Tendencies toward centralized thinking

[guzdial@cc.gatech.edu (Mark Guzdial)]

>At 2:29 PM -0400 9/29/97, billcl wrote:
>>Mitch Resnick has mentioned how people see central control even when none is
>>present.  I am curious if any one knows a method (psychological instrument,
>>experiments, etc) that one can use to measure how strong that tendency is in
>>different individuals.  One way is simply to show them Starlogo demos and
>>ask them what to explain what they saw.  Are there other ways?  Is there a
>>range that people are distributed along?
>>
>>Any thoughts or ideas?
>
>Perhaps the tendency is a manifestation of more general patterns of
>inference - see "The logic of plausibility reasoning:  A core theory"  A
>Collins and R Michalski, cognitive science, 13 1-49 (1989).  The authors
>describe relevant background research.
>
..
>Interesting enough, the literature on restructuring organizations reveals
>tendencies for resisting distributed control...

I think there's evidence that people tend toward designing, as well as
understanding, in a centralized way, at least in building software.

I had a paper in ACM SIGCSE a couple (several now?) years ago on
observations of Sophomore undergrads first learning Smalltalk and C++.
Their tendency is to either (a) create a single class which does everything
or (b) create several classes, but one class controls everything. (And I'm
still having trouble convincing students to do otherwise -- I had students
last quarter building classes that did nothing but hold a value, no
computational responsibility at all.)  My paper was inspired in part by
observations that Mitchel made in his ILE paper on MultiLogo, where he
noted that students had a difficult time understanding behavior spread
across multiple objects (e.g., telling a process "STOP" but confused
because the motor that the process had started was still running).  I'll
bet that if you did an analysis of freely available C++ programs on the
net, most of them are a single class or are highly centralized.

While even the best object-oriented design (i.e., one that distributes
responsibility for functionality across a variety of objects) is not as
decentralized as a StarLogo approach, the observation that students have a
hard time with this limited form of decentralization suggests that
students' tendency is to understand AND design in a centralized way.

Mark

- --------------------------
Mark Guzdial : Georgia Tech : College of Computing : Atlanta, GA 30332-0280
(404) 894-5618 : Fax (404) 894-0673 : guzdial@cc.gatech.edu
http://www.cc.gatech.edu/gvu/people/Faculty/Mark.Guzdial.html