. of International Joint Conference on AI, (IJCAI'91), Sydney, Australia, pages 1100--1106.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... phase) Suggestions of solutions to this problem are for instance the use of fuzzy logic (Derry et al. 1989) probabilistic student models (e.g. Martin and VanLehn, 1993, Villano, 1992, Petrushin and Sinitsa, 1993, Sime, 1993) and use of accumulated pro and con arguments of differing strengths (Murray, 1991). With respect to the level of detail of data, three different levels of granularity of input are defined by VanLehn (1988) as Final States, Intermediate States and Mental States. Input of final states implies that the input reflects the solution to a problem in the form of an answer to a ....

....using hierarchies of knowledge (Lesgold et al. 1989, Greer et al. 1989) Another aspect of evaluation is when evidence is of unequal strengths e.g. recognition of a concept is easier than to recall and use it. Judging competence by using arguments of differing strength has been proposed by Murray (1991). There can also be different degrees of belief in the model, with probabilistic values attached to the items in the model. These three components of the gap, data acquisition, transformation and evaluation must all be bridged in one manner or another before the abstract process of student ....

Murray, W. (1991). An endorsement-based approach to student modeling for planner-controlled tutors. In Proc. of International Joint Conference on AI, (IJCAI'91), Sydney, Australia, pages 1100--1106.

....for each fact in the model of the learner. This system accepts justifications for the beliefs, which will be nodes in the dependency network kept by the AMMS. In fact, all the systems that cope with revisions in learner models have some kind of Reason Maintenance system included (see [6] 5] [8]) However, apart from keeping the justifications of the facts in the model, the learner modeling system has to decide how to revise the model as well. It has to decide which facts to keep in the model, and which ones to eliminate. This choice is based on rationality criterion which depend on the ....

....The revision process is made according to the most trustable justifications. This notion of trust is based on the confidence the modeling systems have in the process of generating the beliefs to be included in the model. In Murray s work, for instance, this confidence is based on endorsements [8]. 4 Learner Changes Whereas a system contradiction is automatically followed by a revision, a learner contradiction is reported in the model, which will change when the learner changes himself. The learner however can change his beliefs in many different ways. What we propose here is a theory ....

William Murray, `An endorsement-based approach to student modeling for planner-controlled tutors', in 12th International Joint Conferenceon Artificial Intelligence, pp. 1100--1106. Morgan Kaufmann Publishers, (1991).

....the desire to adapt quickly to changes in the student s knowledge, and the desire to ignore minor slips. There have been some suggestions of solutions to this problem, for instance the use of fuzzy logic (Derry et al. 1989) and use of accumulated pro and con arguments of differing strengths (Murray, 1991). There is also the problem of gathering sufficient data. What is sufficient naturally depends on the knowledge we are teaching and how it is represented. In cases where the Domain Knowledge (w. bugs) Domain Instructional Knowledge Mental State Input Intermediate State Input Final State ....

....Perturbation Correct knowledge Known erroneous knowledge Undecided DK Correct Domain Knowledge SK Student Knowledge Bugs Bug Library Assumed erroneous knowledge 12 21 11 94 concept is easier than to recall and use it. Judging competence by using arguments for or against has been proposed by Murray (1991). 4.4 Computational aspects on student modelling Some aspects on computability have already been mentioned. For instance, predicting possible answers from known domain knowledge is a way to reduce the search space of explanations to student answers. If we equip the student model with detailed ....

Murray, W. (1991). An endorsement-based approach to student modeling for planner-controled tutors. In Proc. of International Joint Concerence on AI, (IJCAI'91), Sydney, Australia.

....rule A, that rule (A) endorses that hypotheses. In the AMMS this is dealt with by creating a justified assumption which has the A endorsement in its justification. 3 This notion of trust will be defined later in this document. 4 The idea of endorsements [4] has been used in Murray s work [11]. In this document, we will use the idea of an endorsement to be an explicit recorded argument for an hypotheses, but we give our own formal definition within the framework presented. The premisses are certain acquired facts (which can be, for example, some actions performed by the learner) When ....

....are used to guide the revision process. On the other hand, because the justifications are based on the oracles (actions of the learners) they can handle learner changes. In Murray s work, the endorsements are arguments for and against the beliefs in the learner model. The endorsements presented in [11] are grouped in classes with different reliabilities. This correspond to an ordering among them, as in the AMMS case. The approach of Van Arragon [1] to learner model acquisition is based on defaults as well, and uses the Nested Theorist to formalize that acquisition. In his work it is shown that ....

William Murray, `An endorsement-based approach to student modeling for planner-controlled tutors', in 12th International Joint Conferenceon Artificial Intelligence, pp. 1100--1106. Morgan Kaufmann Publishers, (1991).

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC