Archive | Politics of Technology RSS for this section

The Problem of Data-Driven HR

A big thing nowadays is the proposition that “Big Data” will transform human resources, or HR – who to hire, who to fire, who to promote, etcetera.  It’s an interesting and apparently seductive idea, that by capturing more and more data employers will be able to tell everything about their employees.  I was discussing this last night and have been trying to think it through with my “social scientist” cap on.  I think that if analytics does come to drive a lot of HR decisions it will do so much more slowly than we think.  For a few reasons:

  • Garbage in, garbage out. Employee ratings systems are notoriously terrible, and are incredibly biased estimators of employee ability, or value added, or what have you.  Employee quality is a notoriously slippery latent variable, and hard to even coherently define, much less measure.  If we can’t trust the end measurement, we can’t trust anything we find about its determinants.
  • The complexity of the problem invites spurious correlations. There is an absolute ton of data available about people, ranging from the simple – demographics, college GPA – to the complex, like social media activity and text-processing of writing samples.  This is great, but it means that as long as there are a sufficient number of employees you can always find some relationship that is “statistically significant” by hunting through enough variables.  Doing this virtually guarantees you will be directing HR decisions based on spurious relationships. For example, if you take 100 variables known to have no relationship to the outcome and a large, almost certainly you will find 5 with statistical significance as predictors at the 95% confidence interval.  That’s what the confidence interval means – you’re wrong 5% of the time.
  • Non-comparability of employees. This goes back to the first point – for some businesses, there are armies of employees doing similar tasks.  For most businesses there aren’t, and it’s difficult to reduce performance to the same axis.  A software developer and a salesperson simply can’t be measured on the same scale, it’s not meaningful.  It’s often even difficult to compare two salespeople, especially if individual sales are large things like software packages or consulting projects – the more skilled employees tend to take on more challenging assignments.  For companies comprising large amounts of white-collar works doing distinct tasks, the whole mission may be hopeless.

On the upside, some of these actually seem like problems that can be fixed.  Probably the starting point is to examine what actually drives employee ratings – I have a feeling that the findings might be unsettling.  At most companies, I would expect to see a major impact of things like race and gender, and a fairly loose relationship between ratings and actual performance as measured by whatever objective measures are at hand – e.g., client happiness, productivity, project cost/time overruns.  This seems like one case where you’d actually want to have an outside consultant, because your own company’s managers might not be the most objective.  The point in measuring the bias isn’t to chastise everyone – it’s so that HR can normalize ratings behind the scenes and hope to actually improve measurements of employee quality without having to trust individuals’ judgments as much.

It’s not a coincidence that most of these “Big Data HR” stories are coming out of firms that can objectively measure performance like call centers.  The garbage-in-garbage-out problem is extremely difficult, and most of the firms that are currently using HR analytics without clear objective measures are probably putting out garbage.  That needs to be fixed before anyone can hope to use serious analytics in the HR space, and it’s an incredibly difficult problem that likely has no obvious best solution.  Only after the social science foundations are in place do firms have to worry about the Polanyian problem

Data-Driven Discrimination is Coming Whether We Like It or Not

The availability of data makes discrimination easier.  The example the authors start with, of bus schedules, is a pretty trivial example, but it gets real pretty quickly.  Health insurance is the most famous example of this, as it is a business governed solely by adverse selection.  The ideal business model for health insures is to provide exactly zero healthcare reimbursement, and they would achieve this by denying coverage to anyone who might need it. This type of concern is gradually permeating most of the spheres characterized by information asymmetry – those where individuals are dealing with businesses or individuals who would like to know more about them than the individuals would like to put forward.  There’s a spectrum of reasonableness here – at the top is a criminal record, which is why employers conduct background checks.  At the bottom is political or religious beliefs, things that are immaterial to almost any purpose other than discrimination.  But much of what’s in between is characterized by a fair amount of dispute over what might be considered a reasonable subject of interest, and that ambiguity should concern those of us interested in preventing unfair discrimination.

Some of these categories might well have valid applications.  For example, let us imagine a world where the vast majority of white-collar criminals are middle-aged straight white males.  Suspend your disbelief for a moment, please.  If that’s the case, presumably an employer would be justly cautious when hiring a middle-aged straight white male and would wish for further information to assuage their confidence.  They could use the vast amount of publicly available data to look for high-risk predictors of white-collar crime – past convictions, criminal head shapes, degrees from Harvard Business School, etcetera.  The argument can be made that the innocent have nothing to hide, and indeed that’s true – in fact, because of the same logic, poor black men actually benefit on net from mandatory drug testing by employers.  More information only helps the innocent.  And yet it is a hop, skip, and a jump from that logic to the logic of redlining.  It’s not even a particularly slippery slope, but rather an application of the same logic that underlies price discrimination in many of the areas where “Big Data” is applied today.

I think the political economy of this is unsustainable.  The use of personal data in ways that today seem highly inappropriate will only continue and will become even more personal and intrusive over time.  It’s not just bus schedules – it’s car insurers trawling Facebook to see whether you’re high-risk, and even worse, once Facebook figures out the algorithm for high-risk driving they will be happy to sell that data to whoever has the cash.  People resist the intrusion of the market into their lives, and the monetization of people’s being will eventually trigger intense political backlash and a desire to declare certain uses of data out of bounds.  “Privacy” is just the tip of the iceberg – people will demand protection from the precise measurement and metering of their souls that Silicon Valley is out to create.

I wish I had the answers for what policy might serve these goals, but I suspect policy responses will mostly be inadequate to the scale of this societal change.

Money is Indistinguishable from Magic

Noah Smith points out that Neal Stephenson, an excellent science fiction writer, has a weird obsession with money and some wacky ideas about it.  I view that as pretty forgivable, mainly because money was the main MacGuffin of Stephenson’s Cryptonomicon, one of my favorite fun novels.  I say MacGuffin because the novel wasn’t really about cryptocurrency or lost Japanese gold, those were just the driving forces for an excellent adventure.  But Stephenson is rather obsessed with gold, and succumbs to the all-too-common fallacy that only gold, unlike money, is a “stable store of value”.

One science-fiction author who has done some more serious and interesting thinking about money is Charles Stross, specifically Neptune’s Brood.  This is also an adventure tale, but one that is much more visibly obsessed with the question of money.  The question the book attempts to answer is an inherently goofy one, “how can interstellar exchange work without faster-than-light communication?”.  The basic answer Stross comes up with is “slow money”, money backed by extremely long-term bonds.  The trick is making sure each one is only spent once (a fatal issue with Bitcoin, incidentally) – so the trick is distributed verification.  This requires multiple “handshakes”, or queries to neighboring star systems to check the distributed registry of ownership.  It requires multiple light-speed roundtrips, and so actually spending this money can take decades – hence “slow money”.

Money is a problem that humanity really has yet to crack.  It’s one of our most profound inventions, but it’s pretty unique that we invented it long before we had any idea how it works.  Even back in ancient times, we had come around to the idea that money is a symbol rather than something inherently valuable.  But we haven’t gotten much beyond that, and as Smith points out we haven’t yet figured out what money symbolizes – is it a medium of exchange, a store of value, both, or neither?  People now seem to think it’s both, but not always at the same time, and it’s not always particularly clear when it should be one or not the other.  And even beyond that, the question is deeply tied into political economy – who should control money?  We can now be fairly confident the answer is not “the Spanish treasure fleet“, but beyond that the answers are often unclear.  The Fed is hardly perfect.

The rewards for figuring out money are vast – potentially an end to recessions and inflation.  And more than most technological questions, it seems seductively possible to just sit down, think hard, and figure it out.  It is pretty natural that science fiction writers are obsessed with money.

Keeping The Internet’s History Alive

Dr. Greg Brannon is running for Senate in North Carolina.  Dr. Brannon has some unorthodox beliefs. Those beliefs include some unusual opinions about flouride and brain-implanted microchips.  Dr. Brannon used to expound those beliefs on his website.  Dr Brannon no longer wishes these beliefs to be public. Now, he has the much more reasonable belief that his other beliefs might be a hindrance in a Republican Senatorial primary.  So Dr. Brannon has a problem and he would like FoundersTruth to go away – the website is down, but caches are forever.  So Dr. Brannon requested that the Internet Archive (a private nonprofit) take down the cached copy of his site.  The Internet Archive has, apparently, complied.

There is a serious and unresolved policy question here – as more and more keeping of “public” records devolves to private firms, what is the public interest here?  It seems that keeping Internet history both stored and generally available is a matter of public concern, yet right now this isn’t done.  I understand the Library of Congress does some of this, but not in a easily-searchable desktop version or anything like that.  And the question is even more pressing as the internet is increasingly accessed through apps and other closed services.  Twitter is mostly on the public internet, Facebook somewhat less so – but in either case, the information’s accessibility and retention is dictated entirely by private companies.

It would be nice to see Internet archival and accessibility treated as a matter for public concern and thus public funding.  Surely we all need to know about Dr. Brannon and his bold ideas. However, it seems more likely that information accessibility will either go unaddressed or be a topic for heavy-handed government regulation of internet firms.  It’s kind of a shame, because the costs of this are so low compared to feeding the hungry, caring for the sick, or launching ill-advised military interventions abroad.

Wikipedia Will Never Close the Gender Gap

I think it’s wonderful that women are organizing to close Wikipedia’s gender gap:

They had gathered for the first-ever Art + Feminism Wikipedia Edit-a-thon. At Eyebeam, and satellite marathons across New York and in more than twenty cities internationally, women were researching and writing, an effort with the twofold goal of bolstering Wikipedia’s entries on women artists and of encouraging more women to edit Wikipedia.

Wikipedia famously bears one of the starkest gender gaps in contemporary culture — less than fifteen percent of its editors are women. As a result, the Internet encyclopedia is a lopsided, Axe-scented version of the world, onewhere Sex and the City has fewer citations than a single character from Grand Theft Auto. But unlike other spaces where women are underrepresented, Wikipedia doesn’t have any official gatekeeper excluding us. No one hires you to edit Wikipedia. Which means it’s the kind of thing you can’t really complain about unless you do your part.

That being said, I think the chances of ever getting to parity are extremely slim.  The article points out that women may have less leisure time than men due to greater expectations of housework, childcare, and so on.  The ideal wikipedia editor is one with endless patience, thorough diligence, obsessive attention to detail, and a tendency towards hyper-verbose pedantry.  This could apply equally well to men or women, but it applies very well to individuals with high-functioning autism and/or Asperger syndrome.  Both conditions are far more common among men than women.

A small number of Wikipedia editors are responsible for the vast majority of the content and editing. I would guess this high-profile population contains a very high proportion of individuals with autism-spectrum disorders, much like the ranks of the very top software programmers.  Very few neurotypical people have the drive necessary to become one of the top editors, because it requires literally obsessive behavior.  Organizing efforts to improve the representation of women on Wikipedia is a great step, but this just isn’t likely to move the needle.  Wikipedia editing is a pastime tailor-made for high-functioning people on the autism spectrum, and those people are overwhelmingly male.

The NSA and the Electronic Umbrella

No one paying attention should be shocked that the NSA has placed “backdoors” in the firmware of hard drives and networking equipment. For those with less technical savvy, firmware is the low-level software that mechanically operates equipment: spin this fast, move the needle this way, and so on. It does point emphasize something the US government has long known: a government must maintain total control over its own supply chain for secure hardware. From the circuit boards all the way up to the high-level operating systems, everything must be totally secure from foreign intrusion. For the US and China, this is all well and good, and has been practiced for decades.

But what to do if you’re Canada or Taiwan? Both are active internationally, with modern militaries and hopes for regional influence. Neither has the infrastructure to affordably be maintaining an entire supply chain for secure hardware. That’s really difficult. Canada doesn’t have huge microchip fabs; Taiwan does, but they were founded by foreigners. Neither has the resources of an NSA to develop completely secure software for government use. The US and China have a hard time keeping secrets from each other; the Canadians and Taiwanese, and for that matter the Dutch and Vietnamese and Koreans should have approximately zero confidence in their secrecy. Only a few powers have the intellectual and infrastructural capacity for mostly-secure computing: the US, China, Japan, Russia and possibly the Europeans.

A hypothesis: the nature of electronic surveillance is a force tending towards tight power blocs. If you are Canada, you could try to start up and maintain a secure hardware environment no matter the cost and risks. But if you could guarantee your safety by using the NSA-approved gear…well, it might better to know the Americans are listening than merely suspecting the Chinese. If you’re a poor country relying on Chinese development aid, the choice is even clearer – take whatever the Chinese are giving you with open hands. You may have to toe the Party line, but at least the CIA won’t know about your plan to invade North Trashcanistan.

The electronic umbrella is similar to the Cold War’s nuclear umbrella, but more interesting. Nuclear weapons usage is drastic, infrequent, and incontrovertible. Electronic surveillance is commonplace, continuous, and deniable. In particular, targets tend not to know when they’ve been hacked. A known quantity of surveillance from friends might be much better than an unknown quantity from an enemy. As the implications sink in for the second-tier powers of the world, the borders of hardware and OS could end up just as clearly drawn as that of economic systems a generation ago. – The Zero Percent Solution

Silicon Valley has correctly heaped scorn on the implementation of  Part of it is the general anti-government bias of Valley types, but it is basically correct in this case!   Kevin Roose hits the mean streets of the Peninsula and rounds up some choice quotes, and Eric Ries reads my mind:

“You could take any engineer on the street here and ask them, ‘I have a friend who works for a private company — don’t mention the government — who’s thinking about a five-year, $100 million Oracle installation, and they’ve hired an outsourced contractor to build it for them. It’s going to be proprietary, hosted in their own data center, Oracle-based, with waterfall management. What are the odds that it’s working on Day One?’ And everyone here will tell you: zero percent.”

He’s absolutely devastatingly correct, and this isn’t even the sum of what’s wrong with the implementation.  On top of the factors Ries mentions, it’s got some other wonderful aspects.  It’s not an outsourced contractor, it’s many.  There’s no one at our hypothetical company with the required project management experience, but they don’t have the budget to hire a lead systems integrator.  Oh, and because of the states declining to run their own exchanges, they’ve had a massive scope change without any more time or budget.  And and top of all of this you have several actively hostile stakeholders working to undermine the project.

In any software implementation, when you have a big scope or requirements change, that is very bad.  You have three choices – you can bring on more people, you can take more time, and it will break.  There was no budget to bring on more people, and besides that there are diminishing returns to bringing on more people late in the game – with the code base basically already built, it will take a long time to get new people up to speed.  They couldn’t take more time, because they were pledged to launch on October 1st.  So they went with the third option, which was launching a broken piece of software.

I wrote about this a few days ago and return to it because it’s just so incredibly frustrating.  The government wouldn’t be having these problems if it could actually pay a decent wage to get people with technical talent working there.  They don’t even have to compete with Facebook or Google – there are plenty of older programmers with family who would love the stability of government work if it paid enough for them to live in the insanely expensive DC area.  The reflexive hatred of many in government – Democrats and Republicans both – to paying employees a fair wage is hollowing out the government’s ability to fulfill basic tasks.

This is nuts.  People in both parties love to talk about running the government like a business, but this would be no way to run a business and it’s no way to run a country.

Bitcoin Breaks Breaking Bad

Let me start by saying that Bitcoin is silly, and I think it’s more of a speculative asset than a currency due to built-in deflationary pressure.  Okay, with that out of the way…

Andrea Peterson has a post up at Washington Post’s “The Switch”, arguing that Bitcoin isn’t a big deal because it just replicates the anonymity of cash.   Andrea Peterson is completely wrong here – Bitcoin and other digital currencies are a big deal precisely because they just replicate the anonymity of cash.  She notes that it is marginally more convenient than cash because transfers are speedy and reliable.  Economists call these inconveniences of cash “transaction costs”.  The potential for digital currencies to be a very huge deal is because they are anonymous and hard-to-trace, like cash, but with none of the huge transaction costs.  To see why this matters, let’s step into the world of Breaking Bad.

Much of the entire dramatic tension of the show revolved around the transaction costs of cash.  Walter White made meth and sold it in very large quantities for simply staggering sums of money – that he couldn’t use.  Cash is worthless unless it can be effectively laundered – transacting in large amounts of cash is an excellent way to get the FBI to pay you a visit.  So while he can launder enough money through his car wash to pay his bills, it piles up quickly – one needs an enormous, mostly-cash business in order to launder the millions of money he is hauling in.  Like, for example, a fried-chicken chain.  In real life, casinos are a favored venue.  So Walter ends up with a gigantic pile of money, which he can’t spend, use or transfer.  The end of the series is driven by his increasingly desperate drive to figure out a way to get this money to his children after he’s gone.

It’s easy to imagine how his story would be different with a cashlike digital currency.  Simply convert the cash into Bitcoin (or what have you) and it’s laundered right there.  It can then be stashed away in his children’s name anonymously, and when they come into their inheritance then they can spend it anonymously or figure out some way to launder it further into clean American digital dollars.  Most of the drama of the last season wouldn’t have needed to happen!

The fact is that cashlike digital currencies would be a huge boon to criminals and tax cheats and a huge headache for governments across the world.  On the upside, it’d also be a gigantic pain in the butt to authoritarian states everywhere.  But it’s precisely that “cashlike” nature that would make them immensely important – not a reason why they should be shrugged aside.

Don’t Worry About Global Warming

I had a discussion last night with a friend who asked what was the most unreasonable thing I believe (thanks Tyler Cowen), and the above was my answer.  It’s not that I don’t believe in anthropogenic climate change; I absolutely do.  It’s not that I don’t care; I would describe myself as a raging environmentalist.  It’s more a combination of extreme pessimism and somewhat unreasonable optimism.

First, the pessimism: it is unlikely that any regulatory solution is possible while burning fossil fuels is the world’s main source of power.  Any solution would require full cooperation not only between a large number of major actors (sovereigns), but their cooperation in monitoring the billions of smaller actors (individuals, businesses) that would have huge incentives to violate the laws.  These burdens will fall mainly on rich and powerful actors for the benefit of the poorest and least powerful actors.  It’s hard enough to get the US to pony up a few billion for the Eastern Seaboard after Hurricane Sandy; you really expect us to transform our way of life for the sake of the citizens of Lagos and Mumbai?

Get real.  I don’t like it, but that’s how it is.

Secondly, the reason for optimism is that alternative energy technology is advancing much faster than we think is possible.  Distributed solar is particularly exciting – half of all electricity generated is lost in transmission, so there are huge returns to technology that allows onsite production without suffering from diseconomies of scale.  If solar continues to improve, it’s easy to imagine it becoming ubiquitous much faster than we think is possible today.

For example: if you consider a large enterprise with huge power needs, you don’t even need solar to be cheaper – its predictability is a virtue all its own.  If you install solar panels, you know precisely what your power costs will be in 20 years.  On the other hand, try to imagine estimating the fully-burdened cost of coal-based power in 20 years.  It requires you to make sophisticated predictions about technology, relative factor abundance, and a whole bevy of regulatory concerns like potential carbon taxes.  If you’re confident and get it wrong, you go out of business.  If you’re not confident, you pay exorbitant fees to Wall Street to hedge risks.  Well jeez, wouldn’t you love being free of that?

Finally, if worst comes to worst, we’ll very likely find a way to deal with it.  There are already  a huge variety of potential solutions, with varying risks of destroying the planet.  I firmly believe that if warming gets bad enough, someone will undertake one of these approaches to either sequester large amounts of greenhouse gases, directly cool the planet, or both.  This approach may entail unimaginable ecological costs, or it may work out better than expected, or somewhere in between.  But in the end, global warming is a physical and chemical problem that must be tractable to engineering solutions.  Solutions will have unpredictable and potentially severe consequences,  so it’d be better if we never ended up having to do so, but it may come to that.

The long and short of it is that the political economy and technological situation suggest strongly that we’ll muddle through it somehow.  It’d be nice if we could pass a solution to encourage clean energy development and adoption without harming the economy too much.  Cap and trade or a carbon tax would be great.  But don’t lose sleep over it.

Watching the Watchers

Visiting New York this week, where Stop and Frisk is in the news once more.  I’ve never been able to get that into it, since the matter seems pretty straightforward: giving police expanded authority to harass poor minorities leads to police harassment of poor minorities.  Today, a judge agreed with me and ordered a stop to the program as designed, as well as federal oversight to allow the reformation of the program into something that doesn’t systematically violate the rights of everyone it touches.  However, there’s one small toss-off detail that I think is important: the judge also ordered cops in five precincts to begin wearing cameras to record all of their encounters.  This is part of a pilot program which may extend to the entire city.

This is a wonderful idea.  The police, especially the NYPD, are often accused and often guilty of persecuting minorities and generally acting with wanton disregard to the mere concept of civil rights.   Psychological experiments like the Stanford Prison Experiment or the Milgram Experiment show how normal people put in positions of power often act in downright psycopathic fashion.  Not to mention that the police force can draw many people who aren’t the healthiest; the Onion recently put it best with its article on an “Insecure, Frustrated Bully with Something to Prove Considering Career in Law Enforcement”.  Often police officers feel free to engage in this sort of behavior since there are only two witnesses, an officer and suspect, and they can be confident that their story will be believed.  By making all police-suspect interactions available for later scrutiny, hopefully cameras will not only act as ex post evidence of misconduct, but a powerful prior restraint on misconduct.

I confidently predict that reports of police misconduct in the pilot areas will decrease sharply.  I also wonder if savings on liability insurance and misconduct settlements will cover the cost of the cameras.

It goes to show that the technological Panopticon can actually work for good.  The question of “quis custodiet ipsos custodes”, or “Who watches the watchmen?” is an old one going back to the Roman poet Juvenal, and reflects that misconduct is likely to spread amongst those with the responsibility to police others.  People have sought to fix this with institutional oversight a hundred different ways in a hundred different places, but nothing has been that effective. However, the availability of cheap cameras and unlimited storage space can actually allow for effective oversight of law enforcement.  See, technology is good for something other than spying on all Americans all the time – spying on some Americans some of the time!