Genetics is a particularly difficult topic to teach because it involves complex, interrelated, mostly unobservable processes that occur at different levels. With this in mind, we created a great simulation called GenScope to help students learn about genetics. Great, except that it didn't help them pass genetics tests.

Our research in 26 classrooms showed that students were engaged and able to use GenScope to solve genetics problems, but we were unable to measure any increase in their ability to solve closely related paper-and-pencil problems. Clearly, students were unable to transfer their model-based learning.

Trying to understand why we fell short of our goals, we designed a solution called "hypermodels," which might turn out to be a significant development in educational software.

Levels of GenScope
The illustration on page 13 shows how GenScope represents the linked, multi-level processes of genetics. At the organism level students can view the phenotype (the collection of physical traits), but they receive no direct information concerning the organism's genetic makeup. When they move to the chromosome level, they can observe the genes carried on the chromosomes. According to Mendelian laws, altering those genes can change the appearance of the associated organism.

At the chromosome level, genes are simply markers - their exact nature remains as mysterious to students as it was to Mendel. We know now that the explanation of the genetic mechanism resides at the molecular level. GenScope enables students to drop down to this level to explore the DNA molecule contained within each chromosome and to alter it at will. Such alterations result in mutations that show up as changes in an organism's phenotype and may be inherited by its offspring.

Inheritance is handled at the cell level in GenScope by simulation of the twin processes of meiosis and fertilization. By using a special zoom tool, students are able to look at the chromosomes during meiosis and see which alleles they carry. By controlling gamete formation, and by selecting beforehand which gametes to fertilize, students can control the genotype of the resulting offspring.

Such control is not available to the students at the pedigree level. They must instead rely on probability and statistics to predict the outcome of a cross between two organisms. Deprived of information and control mechanisms that are also unavailable to scientists, students are forced to rely on their internal models of genetics to make reliable inferences from data analogous to that obtainable in the real world.

The Solution: Hypermodels
The difficulties students encountered occur frequently when students learn through the exploration of open-ended tools. Assessments outside the environment cannot measure learning within the tool. Students need prompting to transfer their learning to another environment. And teachers are often unable to provide the guidance students need to fully exploit the model-based environment. Because the problems we encountered with GenScope plague most modeling environments, we decided to solve our problem by creating a software architecture that could be applied to any software tool - a hypermodel.

Our hypermodel architecture, illustrated in Figure 1, consists of three software layers designed to separate functions relating to domain content from more general ones relating to pedagogy, and to give control over both.

At the bottom level is the domain content engine, which consists of a set of loosely coupled components, or views, that can be combined and integrated in a variety of ways. In our most elaborate hypermodel to date, BioLogica™ is the domain content engine. BioLogica™ is similar to GenScope but reprogrammed in Java. In BioLogica™ the chromosome view and the organism view share a common database that contains, among other things, the genotype of every organism. One of these views uses this information to display alleles on chromosomes; the other, operating with a set of built-in rules, determines and displays the phenotype of each organism. Altering a gene in the chromosome view, say, from a dominant to a recessive allele, will be reflected, as appropriate, in the form of changes in an organism's phenotype. In our implementation of the BioLogica™ domain content engine we ensured that its views are interoperable. (Each view is a Java Bean and each is serialized using XML, a common markup language.) BioLogica™'s views are purposely kept simple. They do not, for example, have a user interface of their own, but must be placed on the screen and configured by the next level of the hierarchy, Pedagogica.

Pedagogica, as the name suggests, handles all things pedagogical. It is responsible for all interface details, including the placement of text boxes, buttons, and domain engine views in various locations on the screen. Pedagogica also controls the flow of a student's activity by shifting from one set of views to another in response to a student's actions. Pedagogica can set up "listeners" (software agents that monitor views and other objects) and report on changes in their contents or properties. This enables the software, for instance, to react whenever a new organism is created, or when the student clicks on the image of a gene. Pedagogica can communicate with the student through graphics and text, and pose multiple-choice and essay questions. It also controls the collection and storage of data, maintains and controls access to student records, and manages the registration and login functions.

Pedagogica itself is controlled by the third software layer: the scripting engine, which has the job of interpreting short scripts written in a simple, interpreted language, called EASL (Educational Application Scripting Language) developed by us. These scripts implement the activities that students interact with, setting up the initial problem, configuring the hypermodel to match the problem, observing and reacting to students' actions, and communicating with them as they work through their investigations. Although EASL scripts are relatively straightforward to write, they are full-fledged programs and require greater attention to detailed computer functions than most curriculum developers or teachers are probably willing to put up with. We are working on simplifying EASL so non-programmers can create their own scripts and modify scripts of others.

The First Hypermodel: BioLogica™
The BioLogica™ hypermodel can be visualized as a newer version of GenScope linked to Pedagogica. Superficially, the BioLogica™ hypermodel looks much like GenScope. It uses the same dragon species and many of the same levels and tools. But where GenScope is a general-purpose tool that students can use to investigate genetics, the BioLogica™ hypermodel is a tool with which researchers and teachers can develop scriptable genetics curricula.

The difference between the two applications is most apparent in their interface. GenScope has a plain interface: it opens with an empty organism window and provides options for creating organisms, pedigrees, and populations, and to view cells, chromosomes, and DNA. GenScope's interface is designed to make it as easy as possible for users of varying sophistication to gain access to its many features; it is a tool-driven interface. BioLogica™'s interface, in contrast, is activity-driven: the layout of the screen and the actions and representations available to students are determined by the particular activity that is in progress. Whereas GenScope is intended to run as a stand-alone application, BioLogica™ is a utility for the creation of learning activities. BioLogica™ cannot be run by itself, but requires a script - a short executive program that implements the learning environment. The script embodies both the activity that students are to engage in and the indicators that can be used to judge their performance.

The hypermodel software is typically resident on each computer in a classroom, but the scripts may be located on a central server, either remote or within the classroom. This makes it possible to present students with customized learning experiences. For research, this means we can easily manage different treatments within the same classroom. For educators, this means that the activities can be adapted to match individual needs. Using scripts, we can control and limit the options presented to students. Instead of getting lost, they can focus their attention on a particular problem. They can still explore and learn through inquiry, but in a reduced space that is more easily explored and understood. We can also prompt students to think about what they are learning and to explore the links to other concepts.

Pilot Studies
At this writing, we have conducted three pilot studies with scripted BioLogica™ hypermodels and are about to embark on two more. While the data from these trials have yet to be fully analyzed, the initial results are encouraging. GenScope's main feature was its appeal to students, and adding scripting to BioLogica™ has not changed that. Moreover, there is preliminary evidence that the activities we created have overcome many of the difficulties we encountered with GenScope. Encouraged by this success, we are now converting probeware and our molecular dynamics package Oslet (see article, page 4) into hypermodels. We encourage other model makers to do the same so that there will be hypermodel tools for every subject that can use models.

Another benefit of our hypermodel design is its capacity for embedded student assessment. A student using the BioLogica™ hypermodel can be challenged in a lesson to discover whether a particular trait is sex-linked. Pedagogica can note which screens the student uses in answering the question, what order the screens were accessed, and how long each was viewed. From this, we can infer whether the student was guessing, how well the student understood the concepts, and whether the student was lost. This kind of embedded assessment could provide invaluable high-level feedback and reduce the amount of time spent on formal assessment. It is also a promising research tool that allows us to obtain at a distance detailed information about student thinking, knowledge, and problem-solving strategies.

The most significant contribution of information technologies to improved science learning is likely to come through the increased use of powerful, content-based modeling and data analysis tools. Well-designed models should help students learn fundamental ideas and give them computer-based alternatives to formal mathematical techniques. The hypermodel architecture could be the key to realizing this dream in real classrooms.

Robert Tinker is president of The Concord Consortium (bob@concord.org).
Paul Horwitz is director of the CC Modeling Center (paul@concord.org).

Copyright © 2001 The Concord Consortium, All rights reserved.
Last Updated: April 21, 2001
Submit questions and comments regarding this site