Software to replace thinking
It’s not enough that technology is disrupting the financial model that has long sustained journalism. Now researchers are developing software that offers a new way to read.
Great. That leaves just one more step — computer code that second-guesses everything it bumps up against — and my line of work is done for.
But first, that reading breakthrough: The announcement came from Cornell University the week before Election Day, an unfortunate bit of timing, since better reading skills on the part of voters might have made this year’s campaigns a bit less contentious. Manipulative campaign advertising wouldn’t work, you know, if it weren’t for an electorate that hasn’t read up on the issues.
Researchers in Cornell’s Communication and Information Science departments have applied a concept known as “selectional preference” to content from major media outlets and political blogs. In linguistics, “selectional preference” is a way of identifying relations among words in order to help analyze how we think about a concept.
Cornell’s announcement of the new online tool, which it calls “Reflext,” cites the word “water,” and its typical association with such verbs as “pour,” “flow,” “freeze” and “splash.” Fine. You wouldn’t expect a noun like “water” to be associated with verbs like “snowshoed” or “indemnified.”
But take the word “Medicare.” In the text studied by Cornell’s researchers, it often appeared with such words as “slash” and “end,” but also with “reform” and “save.”
Those are very different ways of explaining what may be the same policy choices. One person’s plan to “reform” something may be another’s view of “ending” it. You can label either a tax cut or a tax hike as “reform.”
It’s by analyzing the frequency of such word associations that the Cornell researchers say they can better grasp what Americans are thinking.
“Examining how we talk about issues can provide insights into how we think about them,” said Eric P.S. Baumer, a researcher working on the project.
Long before Cornell began this sort of research, journalists were struggling daily to figure out just the right word to describe what they learn. A word that’s neutral may be dull; a more powerful word may be inaccurate.
This isn’t just about avoiding what is trite. We’re well-practiced in that from the get-go. One of my early editors shoved back a piece of copy I had written reporting that ambulances had “rushed” fire victims to the hospital. “Ambulances always rush,” he said. “Don’t write stuff that isn’t news.” In my brief career as a sportswriter, I covered only one golf tournament, which is fortunate, because I quickly ran out of verbs: Golfers, I learned, either “shoot” or “fire.” Boring.
This issue was well-known to Samuel Clemens, whose career as the writer Mark Twain began in newspapering. “The difference between the almost right word and the right word is really a large matter,” Twain wrote to a friend. “It’s the difference between the lightning bug and the lightning.”
What Cornell researchers hope to do, however, is go beyond precision of word choice. They aim to detect nothing less than what is said between the lines — the patterns in language use that suggest how people are discerning issues.
It’s the next step beyond a word cloud, a rather simple bit of software that gives visual weight to words that appear most often in a text. In a word cloud of the U.S. Constitution, for example, the word that emerges as most prominent is “States,” followed by “President.”
What might a “selectional preference” analysis of last week’s news show? Perhaps the connection of “general” to “scandal,” or “fiscal cliff” to “taxes,” and “entitlements?” During a campaign season, it might reveal word manipulation on the stump.
But blaming candidates and their backers for language that distorts reality diminishes our own responsibility. Persuasion has been an energetic human practice since the first cave-dwellers’ mating rituals. It’s the task of person receiving the message to sort the truth.
We who labor in journalism can help, and the Cornell research might be useful. But there’s no substitute for careful reading: watching out for value-laden words, for phrases that seem to twist your mind and for rhetorical flourishes that shroud complex issues in simplicity.
That doesn’t take software — just the human brain we’re all already packing.
Rex Smith is editor of the Times Union.