Then They Came for my Job — and No One was Left to Speak for Me

Everybody loves Martin Niemoller. Wikipedia will tell you he is famous for his “provocative poem about the cowardice of German intellectuals following the Nazis’ rise to power and subsequent purging of their chosen targets, group after group.” Here’s the original form of that poem, in case it hasn’t come across your Facebook feed lately:

First they came for the Socialists, and I did not speak out —
Because I was not a Socialist.

Then they came for the Trade Unionists and I did not speak out —
Because I was not a Trade Unionist.

Then they came for the Jews, and I did not speak out —
Because I was not a Jew.

Then they came for me — and there was no one left to speak for me.

It’s a lovely piece to quote: it reckons with the past, signals enlightenment and moral clarity, and challenges us all to live up to its ideals. You pretty much can’t help feeling a little more ethical and upright when you read it. Still, before we get too complacent, I’d like to ask you to accompany me, and Niemoller, on a short tour of recent American economic and political history. I want to see if we’ve really learned to speak up for others as well as we think we have. Humor me, you don’t really mind, do you?

First they came for the Socialists — this feels like a quaint anachronism now, but the Socialist party was highly relevant politically in the United States in the early 1900’s. The Socialists had two representatives in Congress and high water marks of 6% of the popular vote in the 1912 presidential election for socialist Eugene Debs and almost 17% in 1924 for socialist-supported progressive Robert LaFollette. (By contrast, Gary Johnson won 3% and Jill Stein won 1% of the popular vote in 2016.) Within 25 years, discredited by American politics and world events, the Socialists were more or less irrelevant in the U.S., winning less than 0.1% of the presidential vote from the 1950’s on. Some people did speak out, but the tide was against them. You and I weren’t around then, in any case.

Then they came for the Trade Unionists —  much of this happened in our lifetimes. Labor unions were highly influential in the United States for most of the 20th century, especially during what’s known as the Great Compression, the period following the New Deal reforms when income inequality declined dramatically. From Wikipedia:

This “middle class society” of relatively low level of inequality remained fairly steady for about three decades ending in early 1970s, the product of relatively high wages for the US working class and political support for income leveling government policies.

This is clear in this chart that tracks income inequality over time. This income leveling correlates well to labor union membership, which grew steadily during the New Deal, and eventually reached almost 35% — more than 1 in 3! — of salaried workers in 1954. Union membership gradually declined from that point. The best statistics start in 1983, when around 20% of the work force still belonged to unions, and they show another 50% decline to just 11% of the work force in 2015. For my generation (I am in my mid-40’s), and for the so-called new economy, I think we can say that we’ve played out Niemoller’s script pretty much as written: few of us were unionists, and few of us have paid attention as union influence declined throughout our working lives. (Uber rides are cheap and convenient, so who cares if the drivers have no rights and the company can raise its take from fares at will? I bet a hundred years ago there would have been a strike.)

Then they came for the Industrial Cities. Here I am talking about both big (Detroit, MI, population 1.85M in 1950) and small cities (Youngstown, OH, population almost 170K in 1960). Since then, both cities have shrunk by about 60%! The history of these cities (and many more like them) in the 20th century is a complex mix of economics, sociology, and racism, but here’s the world’s shortest introduction:

U.S. cities and industry in the North and Midwest grew rapidly in the late 1800’s and early 1900’s, powered by a huge wave of  European immigrants. Then World War I created a labor shortage, and blacks from the South migrated north to fill the gap. This began the Great Migration of blacks to cities in the North and (eventually) the West. Apart from a short break during the Great Depression, the migration continued for the next 40-50 years, transforming the black population from 80% rural-20% urban to the exact opposite — 80% urban-20% rural — in little more than half a century.

As they came to the white-dominated cities, incoming blacks were steered to neighborhoods with lower economic opportunity and investment, away from whites, a morally bankrupt process known as redlining. (For a history of redlining that goes beyond Wikipedia, read Ta-Nehisi Coates’s mighty essay, The Case for Reparations.) Meanwhile, whites tended to concentrate in their own segregated parts of the same cities, and eventually to move away from the cities altogether as road networks improved (white flight). Some of the industrial jobs that brought blacks to the cities went to the suburbs with the white population, some began to go overseas, and some were lost to automation, with the latter trend showing no signs of abating. In just 20 years (1967-1987), Philadelphia, New York City, Detroit, and Chicago all lost over 50% of their manufacturing jobs. Some cities managed to reorient around the technology or financial sectors (Boston, New York), while others lost significant population and income that they’ve never regained. (For more on the relationship between lost industrial jobs and inner city poverty, you can read William Julius Wilson; here is a short summary paper.)

Who spoke out for these cities, for their minority populations and for their financially starved schools? Not those whites who moved to the suburbs to get away from blacks, and not the mainstream Democratic party of the 1980’s and 90’s, which decided it needed to distance itself from inner city concerns in order to win elections. Maybe you did?

Then they came for the Heartland. Or perhaps for the U.S. manufacturing economy, depending on whether you prefer to view things geographically or economically. This has been the subject of much debate since the election, because Trump’s margin of victory came from three “Rust Belt” states (PA, MI, WI) that he had been expected to lose, and because it’s believed that his margin of victory in those states came from dissatisfied white working class voters either seeking change or lashing out either at bullying elites or at bullied minorities or immigrants. (This is probably a lousy explanation of voter dynamics in Pennsylvania.) Frustratingly, these discussions have devolved into arguing over whether these voters are motivated by economic or social-cultural factors (as though it were easy to separate the two) and whether they deserve sympathy (as though we’ve never seen anyone bully others while being bullied at the same time).

No matter how you view the politics here, we should be able to come to some kind of agreement on the economic facts. Here’s a recent article suggesting that the Rust Belt is not a struggling region, because other parts of the country are worse off: the states we are talking about are all in the middle third of U.S. states by median income. Sure, they’re not Connecticut or California, but they’re not Arkansas or Mississippi either. This is true enough as far as it goes, but it fails to consider decline. As a quick back-of-the-envelope exercise, I pulled up the income data and sorted the states top to bottom by median income as of 2015 and as of 1995. Here’s what I found:

  • The state with the biggest decline in ranking over the last 20 years (from #5 to #28)? Wisconsin.
  • The state with the second biggest decline (from #14 to #31)? Michigan.
  • Five of the ten states with the biggest drop in ranking form a contiguous region at the heart of the Rust Belt: Wisconsin, Michigan, Illinois, Indiana, and Ohio. (There are no other clusters; the other five states are completely non-contiguous.)

At this point, these Rust Belt jobs are more likely lost to automation than to other states or countries (though there is some of the latter) or incoming immigrants from lower-income countries (virtually none coming into the Midwest). So it’s hardly clear what to do. Still, there’s not much sympathy for these people’s plight on my Facebook feed, where most people (1) have college or graduate degrees and professional occupations, (2) live in desirable areas, and (3) are frustrated and terrified, as I am, to see Trump in power.

Then they came for — OK, who’s next? Are you? What job will you have in twenty years, or what job will your children have, when:

  • Economic opportunity appears to be declining, not increasing, across generations.
  • The technology sector, which is not very labor intensive, and also finds it easy to move jobs around or offshore, is coming to dominate the (non-service) economy. (Just to give you an idea, Facebook is the 7th most valuable company in the world and employs only 13,000 people, a fraction of the workforce of the industrial behemoths of 50 years ago, whose employees numbered in the hundreds of thousands.)
  • Automation is increasing, and potentially applies to more and more jobs, both in and out of the tech sector, as technology itself becomes ever more capable.
  • The gig economy is rising, and there doesn’t appear to be an obvious limit on what kind of jobs can be gig-ized. (Gig workers have even less leverage than long-term employees.)
  • The financial industry, which has grown rapidly over the last several decades, provides relatively little economic value relative to the human capital invested in it, and might be ripe for contraction. You could speculate that the same might be true of marketing as well (another industry where too many of the well-educated settle), or at least that it could be done more cheaply. If our news stories can be replaced by crap written on the cheap in Eastern Europe, why not the ads?

Shifting into my ominous Rod Serling voice: with the manufacturing economy gone, both in the cities and out, with more and more people shifted into low-wage service jobs, with labor as powerless as it seems it’s ever been, with our economy sliced up into strata, and the humans in each one of those strata cut away in turn — who will be left to speak for you when they come for you? And how will you feel with Niemoller’s warning, turned into prophecy now, running on a loop over and over, not just on your Facebook feed but inside your head?

Cathy’s Book is Out!

Cathy O’Neil’s book Weapons of Math Destruction is out, and it’s already been shortlisted for a National Book Award! Here is a review of the book that I posted on Amazon.com:

So here you are on Amazon’s web page, reading about Cathy O’Neil’s new book, Weapons of Math Destruction. Amazon hopes you buy the book (and so do I, it’s great!). But Amazon also hopes it can sell you some other books while you’re here. That’s why, in a prominent place on the page, you see a section entitled:

Customers Who Bought This Item Also Bought

This section is Amazon’s way of using what it knows — which book you’re looking at, and sales data collected across all its customers — to recommend other books that you might be interested in. It’s a very simple, and successful, example of a predictive model: data goes in, some computation happens, a prediction comes out. What makes this a good model? Here are a few things:

  1. It uses relevant input data.The goal is to get people to buy books, and the input to the model is what books people buy. You can’t expect to get much more relevant than that.
  2. It’s transparent. You know exactly why the site is showing you these particular books, and if the system recommends a book you didn’t expect, you have a pretty good idea why. That means you can make an informed decision about whether or not to trust the recommendation.
  3. There’s a clear measure of success and an embedded feedback mechanism. Amazon wants to sell books. The model succeeds if people click on the books they’re shown, and, ultimately, if they buy more books, both of which are easy to measure. If clicks on  or sales of related items go down, Amazon will know, and can investigate and adjust the model accordingly.

Weapons of Math Destruction reviews, in an accessible, non-technical way, what makes models effective — or not. The emphasis, as you might guess from the title, is on models with problems. The book highlights many important ideas; here are just a few:

  1. Models are more than just math. Take a look at Amazon’s model above: while there are calculations (simple ones) embedded, it’s people who decide what data to use, how to use it, and how to measure success. Math is not a final arbiter, but a tool to express, in a scalable (i.e., computable) way, the values that people explicitly decide to emphasize. Cathy says that “models are opinions expressed in mathematics” (or computer code). She highlights that when we evaluate teachers based on students’ test scores, or assess someone’s insurability as a driver based on their credit record, we are expressing opinions: that a successful teacher should boost test scores, or that responsible bill-payers are more likely to be responsible drivers.
  2. Replacing what you really care about with what you can easily get your hands on can get you in trouble. In Amazon’s recommendation model, we want to predict book sales, and we can use book sales as inputs; that’s a good thing. But what if you can’t directly measure what you’re interested in? In the early 1980’s, the magazine US News wanted to report on college quality. Unable to measure quality directly, the magazine built a model based on proxies, primarily outward markers of success, like selectivity and alumni giving. Predictably, college administrators, eager to boost their ratings, focused on these markers rather than on education quality itself. For example, to boost selectivity, they encouraged more students, even unqualified ones, to apply. This is an example of gaming the model.
  3. Historical data is stuck in the past. Typically, predictive models use past history to predict future behavior. This can be problematic when part of the intention of the model is to break with the past. To take a very simple example, imagine that Cathy is about to publish a sequel to Weapons of Math Destruction. If Amazon uses only  purchase data, the Customers Who Bought This Also Bought list would completely miss the connection between the original and the sequel. This means that if we don’t want the future to look just like the past, our models need to use more than just history as inputs. A chapter about predictive models in hiring is largely devoted to this idea. A company may think that its past, subjective hiring system overlooks qualified candidates, but if it replaces the HR department with a model that sifts through resumes based only on the records of past hires, it may just be codifying (pun intended) past practice. A related idea is that, in this case, rather than adding objectivity, the model becomes a shield that hides discrimination. This takes us back to Models are more than just math and also leads to the next point:
  4. Transparency matters! If a book you didn’t expect shows up on The Customers Who Bought This Also Bought list, it’s pretty easy for Amazon to check if it really belongs there. The model is pretty easy to understand and audit, which builds confidence and also decreases the likelihood that it gets used to obfuscate. An example of a very different story is the value added model for teachers, which evaluates teachers through their students’ standardized test scores. Among its other drawbacks, this model is especially opaque in practice, both because of its complexity and because many implementations are built by outsiders. Models need to be openly assessed for effectiveness, and when teachers receive bad scores without knowing why, or when a single teacher’s score fluctuates dramatically from year to year without explanation, it’s hard to have any faith in the process.
  5. Models don’t just measure reality, but sometimes amplify it, or create their own. Put another way, models of human behavior create feedback loops, often becoming self-fulfilling prophecies. There are many examples of this in the book, especially focusing on how models can amplify economic inequality. To take one example, a company in the center of town might notice that workers with longer commutes tend to turn over more frequently, and adjust its hiring model to focus on job candidates who can afford to live in town. This makes it easier for wealthier candidates to find jobs than poorer ones, and perpetuates a cycle of inequality. There are many other examples: predictive policing, prison sentences based on recidivism, e-scores for credit. Cathy talks about a trade-off between efficiency and fairness, and, as you can again guess from the title, argues for fairness as an explicit value in modeling.

Weapons of Math Destruction is not a math book, and it is not investigative journalism. It is short — you can read it in an afternoon — and it doesn’t have time or space for either detailed data analysis (there are no formulas or graphs) or complete histories of the models she considers. Instead, Cathy sketches out the models quickly, perhaps with an individual anecdote or two thrown in, so she can get to the main point — getting people, especially non-technical people, used to questioning models. As more and more aspects of our lives fall under the purview of automated data analysis, that’s a hugely important undertaking.

 

Ownership

Watching Game 5 of the Giants-Cardinals series. Top of the first, Cards have runners on first and second, one out. The atter hits a line drive to third base. Giants’ third baseman Pablo Sandoval leaps up, catches the ball (the batter’s out), throws quickly to second. Looks like the ball gets there just a hair before the runner on second can dive back in. Umpire calls the runner out at second — double play! Here comes the Cards’ manager to argue with the ump.

Except that this year, baseball uses instant replay. Here’s how it works: a manager has the right to challenge most plays, asking for an umpire’s call to be overturned based on a review of the replay. The caveat is that if you lose a challenge (meaning that after a review of the replay, the call on the field is upheld), you also lose the right to challenge for the rest of the game. So when you challenge, especially early in the game, you better be sure that the umpire’s wrong.

The Cards’ manager speaks briefly with the umpire. Meanwhile, someone on the Cards’ side is reviewing the replay. They must decide that the umpire’s probably right (they don’t challenge and risk losing the right to challenge in the future), because the manager returns to the dugout. It all takes less than a minute.

When replay was introduced, the worry was that managers challenging calls all the time would slow down the game. Only here it feels like it’s actually sped up the game. If you’ve watched enough baseball, you’ve seen many long arguments, frustrated managers venting endlessly to umps who never had any mechanism to change their mind. But with the new rule, the manager has (1) a lot more control, and (2) a strong incentive to be correct. The Cards’ manager gets to make a decision, he decides the call against his team was right, and we quickly move on.

Empowering people, based on the right incentives, can go a long way.