Student model

Ary discussion of the problematic issues of chemical education will include concerns about how to model for students (and for otrrselves ) this reality of atoms, iorrs and molecrrles. We shotrld firstly appreciate that there is a terrsion between the pragmatism of accepting this reality , which we try to model, and the truth that this reality is itself a grand model. Perhaps, therr, orrrtaskisto develop useful students models of the chemists models. 

Individually sequenced lessons versus a common path. An early decision with multiple later consequences was to maintain a common sequence of lessons for all students. This choice was predominantly driven by the experimental nature of the system. That is, we intended to use SPS as the basis of research study, and we desired to control as many variables as possible, including order of instruction. Nonetheless, we developed SPS with an eye to promoting individualized instruction. The modular construction of SPS allows great variation in the presentation of individual lessons or blocks of lessons, and the student model that allows the tailoring of instruction to the individual is already in place. 

The model of Figure 13.4 shows two layers at the middle level, and these layers contain identical sets of nodes. The nodes for the student models were identified from the student interviews conducted at the end of the first SPS session, and they are the bases for the students cognitive maps. Three trained individuals read the transcript of each student s interview and determined which nodes... 

As a test of the model s adequacy, a simulation was carried out in which the plenary cognitive map of Figure 13.1 was used as the student model. All 100 items available in the SPS identification task were presented sequentially to the model, and its responses were compared with the correct answers. The model performed with 100% accuracy, successfully identifying the situations of all items. 

There is a big scope of displaying of specimens and objects collected by students. Models and projects made by students... 

Further standard voltages could be measured and placed into an electrochemical sequence with other pairs of half-cells, as can be found in many chemistry teaching books. The extent of the introduction of the hydrogen half-cell as a standard electrode can be determined in each individual lesson. If one consequently describes all redox reactions in relation to the metal sequence, redox or electrochemical sequence with ions and offers the students model drawings (see Figs. 8.3 and 8.4), then the electron transfer and the redox definition, in terms of involved smallest particles, becomes even clearer and the mixing at the language level of substances and that of particles can be effectively suppressed. 

Many college and university chemistry laboratory courses now include experiments in which students model molecular structures and properties on personal computers. Typically, the software used combines molecular mechanics and molecular orbital methods. A preliminary model is created on the screen based on a structural drawing, a tentative geometry calculated by molecular mechanics, then refined by molecular orbital methods to produce an optimized structure. 

Individually the student decides which links to follow, i.e. the student is in control - not a passive object under the influence of a teacher. At least the student will hardly realize that he/she is being manipulated (cf. the previously mentioned student model discussion). 

The student decides, so his/her motivation will determine what happens next. This is a crucial point in utilizing hypermedia as instructional tools, and this issue is also decisive in outlining the principles of a student model. The question is how much initiative will be left over for the student - and is building a student model actually cheating the student ... 

The contrasts just presented lay the foundation for considering the emergence of hybrid systems which can combine the virtues of both ITSs and HMSs. The central component of such hybrid systems will be the student model. 

In the hypermedia area, the availability of a student model within an HMS would in principle enable the system to tailor itself to the particular characteristics or interests of the student and to the perceived needs of the moment. This intelligent adaptation of an HMS would not aim to be directive, as an ITS often is, but would instead involve a selective orientation of the information made available or a tailoring of the HMS interface, all with the aim of letting the student make well-informed and wise decisions about where to proceed. One interesting illustration of this (Nielsen, 1990) is the potential for dynamically differentiating between anchors (buttons) in a frame, adjusting their prominence according to estimated relevance to the student, as determined by an examination of the student model. 

As indicated above, the process of student modelling involves capturing the flow of interaction during a session, and interpreting that flow for adaptive purposes. In an HMS, a trace of the student s browsing can easily be captured (Macleod, 1990). However, it is the interpretation of this trace that is difficult for a system to perform. 

The design of hypermedia systems involves the design of various fundamental components, such as the information model and the interaction model (Jonassen Grabinger, 1990). What we see the need for now in seeking to extent the hypermedia paradigm further is the design of yet other components, particularly the student model and a didactic model (Mandl, 1990). 

As there is for the time being no universal student model capable of explaining reality (will it ever be possible ) and as it is not operational to consider each individual case, we think that one way out of this difficulty is to try, similarly to what happens in differential psychology, to pragmatically define a limited number of learning profiles capable of explaining the vast majority of cases (not necessarily all ). 

The general idea is to test these programs for consistency of the learning profiles. These learning profiles, as our research has been proving, are more dependent on a situation than on a student model. For example the strategies of exploration of information and acquisition of information vary radically if the experiments are present or absent. 

As already said, the student model plays a key role in ITSs. Traditional CAI systems used a quantitative model of the student, which was only based on statistical data, concerning the number of correct answers, the number of mistakes, and possibly response times. 

In a bug model the learning status of the student is represented in terms of incorrect knowledge (also called misconceptions) that represent possible student errors (bugs). The student model contains not only correct knowledge, but also incorrect knowledge. The main problem of this approach is the impossibility of foreseeing an exhaustive error set. 

John Self, in a recent research [16], reviews the actual uses which have been found for student models in existing ITSs. The student model is considered as a 4-tuple , where P describes the procedural knowledge of the student, C the conceptual knowledge, T the individual traits and H the history. The set K = P u C describes what the student knows. 

Self identified twenty functions of the student model, grouped in six categories ... 

The updated student model is then examined by the therapy module, which attempts to tailor the remedy (the right therapy ) to the needs of the student. It determines the best teaching action to perform, taking into consideration the correct and incorrect knowledge present in the student model. The application of the selected teaching action generates explanations and examples, extracted from a predefined library, and proposes a new problem both to the student and to the expert module. 

---