It's people!: a newsroom in a box

Just a response to a paper I’ve read related to human-computer interface design — one that hit me where I live, or used to.

“Soylent: A Word Processor with a Crowd Inside” describes a software project that amends the dreaded Microsoft Word with some crowd-sourced editing assistance. “Writing is difficult,” the authors observe — yeah, welcome to my world — before adding: “When we need help with complex cognition and manipulation tasks, we often turn to other people” (1). Sometimes we have support systems in place for this assistance, but sometimes not. The Soylent project crafts just such support for any writer-user, utilizing Mechanical Turk workers to farm out editing, proofreading, and formatting tasks to others.

Need someone to read over your paper — because you need suggestions as to what can be cut, because you want to make sure all the proverbial i’s are dotted and t’s crossed, because if you comb through your citations one more time your head will explode — but maybe you’ve called in that favor already or don’t want to risk bothering a colleague? Launch Soylent, which hires its invisible labor force to handle the work for you, perhaps in the dead of a deadline night.

What struck me about this project is how it attempts to replicate something electronically that has existed professionally for more than a century: the newsroom.

As someone who was a newspaper writer, critic, and editor for 20 years, one might expect me to have a natural and intensely negative response to this. Computers editing? Hogwash! But that’s the trick here: computers do not perform but rather facilitate the very human labor.

Having transitioned to academia, too — and witnessed the often urgent need for writing assistance felt by both students and professors — I absolutely see the value of this project, particularly in the fluid, painless, and (importantly) pedagogical way the Soylent creators have integrated the outsourced tasks into the existing Word interface.

Soylent features three parts: Shortn, “a text shortening service that cuts selected text down to 85% of its original length typically without changing the meaning of the text or introducing errors”; Crowdproof, “a human-powered spelling and grammar checker that finds problems Word misses, explains the problems, and suggests fixes”; and the Human Macro, “an interface for offloading arbitrary word processing tasks such as formatting citations or finding appropriate figures” (1). The interface simply connects the writer to Turkers willing to handle these various (paid) tasks on demand.

I try not to be romantic about the art/craft/task of writing, and I’ve especially learned thus far in academia just how unromantic it can be, sometimes less about expression than sheer shoveling of data. The editor in me wants to caution about matters of style and context of skills — i.e., you likely wouldn’t want a newspaper copy editor working over your academic research paper and vice versa — so the matching of task and Turker could instigate problems as well as solving them, which the researchers here seem to be aware of in addressing the potential of lazy contributors. But the fact that this project designs a method for humans to interact with humans to complete the task at hand is what’s important; the machine and the code are the intermediary to the solution, not the solution itself.

Through the lens of my own experience, I recognize that the beauty of this software is that it creates the experience of a newsroom for any writer. What Soylent does is already built into the organization of a modern news-gathering organization. Newsrooms are large collections of people, each with a certain, ordered task along the assembly line of producing and publishing information. If a newspaper writer is stuck shortening something to fit into the assigned space on page A7, he or she may turn to the other writers sitting nearby. All newspaper text is proofread by copy editors as part of the production process (well, layoffs and desperate decision-making has made that claim less sturdy than it used to be, but the standard is still there). Higher “macro”-type editing functions can be performed by the line editors, section editors, or managerial editors available.

The important part of the newsroom experience, too, is that you learn from your mistakes through this built-in collaboration. But in this way, I’m wondering if the anonymity of Soylent and Turk might actually be an improvement on the direct human-contact system. Not only does Soylent provide a writer with options from which he/she can learn — I did not use the word “pedagogical” above lightly — but seems to completely remove any communication of judgment from the transaction. The Turkers simply offer: “Here are some things to consider; change ’em or don’t.” Soylent seems to err on the side of pointing out improvements, not piling up your errors, which should only add to the improvement of the writer’s skills. Now I’m eager to design a study of writers using this software and collect data to see if that does, in fact, occur.

What implications could such an idea could have on the future of journalism? In just a decade, the business has contracted severely, as managers try to make up for lost print-ad revenues by sometimes indiscriminately laying off journalists — with copy editors often the first at the chopping block. Could a newspaper actually incorporate Soylent into its production process, in place of a copy editing staff? Could news organizations join together to create their own single-source staff of copy editors, in place of the Turkers (which would be ideal, at least for the above-mentioned style and context considerations)? The mind reels — and shudders.