I recently did a diavlog with my friend Tim Lee on the new BloggingHeads spinoff site TechHeads, during which I had a thought that seems like it might be worth spinning out. We’re all accustomed to seeing horror stories about ludicrously broad, bad technology patents that have given rise to a wasteful arms race between real tech companies and patent trolls. A growing body of scholarly literature suggests that two decades of software patents, in particular, have been a hindrance rather than a net plus for innovation, and I think it’s worth thinking a bit about why that’s especially likely to be the case.
It has probably always been the case that our intellectual property policy has been skewed by a romantic vision of the lone genius in his workshop crying eureka! as some radical breakthrough strikes—when real innovation is more a matter of gradual, cumulative evolution. But I suspect it’s also true that one kind of innovation is more likely when you have a small number of pioneers working in a new field. A Newton or an Einstein born today wouldn’t be as likely to make similarly earthshattering contributions, partly for the simple reason that Newton and Einstein already did it (you can’t write the Principia twice), and partly because with thousands of people making small, incremental additions to our knowledge pool, it’s harder for even a genius to outpace them in a single bound. But we still want our genius—an identifiable individual to credit with the latest leap forward.
The thing is, as technologies mature, and innovation is more likely to proceed by a series of increments that shouldn’t be individually patentable, it also becomes much harder to police the quality of patents, which makes it more likely bad patents will slip through. To see why, consider the old economist’s aphorism that “division of labor is limited by the extent of the market.” That is, the more people are connected in an extended system of economic cooperation, the more specialization becomes possible. Robinson Crusoe has to be a generalist—has to know how to produce everything he’s going to consume. In a small isolated village, most people need to be food producers first and foremost, and a few others can focus on providing other necessities. In a globally networked marketplace, an incredible amount of specialization is possible—which is why, notoriously, there’s almost no type of porn so bizarre or niche that someone hasn’t created a website dedicated to providing it.
When my parents were children, “computer programmer” was not even an occupational category. Now we have thousands upon thousands of people who specialize in working on very specific types of software: Web applications or encryption or databases or first-person-shooter games. Crucially, an innovation developed by one person can be circulated to a whole community almost instantaneously—which makes it less likely that we need a “genius” to make huge innovative leaps because small components of his big idea, if they do already exist, are scattered across obscure journals in other languages where he’ll never encounter them. Predictably, this enables an enormous amount of very rapid progress.
It also makes it a lot harder to evaluate patent quality. Patents are not supposed to just be a pointless monopoly granted to the first person who happens to file a description of a particular invention with the Patent Office; the justification for granting the monopoly is that it (in theory) elicits innovations that would not exist but for the incentive a patent provides. To that end, patents are only supposed to be granted for inventions that are “non-obvious.” But of course, “obvious” is a relative term: You wouldn’t want to give an industrial engineer a monopoly on some minor improvement to a production process just because it might not be obvious to me. The standard, rather, is whether it would be obvious to a person “skilled in the relevant art” applying reasonable diligence and effort.
The more highly specialized professionals are in rapid communication with each other, the more likely it becomes that you’ll see innovations that are “obvious” because they involve combining various disparate kinds of incremental prior innovative steps, but which don’t have “prior art”—meaning nobody has taken that exact step before, because it required a bunch of other pieces to be in place before it was viable. So searching for “prior art”—if that means exactly the same preexisting invention—becomes a less reliable guide to what is “obvious” in the relevant sense. But as specialization increases, it also becomes vastly more difficult for a patent examiner with broadly relevant training (engineering and electronics, say) to use his own understanding and expertise as a guide to what is truly “obvious” to someone trained in the specifically relevant domain (say, engineering mobile cellular data networks). It’s increasingly unreasonable to expect even the smartest and most diligent examiner—even assuming away all the bureaucratic and institutional incentives to err on the side of granting patents—to judge the “obviousness” of innovations across an ever-proliferating array of subspecialties.
As specialization increases the number of eyes on each particular type of problem, it also shifts the degree of non-obviousness that makes the granting of a patent monopoly a wise bargain for society. As I suggest above, when a small number of people are working on a problem, and in only sporadic communication, then you’re more likely to need someone to make a big leap to hit on an innovation rather than relying on the accumulation of many small incremental insights. (Perversely, patents may create an incentive to keep quiet about those incremental insights along the road to a functioning, patentable idea.) But sheer numbers also make it vastly more likely that someone—and more likely many someones independently—will arrive at even less-obvious ideas. If only 1 in 100 people working on a type of problem would hit on a specific solution, it seems a stretch to call it “obvious.” But if there are 10,000 people working on the same class of problems, that’s still 100 independent inventors.
That very much changes the shape of the problem patents are meant to solve. If you’re in a market with a dozen other players, then in the absence of a patent monopoly, your inventive might be to take measures to keep that one-in-a-hundred innovation secret so your competitors can’t copy it. (Send me your raw materials and I’ll apply my new improved production process behind closed doors, rather than making a machine to sell.) The public loses the benefit of learning how the invention works because secrecy is an effective alternative means of maintaining a monopoly. If there are a thousand other players in the same market, the perverse secrecy incentive drops away, because you can be reasonably sure lots of other people are going to hit on the same invention, and it’s likely that your costly efforts to maintain secrecy won’t actually leave you with a monopoly, or even ultimately prevent the knowledge of how the invention works from getting out.
More specialization, in short, means that society can get the benefit of even less-obvious innovations without having to pay the cost of an inefficient monopoly grant. Needless to say, the greater the efficient standard of “non-obviousness”—the larger the logical leap it makes sense to require from “prior art” to a new innovation before that step becomes patentable—the harder it is for a non-specialist examiner to correctly apply that standard.
So there’s an irony here: As a sector becomes more innovative, as even “good” patents become less necessary and more likely to hinder rather than spur innovation on net, and as the economically efficient “non-obviousness” standard for patent issuance rises, patent examiners necessarily become less likely to be able to discriminate effectively between good and bad patents. In other words, the circumstances under which it is efficient to be granting fewer patents are the very circumstances under which patent examiners lose the ability to accurately judge “obviousness,” and therefore become more likely to grant more patents. No wonder it’s such a mess.