TITLE: Workshop 3: Computational Thinking in Physics
AUTHOR: Eugene Wallingford
DATE: October 31, 2008 10:26 AM
DESC:
-----
BODY:
*[A transcript of the
SECANT 2008 workshop:
Table of Contents]
*

As much as computation is now changing biology, it has
already changed physics. Last year's workshop had a
full complement of physicists and astronomers. In
their minds, it is already clear that
physicists must program
-- even students learning intro physics. The question
is, what problems do they face in bringing more
computation to physics education? This panel session
shared some physicists' experience in the trenches.
Bruce Sherwood, the panel chair, set the stage: We
used to be able to describe physics as theory, experiment,
and the interplay between the two. This is no longer
true, and it hasn't been for a while. Physics is now
theory, experiment, *simulation*, and the
interplay among the three! Yet this truth is not
reflected in the undergraduate physics curriculum --
even at so-called "respectable schools".
Rubin Landau
described a systemic approach, a Computational Physics
major he designed and implemented at Oregon State. He
was motivated by what he saw as a turning inward of
physics, efforts to cover all of the history of physics
in the undergrad curriculum, with a focus on mathematics
from the 19th century, and not looking outward to how
physics is done today. (This CS educator felt immediate
empathy for Landau's plight.) He noted his own
embarrassment: computational physicists at major physics
conferences who refuse to discuss their algorithms or
the verification of their programs. This is simply not
part of the culture of physics.
Students learn by doing, so projects are key to this
Computational Physics curriculum. Students use a
"compiled language", which is Landau's way to distinguish
programming in a CS-style language from Mathematica
and Maple. For him, the key is to separate the program
from engine; students need to see the
program as an idea.
Two languages is better, as that gives students a chance
to generalize the issues at play in using computation
for modeling.
The OSU experience is that the political issues in
changing the curriculum are much tougher to solve than
the academic issues: the need for budget, the resistance
of senior faculty, the reluctance of junior faculty to
risk tenure, and so on.
Landau closed by saying that, for physics-minded students,
using computation in physics and then taking a CS course
seems to work best. He likened this to the use of Feynman
diagrams in grad school: students learn to calculate with
them, and then learn the field theory behind them the
next year. His undergrads have several "A-ha!" moments
throughout CS1. I suspect that this approach would work
for a lot of CS students, too, if we can get them to use
computation. Media computation is one avenue I've seen
work with some.
Next up was Robert Swendsen, from Carnegie-Mellon. In
the old days, physicists wrote programs because they did
not know how to solve a problem analytically. Now, they
compute to solve problems that *no one* knows how
to solve analytically. (Mental note: It also lets them
ask new questions.) The common problem many of us face:
we tend to teach the course we took -- something of a
recursion problem. (Mental note: Where is the base
case? Aristotle, I suppose.)
Swendsen identified a few other challenges. Students
are used to looking at equations, though if they don't
get as much from them as we do, but they have no experience
looking at and reasoning about data. They struggle even
with low-level issues such as accuracy in terms of the
number of significant digits. Further, many students
do not think that computational physics is "real" physics.
To them, physics == equations.
This is a cultural expectation across the sciences,
a product of the few centuries of practice. Nor is it
limited to students; people out in the world think of
science as equations. Perhaps they pick this notion
up in their high-school courses, or even in their
college courses. I think that faculty in and out of
the sciences share this misperception as well. The
one exception is probably biology, which may account
for part of its popularity as a major -- no math! no
equations! I couldn't help but think of Bernard
Chazelle's efforts to popularize the notion that
the algorithm is the idiom of modern science.
Listening to Swendsen, I also had an overriding sense
of *deja vu*, back to when CS faculty across
the country were trying to introduce OO thinking into
the first-year CS curriculum. Curriculum change must
share some essential commonalities due to human nature.
Physicist Mark Haugan focused on a particular problem
he sees: a lack of continuity across courses in the
physics curriculum with respect to computation.
Students may use computation in one course and then
see no follow-through in their next courses. In his
mind, students need to learn that computation is a
medium for expressing ideas -- a theme regular readers
of this blog will recognize. Mathematical equations
are one medium, and programs are another. I think
the key is that we need to discuss and work with
problems where computation matters -- think
Astrachan's Law
-- problems for which the lack of computation would
limit our ability to understand and solve the problem.
This, too, echoes the OO experience in computer science
education. We still face the issue that other courses
and other professors will do things in a more traditional
way. This is another theme common to both SECANT
workshops: we need to help students feel so empowered
by computation that they use it unbidden in their
future courses.
The Q-n-A session contained a wonderful thread on the
idea of physics as a liberal art. One person reported
a comment made by a student who had taken a computational
physics course and then read a newspaper article on
climate modeling:
**
Wow. Now I know what that means.
**

I can think of no higher "student learning outcome" we
in computer science can have for our general education
and introductory programming courses: *Wow. Now I
know what that means.*
There are many educated people who don't what "computer
model" means. They don't understand what is reported
in the news. There are many educated people reporting
the news who don't understand the news they are reporting.
That's not right.
-----