by James Bailey |

Stephen Wolfram, in a recent blog entry, makes a compelling case for computational essays. Students have an easy time embedding images in their school work. If they know how to create them, their word processors now allow them to embed movies as well. Computational essays go the final step and embed executable code that the reader can invoke and watch the result. Standard word processors do not allow live code, but the notebooks of the Wolfram Language do.

The effect can be startling, almost like the cast of a Broadway musical suddenly bursting out into song to make a point. The absence of this capability, meanwhile, can be hugely frustrating. Authors of scientific papers routinely relate in text what they saw on their own computer screens.

"

I stayed up most of one weekend to get our first hexagonal [cellular automata] model experiment running on a high-resolution display ... The next morning I found a group of physicists, many of them fluid experts, staring silently at my screen in total shock ... and awe.

"
"

Sometime around 1980 we all got computers... You could simulate actual systems unfolding, so you could watch things unfold on your computer... We watched what happened… If somebody showed me how something was actually changing and unfolding, how new structures arise and fall away and further ones arise, I found that interesting.

"
"

The occasion was the annual Darwin Lecture in 1996, sponsored by Darwin College, Cambridge. We were midway through one of the lectures, listening to a record[ing] of NETtalk learning to pronounce English ... To judge by the silence of the audience, most listeners found the process uncannily realistic.

"
"

Supply your computer with the relevant biochemical data, concentrations, rate constants, and so on, then watch the screen!

"

Fine, but we want to see and hear those patterns, too. And we want to experiment with different initial conditions, to see if different patterns emerge. We want to briefly recapitulate the research that the authors did, to deepen our understanding. Computational essays offer this opportunity.

At the student term paper level, we want to see enough to verify that the student’s conclusions are plausible, that they are grounded in what their code actually shows.

One challenge for computational essays is to make that reader recapitulation easy. We want to see, in minutes, what the author no doubt spent hours and days arriving at. That means that the embedded code needs to offer an intuitively obvious interface: a few sliders or knobs that allow a modicum of experimentation.

The new book Canon/Archive edited by Franco Moretti is an example of a text that cries out for some embedded code. He is analyzing a large corpus of novels. What he sees at the level of individual words is different from what he sees in complete sentences, and different again to the patterns at the paragraph level. There is no way to recapitulate by just leafing through the novels one at a time. The reader wants to play with the same algorithms Prof. Moretti is playing with and apply them to the same open source data. To be continued.

tags:

Commenting is now closed.