Alexander Repenning and Andri Ioannidou (Boulder, CO, USA)

Center of LifeLong Learning & Design

University of Colorado at Boulder

{ralex, andri}@cs.colorado.edu

and

AgentSheets Inc.

{alexander, andri}@agentsheets.com

http://www.agentsheets.com

Engaging Learners through Simulation-Based Design

Engagiertes Lernen Durch Simulationsunterstütztes Design

Summary / Zusammenfassung
An engaged learner is an individual actively participating in a learning activity. Learning by doing is an effective learning process that not only engages learners but, as a consequence, can increase learning retention and even has the potential to make learning activities enjoyable. In the context of web-based learning or computer-aided learning it is often less clear what constitutes an active learning activity. The intrinsic limitations of mouse and keyboard-based human-computer interactions degrade many so-called interactive multimedia titles to page-turning or multiple choice learning activities. We claim that simulation-supported design activities result in significantly more interactive learning activities with great potential to deeply engage learners. / Ein engagierter Lerner ist ein Individuum der aktiv im Lernprozess mitmacht. Lernen durch Tätigkeit ist ein effektiver Prozess der nicht nur den Schüler engagiert aber konsequent die Lernbeibehaltung erhöhen kann und sogar das Potential hat den Lernprozess als erfreuliche Tätigkeit zu geniessen. Im Kontext des web-basierten Lernen oder komputer-unterstütztes Lernen ist es oft unklar was eine aktive Lerntätigkeit konstituirt. Die Limitationen der Maus- und Tastatur-basierten Mensch/Komputer Beziehung degradieren viele sogenannte interaktive Multimedien Titeln zu Seitenumblättern oder Multi-Choice Lernaktivität. Wir behaupten dass simulationsunterstützte Aktivität bezeichnend besseres Resultate im interaktiven Lernen hat sowie ein grosses Potential besitzt den Schüler stärker zu engagieren.

1 Engaging through Design

Lifelong learning has become central to our society. The traditional model dichotomizing educational practice in a time to acquire knowledge –at school - and a time to apply knowledge – at work – is no longer operational. A number of factors contribute to the constant need to learn new things while at work or at home. Members of the information society change jobs as frequently as never before in the history of labor. We have to learn to use new tools, collaborate with new people, start new projects, and start new jobs. Based on new scientific discoveries practice changes. In a field such as medicine, scientific discoveries may end up producing new kinds of treatments or medication. In engineering-oriented fields new technologies, materials, tools and techniques dramatically change practice over time. The need to embrace lifelong learning, not just as a quick weekend workshop but as a fundamental philosophy of life itself, becomes increasingly apparent. How can information technology help us to deal with the overwhelming and increasing need to learn?

1.1 Old Content, New Media

The use of information technology to support learning is still in its infancy. The affordances of the web as a new medium are still being explored. A lot of the newly developed educational content uses information technology as gift-wrapping for traditional content. Presentations are being “webbified” rendering existing slides into web pages. While this does serve the purpose of making content more widely accessible it does not employ technology affordances to create more engaging content. The history of new media usage indicates that old to new media transition often takes a considerable amount of effort and time. Early TV broadcast was little more than radio-show-inspired talking heads with pictures. To develop a sense where technology should go to more functionally support learning we have to revisit previous visions of pedagogical effectiveness.

"I hear and I forget, I see and I remember, I do and I understand."

- Confucius

This frequently quoted statement by the ancient Chinese philosopher Confucius is a plea to engage learners in hands-on activities for effective understanding. Historically geared at learning activities such as learning carpentry or painting, it is less clear how the act of doing transfers to learning through contemporary information technology-based media. Clearly, simple page-turning schemes in which a learner clicks navigation buttons to traverse a web of information – while informative – do not constitute engaging active learning. This poses a big challenge to information technology. How can we build truly engaging educational content that stimulates learning by doing?

1.2 Design meets Simulation

Design activities are constructivist [9] and constructionist [3] activities in which a learner creates conceptual as well as physical artifacts. Design is a constructive process requiring trade-offs to be made. This process exposes the learner to complex design spaces. Instead of exposing a learner to a large number of isolated facts which he or she only "uses" in the context of some multiple choice test, a design activity will connect relevant facts in the context of building an artifact. For instance, even the design of simple electronic circuits makes a learner choose from thousands of existing components and combine them in a way that fulfills some specified requirement.

A design activity is an active design space exploration in which a learner learns about artifacts by assembling and disassembling building blocks. For instance, building a car by assembling individual LEGO® pieces is a design activity. Many design activities are open-ended. In other words, like most authentic real-world problems, design problems do not typically have a single, correct solution. For instance, there is an infinite number of cars that could be build out of LEGO pieces.

Design activities require design environments. In the physical world, rich environments support design activities. For instance, the electronic designer is surrounded by large collections of tools (e.g. oscilloscopes, soldering iron, etc.), components (e.g. integrated circuits, capacitors, resistors, etc.) and a suitable workplace. Virtual design activities require virtual design environments [1]. Virtual design environments are considerably more complex than page-turning interfaces such as “webified” slide presentations, since they must afford complex human-computer interactions. A virtual design environment needs to allow its users to:

• Select components: users need to be able to select relevant components from a component palette. A car designer needs to be able to select car parts such as engines, and wheels.

• Assemble components: users need to be able to combine components and assemble them into functional units.

• Simulate function: After users have assembled components it should be possible to simulate the function of the complete design. For instance, assembling engine parts into a complete engine should allow a user to simulate the running engine. If components are missing or misplaced the simulation should highlight problems and help the user improve the design.

Using authoring tools such as AgentSheets [5, 8], design environments can be built for all kinds of subject areas as well as for all kinds of users. Below is an example of a simple architecture design environment for kids called the Bridge Builder.

The objective of the Bridge Builder activity is to develop an understanding for static and dynamic forces in a bridge. While this activity is geared towards middle school kids, it illustrates the design process. The exact affordances of learning by using a design environment are hard to convey in a paper. The readers are invited to explore their own bridge design by interacting with the Bridge Builder simulation applet at http://www.agentsheets.com/Applets/bridge-builder/. Bridge Builder users are instructed to build a bridge with the minimal number of bricks. The initial bridge presented to the user does work but is not cost-effective because it is uses too much material. Employing the tools provided, users modify the bridge design by adding, removing, and rearranging bridge components. Figure 2 shows typical bridge designs.

Figure 1. An AgentSheets-based design environment. Users explore bridge design by selecting bridge components from the left window, and assemble them into a complete bridge. AgentSheets simulates the bridge indicating strong forces and makes the bridge collapse if the design is too weak.

Design activities are open-ended. Even with a relatively simple design activity such as the Bridge Builder there is no unique solution. In the process of exploring the design space users are likely to rediscover well-known designs. For instance the column-based bridge design (Figure 2 left) was used by the ancient Greeks. Much later in the history of architecture the Romans started to appreciate the advantages of arch-based designs (Figure 2 right).

Figure 2. The user has improved the bridge by removing bricks. The simulation computes tensional forces and visualizes them in shades of red. What is the minimal number of bricks needed?

The combination of design activities and simulations is highly suitable for Computer-Aided Learning. In contrast to instructivist approaches [2], computers used as design environments are not just containers of information presented passively to the learner but a medium that engages the learner in bi-directional interactions. That is, the computer is not just an information container but becomes a thinking tool. Through the use of interactive simulations the potential to actively engage learners dramatically exceed other media such as text, picture, animation and even movies.

1.3 Delivering Interactivity on the Web

To make design environments appropriate for distance learning, it is necessary to wrap them up as web-compatible content. The AgentSheets authoring tool includes powerful generators to create Java output. Figure 3 shows a design environment for exploring the effects of antidepressants created by a psychiatrist. The resulting web pages, geared towards patients not medical doctors, is interactive by allowing its users to experiment with the administration of antidepressants and explore the consequences.

Design environments are valuable computer-aided learning tools, but how hard it is to build this kind of content? The following sections provide a brief overview over the AgentSheets authoring tool.

Figure 3. An AgentSheets simulation of a serotioning synapse wrapped into a web page. Learners can use different tools and change the simulation while it is running in the web page.

2 How Hard is it to Build an Interactive Simulation?

The AgentSheets authoring tool empowers casual computer users with no formal programming training to engage in constructionist learning activities. Combining end-user programmable agents, spreadsheet technology, and Java authoring, AgentSheets allows users to create their own interactive simulations and publish them on the web.

To build a simulation, an author employs the authoring tool to describe agents and causal relationships between agents. Agents are computational representations of domain objects that execute their behaviors in parallel. In the domain of medicine agents may represent objects at different levels including molecules, cells, organs, people, or societies. Such a simulation is much more than animation since its behavior emerges out of the causality expressed.

AgentSheets agents have rich communication modalities in support of truly interactive educational content. Agents can react to other agents, mouse clicks, keyboard input, and even speech input. Agents are capable to compute formulae, open web pages, query web pages, play sound, play MIDI music, and speak using speech synthesis.

A wide spectrum of users, ranging from elementary school students to scientists, have used AgentSheets to create interactive simulations and games in a variety of disciplines including computer science, environmental design, fine arts, robotics, music, biology and history.

2.1 Example: the Virus Attack Simulation

To illustrate how an AgentSheets simulation is created, a brief description of how to build Virus Attack, an educational simulation of a virus spreading through a community, is presented. The point of this simulation is to understand the basic virus mechanism. How fast does the virus spread? Does the number of people infected by the virus go up linearly, or is there some other function behind the rate of growth? What can be done to contain the virus? The principles behind this kind of simulation are not limited to virus spreading. The same laws apply even to the spreading of rumors, fads, and other types of information relevant to society.

In Virus Attack, simulated people move around randomly, representing real people moving around by going to work, shopping, and traveling. A healthy person standing next to an infected person has a certain chance of getting the virus.

2.1.1 Creating Agents and their Looks

An AgentSheets simulation is made up of agents that get created and reside in a Gallery. Users create agents and give them appearances by drawing icons. For the Virus Attack simulation, a number of agents to represent healthy people, infected people, doctors, and background tiles for the people to move around on are created (Figure 4).

Figure 4. The Gallery: a collection of agents defined by the simulation author.

2.1.2 Creating a Simulation World

A worksheet is the world where the simulation unfolds. Users can populate simulation world with agents from the Gallery. For the Virus Attack, the worksheet (Figure 5) is built with “Background”, “Person” and "Sick Person" agents by first selecting the them in the gallery and adding them using the drawing tools provided.

Figure 5. A simulation world representing some abstract space. For instance this space could be interpreted as 4 offices connected with aisles. The green, unhappy agent (lower left) is a carrier of the virus. How long will it take for the virus to spread through this community?

2.1.3 Creating Agents' Behaviors

AgentSheets provides the object-oriented programming language called Visual AgenTalk (VAT) [4] for giving behaviors to agents. These behaviors are created by combining Condition and Action commands (Figure 6) into rules and by grouping rules into methods. For example, for the Person agent in the Virus Attack simulation, conditions and actions were combined in its behavior editor (Figure 7) to enable the Person to move around randomly and with 5% chance get infected when next to a sick person.

Figure 6. Agents’s behaviors are rule-based. Each agent has a large repertoire of commands available to the users through the conditions pallette (left) and the actions pallette (right).

A drag-and-drop interface (called Tactile Programming) makes it easy for users to not only assemble but also test behaviors. At any point during the creation of a simulation, individual conditions and actions, rules, methods or entire behaviors can be dragged onto an agent to test them in specific contexts. AgentSheets steps through the execution of the VAT component being tested and provides visual and acoustic feedback to aid the user in understanding and debugging the program.

Figure 7. Agent behaviors are defined by dragging conditions and actions from the palettes into the IF and THEN part of rules. The “Person” behavior has only two rules; one to define how to spread the virus and the other one to make the agent move around randomly.