Sign up here for the weekly email newsletter

The Human as Interface

Thoreau said in Walden that “we do not ride on the railroad; it rides upon us.” He was talking about the true cost of the ride—that while he embarked on a trip by foot and arrived at his destination that same day, you would have to first get a job and work for a day in order to earn enough money to make the same trip by train. “Instead of going to Fitchburg, you will be working here the greater part of the day. And so, if the railroad reached round the world, I think that I should keep ahead of you.”

Generalizing his assessment, Thoreau found that so much in the modernizing world bore a similar cost. The railroad and the industrialization it embodied was a powerful enough force to remake society in its image because of the demands placed upon us to build, maintain, and operate those systems, as well as our willingness to accept those demands in exchange for speed and efficiency.


 People helping computers talk to people (source

It makes sense why all of that heavy infrastructure weighs us down, why it’s such a chore, just like Thoreau’s image of a train driving on tracks made of people makes such vivid sense, or why a household cluttered with stuff feels so oppressive. But haven’t we escaped the era of stuff and ventured into the era of information, finally mastering the former? Software shouldn’t be able to “ride upon us” like industrial machinery can. If anything it should be loosening the physical world’s grip on us (and in many ways it is). Information storage is basically free and processing power increases exponentially and yet we don’t feel certain that we are freer than we used to be, or any less burdened. Why does email now ride upon us like trains once did?

Commenting on artificial intelligence as it reaches a tipping point in its maturity, John Robb recently observed that “we can’t even design systems that work for human beings”—that is, we’re designing AI as a godlike force that works in mysterious ways, not a true agent of our own objectives, and ensuring that we will somehow bow to it, just like we did to the industrial behemoths that we built in a previous era.

Put another way: Every medium-sized company with a competent customer service operation automates a large chunk of that work. When you call an airline or a credit card company, you pass through a tree of often-frustrating multiple-choice menus before getting your issue resolved. You only get to speak to a live operator after exhausting the menus’ abilities. That process of escalation is the Human as Interface—a reversal of roles.

The Human as Interface is the troubling but darkly funny outcome of our white-hot progress in the digital realm. An interface is traditionally a point of contact between people and computers (or between hardware and software, or two separate software systems) that eases their interaction and translates between two modes of communication.

Software is eating human work so fast that there’s less of a role for interfaces between humans and computers, as the latter can finish more and more of their work without humans dropping in partway through the process to guide them. At the same time, that software is doing even more of the jobs that humans used to do and eliminating the need for those jobs. Finally, the various activities that computers do are becoming so sophisticated that humans can’t only not understand them, we don’t even have a language for describing them. The gap between human and computer abilities is either closing or widening, depending on how highly you regard humans, and there’s a shortage of a different kind of interface or API: the kind that mediates between software and its human users in the transitional phase before a computer can handle that step too.

Thus, machines need people to translate between themselves and their users—the Human as Interface. This is a form of turking, in the sense that it’s yet another role humans only fulfill until software learns how. This type of work is found at every ability level: Customer support reps who handle the overly complex issues that automated systems escalate. Convenience store employees who help customers get unstuck from the self-checkout machines that replaced all the other employees. Explainers who can communicate to a broader audience a concept like machine learning and why it matters. IT help desks.

It’s surely a sign of increasing economic polarization that a small percentage of specialized individuals build and run the advanced systems that transform everyone else into a user in both their work and their free time. For this majority, their jobs await imminent automation, at best, and already function as an interface for machines otherwise (meanwhile everyone’s a user of some kind in their free time). No matter the reason for this condition, it’s hard to pretend that we don’t somehow work for computers, or that software doesn’t ride upon us as heavily as the railroads did.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s