When is intrinsic motivation bad?

Intrinsic motivation – the desire to do a thing for its own sake – is often regarded as the ideal motivation in the context of designing incentive structures. But there are some cases when a reliance on intrinsic motivation leads to adverse selection.

Consider policing. How many cops are sociopaths? The rate in the general population is about 4%. If police were representative a naïve estimate might be that 4% police are sociopaths. But police are not drawn at random from the population – people choose to be police. This has interesting implications.

Suppose police operate under “low-powered” incentives, thus intrinsic motivation is a larger component of why people choose to become police. This could be desirable if the effort exerted by police is difficult to monitor, thinking along the lines of “we trust police to try and catch bad guys even when they aren’t monitored, because police actually want to catch bad guys”. And we might believe that they want that because if they were “in it for the money” i.e. not (at all, there are degrees) intrinsically motivated they wouldn’t have chosen a career in law enforcement in the first place.

The problem is that low-powered incentives can select for intrinsic motivation, but not for particular kinds of intrinsic motivations. You might want to become a cop because you want to serve your community and keep people safe – or you might want a cop because you want to be able to beat, harass, and bully people with impunity. That is to say, you’re a sociopath sitting pretty in the middle of the “dark triad” of narcissism, machiavellianism, and sadism and despite low-powered incentives would so enjoy abusing others that you do the work anyway.

What can you do about this adverse selection problem? If you don’t want to throw the baby out with the bathwater (so to speak) and keep selecting for (public-spirited) intrinsically motivated cops by using low-powered incentives then you need effective and rigorous screening and accountability measures to weed out those who want to be cops for the wrong reasons. If you can’t or won’t do that, then you need to consider whether it would be worth it (or even possible) to switch to “high-powered” incentives and do away with intrinsic motivation altogether, good or bad. In the American context, with an extremely fractured patchwork of law enforcement agencies that vary wildly in professionalism and accountability, this is particularly a problem – a “bad cop” can just get a job at another department. One department could be doing everything right, and another 20 minutes away could be doing everything wrong – and then cops sort into departments that match their behaviors.

Should we rely on the intrinsic motivation of police? To make the judgement, you have to ask yourselves “Do I think there are more public-spirited people who want to become cops than sociopaths?” or “Do I think there are more public-spirited cops than sociopathic cops?”

I can’t answer those questions. But I think the answer is probably that the proportion of cops who are sociopaths is greater than x. Police unions fight tooth-and-nail against every measure of accountability, no matter how small – and I have little doubt that plans to consolidate police departments into larger, more professional organizations would provoke howls of protest. But more fundamentally, I just think that the temptations of the power given to American police – in practice, the power to deprive another of life, liberty, and property with near impunity – is so much more tempting to the sociopath than the public-spirited person’s desire to use a badge and a gun to actually help people. I guess I just think there are a lot more shitty people out there than really good ones, so if you set up an incentive structure that mostly attracts shitty people or good people, you’re going to end up with a lot more shitty people than good.

There’s a lot of jobs like this consider the case of prison guards – we want prison guards to only be in it for the money. People who enjoy being jailers probably shouldn’t be.

Just some food for thought.

The Coming Automation Crisis

The coming automation crisis is not a technological crisis. Rather, it is a crisis of governance in the most basic sense. A fairly recent report from the Mckinsey Global Institute makes this obvious - though they studiously avoid drawing attention to this alarming conclusion.

Overall, MGI anticipates tens of millions of workers will be displaced by automation over the next 10-15 years, depending on the speed of automation. At the high end of their estimates, 73 million US workers (a little less than half of workforce!) would be displaced. Most of those displaced would have to switch occupations entirely. Their middle-of-the-road projection is still dramatic: only quarter of the workforce displaced, and with half of those switching occupation entirely.

The MGI report has plenty of suggestions for policy responses, but the US has essentially no public or private sector apparatus up to the challenge of smoothing this transition. The American government spends the least of any OECD country on labor markets. That money is used for job training and placement, for hiring subsidies and hiring in the public sector, and for unemployment benefits. Of the little money that is spent, only a tenth of that goes to job training – less than half of what Canada spends, and less than a sixth of what Germany spends, and a tenth of the amount France spends. The only tools that that remain are monetary policy and educational policy. 

Using monetary policy to maintain full employment is problematic for practical reasons and political reasons. The political problem is that central bankers tend to be more concerned with inflation than unemployment - witness the recent preemptive interest rate hikes by the US Fed. God forbid we should wait for robust wage growth before panicking about an inflationary spiral!

Even if by some minor miracle the Fed changed course and pursued a policy of full employment at all costs, the for-profit sector in the United States is not up to dealing with this crisis. MGI highlights a few firms who invest in employee development, but these anecdotes are misleading. Overall the share of workers receiving on-the-job or employer-sponsored training has declined sharply since 1996, and stagnated since 2004. You will note that this period includes part of the 1990s boom, and 2001-2007 boom. This is entirely rational behavior on the part of firms, since a worker who receives valuable training could find work elsewhere. And since firms show no loyalty to their employees, it would be sheer madness to expect workers to show loyalty to their employers,

This leaves only educational institutions and workers themselves. And educational institutions are not ready for this either. For people entering the labor force, the future looks bleak. Only about 36% of Americans age 25-29 have 4-year college degrees or more. For the majority American's associate degrees, high-school degrees, and the nebulous term "some college" are the highest level of educational attainment. For such people, quality high-school vocational programs are invaluable. But MGI’s best guess, since US data are hard to come by, is that less than 6% of Americans in secondary schools are enrolled in a vocational program, which would give those with only a high-school degree a fair chance That’s a smaller share enrolled than Brazil(!), and a tenth(!) of the Swiss level of enrollment. As far as post-secondary education goes, the federal state governments have systematically pulled back on support for public higher education for decades, exacerbating spiraling tuition costs and miring a generation of graduates in debt. What about community colleges, where displaced middle- and low-skill workers will most likely go for retraining? Data here are hard to come by, for the same reasons that data on high schools are hard to come by, but the ability of community colleges to handle the crisis does not look good. Tuition is not trivial, though it varies wildly across states - about $5,000 for in-state public schools. Student debt, graduation rates, and transfers to 4-year schools for community college graduates also looks terrible. About 10 million people are enrolled in community colleges in the US in any given year, with about 2.5 million full-time students. Remember we're talking tens of millions of workers displaced by automation over the period, with half that number needing to switch occupations entirely, requiring extensive retraining. I doubt our current educational system even has the capacity to cope with so many students, let alone provide a good education to all of them.

The MGI report helpfully makes two forecasts; a trend-line forecast where things continue as they are going and a step-up forecast where the government takes active measures to reduce the impact of automation on employment. Only the step-up forecast doesn't result in net job losses.

Read all that and tell me you aren't worried.

This is what I mean when I say the automation crisis is a crisis of governance. No problem described above cannot be addressed, even solved, by straightforward policy responses. Not  government that cannot act, not government that does not know how to act, but government that for structural and ideological reasons - political reasons! - will not act. This is the American Crisis today.