The Control Panel

  The control panel is the only place that you actually interact with the model.  You will find several options for selecting the model's conditions, run control, and some advanced output options.  Following is a list of the individual components and a description of what they do.

Pre-Task Appraisals

On the left, you will see a radio-button panel with options for Threat, Neutral, and Challenge.  When one of these options is selected, the model will reload with new parameters that influence the model's behavior in the desired way.  If you wish to run the model through only one task, the selection that you make will remain the model's pre-task appraisal throughout that task, but if you wish to script the model through four-trial blocks (more on that later), your selection will be the model's initial pre-task appraisal (for the first trial) and for the additional trials in the block the model will take on either a threatened or challenged pre-task appraisal based on its previous performance and caffeine level.

Worry

Below the task appraisal option panel, there is a check box labeled "Worried".  Select this box if you wish to make the model worried.  This effect is cumulative with task appraisal effects.

Caffeine Level

Across the bottom of the control panel, you will find a slider labeled "Caffeine Level".  To include the effects of caffeine, simply adjust the the level (in milligrams) to the desired amount.  This can be done by dragging the carrot across the slider, using the arrow buttons at each end, or typing the desired amout into the text box on the left of the slider.  Caffeine effects will only show up if the model is scripted in four-trial blocks.

Run Control

There are buttons on the control panel for running, reloading, scripting, and stepping the model.  Pressing "RUN" will run the model continuously for one trial (four minutes by default, in ACT-R time).  Pressing "RELOAD" will reset the model with the preferences selected elsewhere in the control panel.  "SCRIPT 1" will script the model over one four-trial block, outputting the data obtained from each trial into the script data window (see below).  "STEP" will run the model for one production cycle.  This is typically used to closely inspect the model's cognition, for debugging purposes or to gain deeper understanding of the model's behavior.
    There is also a text box to the immediate right of the RUN button.  This contains the text "240" by default, and it is recommended that this be left as is.  It is possible to change the text and thereby change the duration for each trial, but this feature has not been debugged so try it at your own risk.
 

Menu Functions

Across the top of the control panel is a menubar, containing the items "Model", "Edit", and "Help".  The Help function merely pops up a window that directs your attention to this document, but the other two menus contain functions that will be useful to learn about.

Model Menu

This menu contains the items "script", "run", "reload", and "quit".  The run and reload options perform the same functions that the run and reload buttons on the main part of the control panel (as above).  The quit option will quit your lisp session, destroying the model interface with it.  The script option opens the script options dialog box, which gives you quite a bit more control of model scripting than the "Script 1" button on the main control panel.  Following is a discussion of the script options dialog box.

Script Option Dialog Box

Opened from the "Model" menu, this dialog presents options for scripting the model over varying conditions as many times as you like, or have time for.  There are two main scripting options, one for running the model many times with the same pre-task appraisal each time (Script Single Condition), and one for scripting the model in four-trial blocks with the same initial pre-task appraisal at the start of each block.
    If you wish to use the single condition option, simply select a task appraisal from the radio button panel under the Script Single Condtion label.  Then, if you so desire, you may change the number in the box labelled "Iterations" to alter the number of trials the script will perform.  To start the script, press the "GO" button in the Script One Condtion area.  The script options dialog will disappear and the model will begin its script (this might take a while).  Values for caffeine level and worry will be taken from the controls on the main control panel.
    If you want to script the model over four-trial blocks, just apply the above procedure to the controls under the Script Four-Trial Blocks label.  The difference is that the task appraisal that you select will only be applied to the first trial out of each four-trial block.  The pre-task appraisals for the following three trials will be decided by the model based on its performace on the previous trial.  Also, the number of iterations now applies to the number of four-trial blocks, not individual trials.
    For a note on how to greatly speed up script execution, see the Note at the end of the Getting Started document here.

OUTPUT

All output for the scripts is entered into text files in the model directory.  Specifically, raw data is stored in a file called raw-data, and processed data is stored in a file called data.  The form of the raw-data file is a long column of numbers in the form:

48
3
51
0
45
5
and so on...

As you might have guessed, these indicate pairs of data points for each trial, namely, number of attempts and number of errors.  So the data above would indicate three trials: during the first trial, the model made 48 attempts and got 3 wrong; on the second, the model made 51 attempts and had a flawless performance; and on the third, the model made 45 attempts and got 5 wrong.
    You do not really need to work with the raw data unless you really want to personally inspect the details of the model's performance, because the processed data file, "data", contains average performace measures, as well as standard deviations, in a very readable form.  See Sample Output for a detailed view of what the data file looks like.
    NOTE: If you run scripts multiple times without manually archiving or renaming your data files, you may lose your data.  The macros are slightly forgiving - if you run two scripts without archiving your data from the first one, your data files from the first script will be renamed raw-data.old and data.old, but if you run the script a third time without taking precautions, the data files from the first script will be deleted.
 

Edit Menu

The edit menu only contains one item - the traces function.  This will bring up a new window that will allow you set ACT-R traces graphically.  Following is a description of the trace control window.

Traces

The code for the trace control window is taken from ACT-R Vision, a graphical development environment for ACT-R written by Roman Belavkin at the University of Nottingham.  ACT-R Vision is a great tool (I use it when I do ACT-R on my Sparcstation), and I was interested in incorporating it with the model distribution, but most of its features did not work on the Mac under MCL 4.3 when I tried them.  This being the case, I simply lifted the trace control window (which does work) and stuck it into my code (properly credited, as you may notice).  It should additionally be noted that by the time you read this, ACT-R Vision might very well work with MCL, but I do not know for sure.
    For a detailed description of all of the traces, see the ACT-R 4 manual at http://128.2.248.58/inter/Welcome.html (you will have to log in - just follow the directions on the site).  For now, I recommend investigating the production and partial matching traces if you are interested.  Others may be interesting, if you wish to experiment.  These traces are best utilized in conjunction with the step function - you can step the model one production cycle from the step button on the control panel.
 
 
 

Model Behavior Window

This window displays aspects of the model's behavior, both overt and "internal".

Overt Behavior

The display labeled "Current Number" contains the model's latest subtraction answer.  If the answer is correct a blue check mark appears to the right of the number text, and the answer remains unchanged until the next response is given.  If the answer is incorrect, however, a red "X" mark appears over the number text and the model is "fed" the correct answer (as in the Tomaka et al. experiment), at which point the correct answer appears in place of the incorrect answer and the model resumes subtraction.
    The # of Attempts and # of Errors boxes keep track of number of subtraction attempts the model makes during a single trial, and how many of those attempts resulted in incorrect responses, respectively.  Task-Time in s indicates the total time spent on the task in ACT-R time (measured in seconds).

"Internal" Behavior

I have chosen to designate the following components as displays of the model's internal behavior because these components allow the observation of cogntive activity that would not be observable in human subjects.  They are representations of mental activities that are not reported, nor could they be reported, by humans undertaking the serial subtraction task.
    Specifically, the components in this category are the matrix labeled "Number in Memory", which I will call the chunk matrix, and the blue circle labeled "Worry Indicator".  The numbers in the chunk matrix represent declarative memory chunks encoding knowledge of numbers (e.g. Eight ISA number value eight).  When these chunks are retrieved during mathematical processing, they are outlined in the chunk matrix with a red square.  For example, if the model is processing the rightmost column in the expression 5178 - 7, and is retrieving a known subtraction fact (that eight minus seven equals one), eight will be highlighted followed by one.
    The worry indicator simply tracks the firing of the worry production if worry is turned on in the model.  When worry fires, the worry indicator turns red.
 

Appraisal Window

This window is fairly self-explanatory.  At the bottom is a box labeled "Pre-Task Appraisal".  The text in this box indicated the pre-task appraisal made by the model at the beginning of the task (or, most likely, that you selected from the control panel).  Throughout an individual trial, there are no changes to this window until the end of the task, at which time a post-task appraisal is made.  If you are running the model for only one trial, the post-task appraisal will appear at the end of the task and remain static.  If you are scripting the model, the post-task appraisal will only be displayed briefly at the end of each trial, and at the start of the next trial, the post-task appraisal for the previous trial becomes the current pre-task appraisal, leaving the post-task appraisal box empty once again.
 
 
 

Script Data Window

The script data window is useful for viewing he results of one four-trial script.  This window is typically used in conjunction with the SCRIPT 1 button on the control panel.  As each trial in the four-trial block is completed, all of the values (appraisals, number of attempts, and number of errors) are filled in to the appropriate box.
    Though output still goes to this window when longer scripts are run, the resulting data files (see above) are much more useful than trying to keep up with all the values as they are displayed.
 

Last modified January 20, 2002