The Algorithm Ate My Homework

Algorithms are useful tools. But who's watching how they're used?

If I were to bump into someone who blurted out that I look 77 years old, I’d be inclined to clench my septuagenarian fist and punch him in the chops. Like any other self-respecting woman of a certain age, I’m touchy about this sort of thing.

The one saving grace in this hypothetical situation is personal contact. Barring fisticuffs, I could look the shmuck in the eye and let him know that I’m really offended. I know who’s accountable for this really bad guess.

On a good day, I might even swallow my pride and ask him why he thinks I look 77.

But what if my accuser isn’t a person? What if it’s a machine telling I look like I was born during FDR’s second administration?

This changes everything – at least in my mind.

Earlier this month in Japan, I encountered into an image recognition machine called “FieldAnalyst.” It was being demonstrated at NEC’s booth at Embedded Technology 2016. The system was running on Renesas’ chip. I was there to cover the show, watching the demo and taking pictures. 

I suddenly noticed on its screen that the FieldAnalyst had just scoped me out and pegged me as a 77-year-old female.

Imagine my horror.

See me -- holding a camera -- in the screen pegged as a 77 years old female. (Photo is shown above)

Was it my baggy eyes? I’d been working like a sled-dog over the last few days with very little sleep and a bad case of jet lag. Or was it the gray hair that I haven’t had time to color?

I couldn’t help but ask: “Why me? Why 77? Where does this stinking machine get off even bringing up the subject? Who asked it?”

If I were to pose all these questions — to the machine? to a person? — the inevitable answer would be: “It’s just an algorithm. It isn’t perfect yet.”

Right.

What got under my skin isn’t really my age. It dawned on me that I had nobody to whom I might complain. How do I argue with algorithms? How do I make algorithms accountable for its pre-programmed astigmatism?

In particular, when companies who make these infernal machines tell me, hey, whaddya gonna do, it’s just an algorithm, how do I respond? What’s my recourse? There’s no perfect retort to an algorithm that cannot feel the sting of my wit. Or feel anything? Or say, “Oops.”

But in today’s world, we’re surrounded by a sea of algorithms.

These days, machines make a lot of decisions for us. They seem to know which news stories I want to read or which books I might want to buy next. Google Map knows the best route I can take from point A to point B. Or thinks so. Machines seem to be even capable of predicting who is likely to commit the next violent crime.

And oh, by the way, they know how old I am. Or think they do.

A multitude of tech companies -- Google, Facebook, Microsoft and Amazon included – are striving to refine their algorithms. As a reporter of EE Times, I approve.

After all, algorithms are king. They are the tech companies’ proprietary tools and secret sauce. They’re vital to making their services and products smarter.

But here’s the rub. I see little evidence that any of these companies are being held accountable for their algorithms, let alone for their discretion (or lack thereof) in how they’re used.

Too often, companies use algorithms as a shield — an excuse — for not having to explain what the algorithm did, or failed to do. I hear the voice of Freddy Prinze: “Hey, ees not my job, man.”

ProPublica, an independent non-profit newsroom, recently launched a series of experiments that allowed them to look inside a “black box.”

By “black box,” what do they mean? The reporter explained: “When algorithms used to make those decisions aren’t transparent to the user, we call it a black box… All too often, it’s impossible for outsiders to know what’s going inside a black box.”

Next page: Pandora's black box

To read on, please click here.