M. David Merrill's

Instructional Transaction Theory (ITT):
Instructional Design Based on Knowledge Objects

CONTINUED:

Knowledge objects are containers consisting of compartments (slots) for different related elements of knowledge. The framework of a knowledge object is the same for a wide variety of different topics within a subject domain, or for different subject domains. The contents of a given compartment differ but the nature of the knowledge element in a given compartment is the same.

All knowledge objects have a set of information slots including: name, portrayal, and description. The name contains the symbol or term that references the knowledge. The portrayal is one or more multimedia objects (text, audio, video, graphic, animation) which will show or represent the knowledge object to the student. The description slot is an open compartment into which an author can place any desired information about the knowledge object. It is possible for the description slot to be subdivided into several subslots. These might include: function, purpose, etc. and may be defined by a given user.

A knowledge object may also have links to other knowledge objects. The nature of these links will be described in more detail in later sections of this paper.

Learning Environment

Goal. The goal of a learning environment is to enable the student to explore some device or setting. The objects in the environment behave in a way similar to their behavior in the real world. The student is able to act on objects in the environment and see the consequences of their actions. An open-ended learning environment allows free exploration within the constraints of the learning environment. So-called simulations that allow only a single action and are constrained as to the path that the student must take, are merely interactive demonstrations not learning environments.

Knowledge Structure. We have identified four types of knowledge objects: entities, activities, processes, and properties. (See Jones & Merrill, 1990). Entities represent objects in the world and can include devices, persons, creatures, places, symbols, etc. Activities represent actions that the learner can take to act on objects in the world. Processes represent the changes that occur in the properties of objects in the world. Processes represent events that occur in the world that change an entity. Processes are triggered by activities or by other processes. Properties represent quantitative or qualitative attributes of entities.

Each of these four types of knowledge objects include information slots for name, description (one or more slots), and portrayal (one or more slots).

For learning environments a knowledge object for an entity is expanded to include slots which point to one or more property knowledge objects. A property knowledge object, in addition to a name and a description, has a set of possible values. Each of these possible values is associated with a portrayal or indicator, thus a property knowledge object has a portrayal or indicator for each of the values that the property can assume. Thus when the value of a property knowledge changes its portrayal also changes.

A process is defined as a change in the value of some property of some entity. We say that this change in property value is the consequence of the process. Processes are also conditional, that is, a given process will not execute unless its conditions are met. A condition for a process is defined as a value on some property of some entity. A condition for a process is thus a value of a property. If in a given situation the value of the property is the same as the value specified for the condition for the process, the process executes; if the value of the property is not the same as the value specified for the condition for the process, the process does not execute. Finally a process can trigger another process. This provides for a chain of events (processes) each one triggered by the previous event (process). When a process is triggered, it evaluates its conditions, if they are true it executes, if the condition is false it does not execute; it then triggers the next process in the chain whether or not it executed its own consequence.

[In the example: one process is; disconnect air line. It is triggered by the activity undo air line. The air line (an entity) has a property, connection, with two values: connected and disconnected. The portrayal for the value connected is a graphic of the air hose connected to the double seat valve; the portrayal for the value disconnected is a graphic of the air hose disconnected from the double seat valve. This process, disconnect air line, has two conditions: (1) the value of the property connection of the entity air line is connected and the value of the property position of the entity air compressor switch is off. If the actual value of the air compressor switch is on, then the process, disconnect air line, will not execute and nothing happens when the student executes the activity, undo air line.]

The relationship among processes, entities, and activities enables the construction of learning environments from knowledge objects. In ITT this set of interrelationships is called a PEAnet (Process, Entity, Activity network). Figure 3 illustrates these PEAnet relationships. The learner executes some activity on a controller (itself an entity or part of some other entity). This action triggers a process. If the conditions of the process are true, then the process changes the value of a property. When the value of the property changes the portrayal of this value changes thus causing a consequence for the process to be indicated to the learner.

Figure 4 lists the entities, properties and property values involved in the Valve Removal Learning Environment.

Figure 5 illustrates the PEAnet relationships involved in the Removing Valve from Housing learning environment. The sequence of activities required for the student to remove or install the valve from the housing are indicated down the left side of the diagram, the processes triggered by these actions are shown in the center column, and the consequence (change of value of properties) are shown in the third column. The actions include: flip the air compressor switch, undo/connect the air line, undo/connect the cleaning connection, undo/connect the valve indicator line, remove/insert the flange bolts, insert a flange bolt into the tap hole, remove the flange bolt from the tap hole, pull out/insert the double seat valve.

Some processes have more than one consequence. [For example, there are two consequences for the process toggle. When the process, toggle, executes it changes the value of switch position to off if it is on, or to on if it is off.] The letters on the right side of the diagram indicate conditions (property values) for a consequence to execute. If the condition is not met the process does not change the value of the property. For many processes there are alternative consequences. When the value of some property has one value one consequence is executed, when the same property has a different value another consequence is executed. [For example, there are two consequences for the process, disconnect air line. The first consequence, set connection of air line to the value disconnected, is executed when the value of the switch position is off and the second consequence, show an accident message, is executed when the value of the switch position is on.]

Simulation Engine. PEAnet representation makes it possible to write a general simulation algorithm (sometimes called a simulation engine) which runs any learning environment constructed on PEAnet architecture. This algorithm monitors for an action (usually some mouse action on the screen), it interprets this action, meaning that it determines from the location of the mouse action which action has occurred, it then checks the conditions of the process triggered by this action, if the conditions are true it executes the process (changes the value of the property specified by the consequence), and it displays the portrayal corresponding to the new value of the property, and if a process-to-process trigger is specified it triggers the next process in the sequence. This instructional algorithm is written once and used over and over for different learning environments. Any situation, device, or phenomena which can be represented by properties and their values can be represented in PEAnet architecture and run by the simulation engine.

The set of actions down the left side of Figure 4 do not represent a linear sequence. [For example, if the position of the air compressor switch is off, then the air line, cleaning line, or valve indicator line can be safely disconnected in any order. The learning environment allows the student to attempt to disconnect any of the lines even if the position of the air compressor switch is on, however, the system then displays an accident message indicating danger to the technician or damage to the valve as a result of this action.] The student can also reverse any action. [For example, the student can toggle the switch, reconnect a line, or reinsert the flange bolts at any time.] Thus any action which the student could do in the real world environment is also possible to do in the learning environment with some representation of the consequence of this action.

Exploration. Without instructional overlay exploration is the only learning activity enabled by this learning environment. The student can operate the device or explore the environment experiencing consequences representative of those that would occur in the real world. Exploration represents a presentation alternative for a procedural or execute transaction.

It has been found that exploration without guidance is often insufficient for adequate learning. Exploration alone is inefficient. The student engages in trial-and-error behavior and makes many wrong moves before coming across the appropriate sequence of actions to solve a problem. Often the student is unable, from exploration alone, to discover the most efficient sequence of actions to accomplish a given goal in the environment. From unguided practice alone most students are unable to discover the underlying conditions which would prepare them for prediction or troubleshooting tasks. Hence, a learning environment without instructional overlay is only part of an instructional transaction and is therefore incomplete.

In order for a learning environment to be effective it is necessary to provide different forms of learner guidance. One type of guidance is propaedeudic instruction such as learning the names, functions, and locations of the parts of the situation or device. Learning a procedure is more efficient if learner guidance takes the form of a demonstration in the learning environment itself followed by scaffolded practice. To acquire prediction or troubleshooting skills the system must provide guidance in the form of explanations in the context of the student exploration. An explanation indicates those conditions which were met or not met when a given process executes or fails to execute. Finally, prediction and troubleshooting skills are developed when the student is required to predict the consequence of a given action or set of actions, or to find the conditions which prevented a given consequence or set of consequences from occurring (troubleshooting). In the following sections we will describe each of these instructional transactions as they are implemented using knowledge objects in a learning environment.

Identify (Component) Transaction

Goal. The student will be able to identify the name and location (with regard to some whole) of a given part of an entity(artifact, device, system, location, communication, etc.)

Knowledge Structure. In addition to the information slots (name, description, and portrayal), knowledge objects for component transactions require three additional slots: location, part of, and has parts. The location slot indicates the location of the portrayal with regard to some referent (the object to which the part belongs) knowledge object. The part-of slot contains a pointer to the referent knowledge object. The has-parts slots contains pointers to knowledge objects representing each of the component parts of the referent knowledge object.

The location slot has a parameter of mode with two values (graphical, temporal). If the portrayal of the KO is text or graphic, then the location mode is graphical (which means that it can be located on a computer screen). If the portrayal is audio or video then the location mode is temporal meaning that the portrayal of the parts are some time segment of the portrayal of the referent knowledge object. (For example a segment of a video or a segment of a musical rendition.)

Presentation. The presentation mode for the IDENTIFY transaction is as follows: Show the name and portrayal of the referent knowledge object. [In our example the name is in the title bar and the background for the valve including the pipe, the bottom plate of the valve and the switch plate is the portrayal of the referent knowledge object.] Show the portrayal of each part of the referent knowledge object. [In our example the illustration of each part is the portrayal. For example, the valve itself, the flange bolts, the air hose connection, the air hose, etc.] If explore mode is enabled: on mouse enter show the name of each part. On right click show the description of each part. If lecture mode is requested ["Tell me about some of the parts."] : highlight each part in the selected item list, show its name, show its description. For temporal portrayals (video or audio) some graphic portrayal usually accompanies the temporal portrayal to identify the parts in time.

Practice. The practice mode for the IDENTIFY transaction is as follows: If locate parts is selected: present a part name, the student clicks on the part, provide right/wrong with correct answer feedback. If wrong retain item in the list. If name parts is selected: highlight a part, present a list of part names, the student clicks on the name of the part, provide right/wrong with correct answer feedback. If wrong retain item in the list. If identify function is selected: present a function description, the student clicks on the part, provide right/wrong with correct answer feedback. If wrong retain item in the list.

Learner Guidance. During practice, if a student points to an incorrect part, the correct part is highlighted for the student. During practice, if a student selects or types the wrong name for a given part the correct name is given.

Parameters. Presentation and practice strategies can be controlled by a number of parameters. In a given transaction these parameters can produce a variety of different presentation or practice combinations. Some of the parameters for an IDENTIFY transaction include the following: show name {yes, no}; name mode {text, audio}; show portrayal {yes, no}, portrayal mode {text, audio, graphic, video, combination}; show description {yes, no}, and description mode {text, audio}. By employing these parameters a given part may be represented to the student in any of over 128 different combinations. [In the example the description was a combination graphic and description].

Execute (Activity) Transaction

Goal. The student is able to execute a series of actions which lead to some goal.

Knowledge Structure. The PEAnet structure for a learning environment enables the student to execute the activity. The PEAnet consists of a set of activities each of which triggers a process which leads to some consequence (change in property value of some entity in the system). In the example, the PEAnet structure of Figure 4 is the knowledge structure for the activity of valve removal.

Inference Engine. PEAnet knowledge representation makes it possible to write an inference algorithm (inference engine) which can determine an appropriate path to a goal from any set of initial conditions (values of properties). The author or student indicates a goal for a procedure. A goal is a value for one or more properties in the system. The inference engine then determines which property sets the goal property to the goal value. It then determines the process and trigger for this process which sets this property to the desired value. If all of the conditions for this process are true then it determines that the necessary to trigger this process is necessary to accomplish the goal. If the conditions for the goal process are not true, then the inference engine determines what process is necessary to make the false condition true. This backward chaining inference is continued until the inference engine has determined a path, or set of actions necessary to move from the initial state (initial property values of the system) of the system to the goal (desired property values of the system).

Presentation. An appropriate presentation for teaching a procedure (sequence of actions) is a demonstration or "Simon Says" simulation. The inference engine determines a path (sequence of actions) from the initial state of the system to the goal. The guide then directs the student to execute each of these actions in turn with the direction "Do ." If the student does this action its consequence is executed and the guide directs the student to do the next action. If the student does some other action the guide presents a message, "That is not the . Try again." This guided demonstration is continued until the goal has been reached. This type of presentation is sometimes called "guided practice." We feel that this is a misnomer, that this type of learning experience is not practice but presentation.

Another type of presentation is a "hands-off" demonstration. In this type of demonstration the guide does the first action by automatically moving the cursor to the appropriate entity part and simulating a click enabling the system to show the consequence, the guide then does the next action, etc. until the goal has been reached. We feel that this type of presentation is less appropriate than the "Simon Says" demonstration since it is passive. "Simon Says" requires the student to be actively involved in the demonstration focusing the students attention on the action to be taken and the consequence of this action.

Practice. We identified two additional levels of practice. "Next step" practice is the same as "Simon Says" demonstration except the guide does not indicate the action to be taken but merely gives the direction, "Do the next step." The student must remember which step is next and then do the action. If the student does a different action they are given the same reminder as with the "Simon Says" demonstration, that is, "That is not the ."

In "You do it" practice the system is put in open-ended mode enabling the system to execute any consequence to any process trigged by any action within the limitations of the system. However, the guide indicates the goal of the performance and directs the student to take those actions necessary to accomplish the goal. The guide also directs the student to click-on a finished button when they believe they have accomplished the goal. When the student clicks on the finished button, they are told if they have correctly finished the task (accomplished the goal). The guide then shows the student the path (sequence of actions) that the inference determined and shows the student's path (actual actions taken by the student). Unnecessary actions on the part of the student are highlighted to facilitate the student's comparison of their path with the guide's path. In systems where there are multiple paths to the goal the student's path may not be the same as the guide's but yet still accomplish the goal. The system recognizes that the goal has been correctly accomplished. The system records the student's actual path and the number of steps required for the student to accomplish the goal.

For complex systems additional practice is desirable. A number of different problems can be defined. A problem is a different set of starting conditions. Selecting a different problem resets the starting conditions to some predetermined values.

Learner Guidance. The scaffolding nature of the practice from a "Simon says" demonstration to a "You do it" practice is one form of learner guidance. This successive progression from highly guided interaction to unguided interaction has been recommended by a number of different instructional design theorists and has been supported by empirical investigations to facilitate the learning of procedural skills.

Providing an "explanation" to the student as they learn a sequence of actions has also been recommended and shown to facilitate the acquisition of procedural skill. PEAnet knowledge structure makes it possible to build a system which can automatically provide explanations under a variety of circumstances. In an explore mode the student can request an explanation from the guide. In a "Simon says" demonstration mode or "next step" practice mode the learner can turn on the explanation which is then displayed as each step in the procedure is executed. In a "you do it" or performance practice mode the explanation can be turned on and displayed as the student executes each action.

An explanation has two parts: what happened? and why? In terms of the elements of knowledge objects what happens is the consequence or change of property value caused by the process. An explanation template enables the guide to provide this what-happened explanation. One template is as follows: When the student requests an explanation, the guide present this message when the process executes, "When you the of is changed to "; and this message when the process does not execute, "when you nothing happens." A system may have different explanation templates for different kinds of values and different operations for changing those values but these details are beyond the purpose of this paper.

The "why" part of the explanation is implemented by presenting the condition(s) which had to be satisfied in order for the process to execute. The "why" part of the explanation presents a text template of the following form when the process executes, "This happens because the of is "; or the following when the process does not execute, "This happens because the of is not .

In the example, suppose the student attempts to remove the valve before removing the flange bolt from the tap hole. The explanation presented by guide reads as follows: "When you nothing happens. This happens because the of the is not ."

The buttons that appear below will be found at the top and bottom of each page of the discussion. The first button will take you back to the previous page (in this case, to the beginning of paper #22). The middle button will take you to the ITForum home page. The last button takes you forward into the remaining sections of the paper.