[ACT-R-users] Chunk Definition

[Dan Bothell]

--On Monday, September 01, 2008 4:48 AM -0700 sima najafi 
<nj_sima at yahoo.com> wrote:

>
> Dear All,
> I'm a beginner to ACT-R. nowadays I try to solve a second degree equation
> with ACT-R.
> I define the following chunk type:
> (chunk-type EqDeg2 a b c root1 root2).
> for sovling a second degree equaion normally we use the following formula:
> X = -b +/- (?(b^2 -4ac))/2a).
>  I saw some sample models in which lisp functions for computing the
> answer of the equation are being defined whenever such these situations
> occure.
> now i have a question:
> is it correct to define a formula in this manner or we should encode the
> formula
> in rules and chunks such as add model in unit 1 of ACT-R tutorials?
> I want that model be plausible psychologycally but dont' have enough
> knowledge in psychology.
>


Generally, whenever one uses calls to "code" outside of the model to
perform computations or operations that are attributed to the model
it brings the psychological plausibility of the model into question.
Note however that just because a model doesn't use any outside calls
doesn't mean it is automatically psychologically plausible.  The
plausibility of a model depends on many factors, and all the choices
one makes in modeling need to be justified to support the model's
plausibility.

For the task you mention, solving an equation, I think the plausibility
of using code to compute an answer would depend a lot on what the overall
task you are modeling is.  If solving the equations is the task being
modeled, then having code outside of the model compute the value doesn't
seem like a very plausible model to me since it's bypassing all of the
cognitive components of the model to achieve the result.  However, if the
task were instead a memory learning task for recognizing or remembering
answers to equations, then one could argue that the process of solving the
equation doesn't need to be handled in detail by the model because that
is below the level which the model is intended to address.  Other tasks
would have to be considered accordingly, and different people may have
different views as to what they would consider plausible models for such
tasks.

There are many publications of ACT-R models which perform mathematic and
algebraic tasks available on the ACT-R publications page which you may
want to consult for additional information on the specific topic:

<http://act-r.psy.cmu.edu/publications/index.php>

Hope that helps,
Dan