Tuesday, December 11, 2012
Incentive Granularity and Software Patenting
For my last post, I would like to address a comment repeatedly seen on my prior post: “Show me an invention that would not have happened for the entire patent term, and maybe then we can discuss whether the patent system does any good.”
I’m not convinced this is the right level of granularity. But first, a couple caveats:
- I tend to think the patent term is too long for the speed at which technology develops today, especially computer software. This may not be true for pharmaceuticals, which leads to tension in the system.
- Of course we should look at whether individual patents were incentivized by the patent grant. It would be a bad system indeed if we protected everything that would have occurred anyway. Note that I think the “inducement” standard proposed by Duffy & Abramowicz and discussed in my previous post has some real merit.
But even with these two caveats, that’s not the question we should be starting with. The goal of the patent system is to promote progress of the useful arts. That might happen by encouraging investment in start-ups. That might happen by encouraging research & development funding. That might happen by inventions that come earlier than they would have, even if they would have otherwise come within 20 years. That might happen by allowing inventors some breathing room to invest in commercialization and dissemination of the invention. That might even happen by ending duplicative (wasteful) races carried out in secret. And all of these things might create costs, perhaps tremendous costs for some who come later.
To be sure, there is great (and I do mean tons of) study and debate about whether any of these benefits actually materialize and outweigh the costs. The analysis, though, takes place at a higher level than whether each and every invention would have come about within 20 years. That analysis – or something like it – certainly has its place, but not when assessing the system as a whole. And that's all I have to say about that.
Thanks again to Prawfsblawg for having me back. I enjoyed my stint, in what may be the most active commenting I’ve received (which may not be a good thing!).
TrackBack URL for this entry:
Listed below are links to weblogs that reference Incentive Granularity and Software Patenting:
Michael, I think the move you make--namely "going meta"--ends up with a useless and unwieldy inquiry. First, a small caveat, which is that I don't read Abramowicz & Duffy as basing their standard on whether the invention would have existed 20 years later but only whether it would have existed as early as it did but for the patent system (whereas my version of the idea based it on a cost-benefit balancing keyed off independent invention). But if your point is that you fundamentally don't think the timing of the invention is the right metric for assessing the patent system because you think this improperly fails to consider other types of "utility," the problem becomes that this renders everything at large. The end of this slippery slope is something like Eldred where you can construct a theory that retroactively granting patents to long-dead people will promote progress--which it will, as long as you define "progress" in a sufficiently creative way. If "utility" doesn't mean what the patent system has long thought it to mean (i.e. inducement of inventions earlier than otherwise would have occurred) but instead means what an individual judge or scholar wants it to mean, then utilitarianism has no meaning and provides no theory for organizing the patent system.
Posted by: TJ | Dec 11, 2012 11:41:34 AM
TJ - I read Abramowicz & Duffy the same way. And, yes, I'm saying that if giving patents to long-dead people made things better off we should do it. Indeed, that's what we do for copyright, no? Not on the front end, but certainly on the back end. And we argue about whether that promotes the progress or not, but we don't ask in each case whether the copyright is justfied for the term. That's just too granular.
Of course, if we could come up with a patent system that built that granularity in using systematic means, then great. And so we have inducement ideas, and even requirement of use in trademark to maintain ownership (not under the IP Clause, I know). I'm not saying that you don't want the granularity or try to achieve it. I'm saying that's not necessarily how you measure it. And maybe it is too meta, which is why no one seems to come up with data that answers the question clearly in either direction.
Posted by: Michael Risch | Dec 11, 2012 11:50:16 AM
This seems like a bit of a cop-out. If the mean duration before invention in the absence of patents were even five years, across tens of thousands of patents a year, by sheer chance you'd expect to see at least one with a MDBIAP of 20 years.
It is somewhat odd that a scholar so deeply interested in the question can't point to even one thoroughly convincing software patent, where even opponents are forced to say 'yeah, but'.
Posted by: brad | Dec 11, 2012 3:29:56 PM
As an addendum I would think some of the public key cryptography patents would be a good place to look. The breakthroughs were patently non-obvious and useful (no pun intended!), but you have to face head on the issue of: is it just math.
Posted by: brad | Dec 11, 2012 5:47:02 PM
I agree on the cryptography - on both counts. That really is just math. I didn't say I couldn't point to one - only that it's not the right question to ask. I am also certain that any one I point to will be rejected by the anti-software patent community, so I am not wasting my effort. I also don't want to bog down this post with the debate.
Posted by: Michael Risch | Dec 12, 2012 3:31:40 AM
Michael, it is true that nobody explicitly asks whether any individual copyright is justified for the entire life-plus-seventy term based on an inducement theory, but that is because the answer is so plain and obvious. The conclusion that usually follows--at least among the people that I talk to (with the exception of non-utilitarian people)--is that this illustrates how the entire copyright system is fundamentally broken. I read your commentator as making the precise same argument on the patent front.
But I think you are missing my point, which is that you are pulling a move that I'll call anti-reductionism. The contribution of the inducement standard is that it provides a theory of what constitutes "progress," in the same way that the Hand Formula provides a theory of what constitutes "reasonableness." What you are doing is saying that such a reductionist theory does not 100% capture all the different conceptions of the word. But then you don't provide your own theory. You just call for more study of whether the system promotes "progress."
The problem, which you kind of concede but don't seem to fully appreciate, is that there is nothing to study if you make the anti-reductionist move. The whole contribution of the inducement standard is that it provides a concrete measure of "progress" to study. If you say that inducement is not the right measure of "progress" but don't provide a comprehensive theory of what is the right measure of progress, then there is nothing to empirically study because we don't know what to measure, and any studies that are produced can always be disputed ex post by saying they used the wrong yardstick. It then strikes cynics as a cop-out when a person who does this calls for more study, because the sneaking suspicion is that the person don't really want any study and just wants to delay because they like the status quo but can't provide any reasoned justification for their preference. The ceaseless calls for empirical study become a tactic to delay and to shift the burden of proof--an impossible one--to the other side. Not accusing you of this in particular, but you should be aware of why I hate this move.
Posted by: TJ | Dec 12, 2012 1:47:05 PM
I said pick your own utilitarian criteria, and then show me just one example of a deserving patent. You've done neither.
Nobody is suggesting we assign varying patent terms to individual patents based on any (ex ante or ex post) assessment of how far the invention is ahead of its time. That would be extremely unwieldy. We're suggesting that in the vast majority of patent cases, perhaps in every single software patent case, the patent term is too long (or no patent should issue at all) under any utilitarian standard.
This is simple to test empirically: take a sample of patents, and ask a PHOSITA plus a little knowledge of the industry's history, how long until each was independently invented? In the case of software patents, the answer is usually a negative number. But even in hardware patents, it's clear that the answer is usually less than the patent term.
Now, maybe, there is some further utility to patents - like "ending duplicative (wasteful) races carried out in secret". (Applying that particular idea to software patents is laughable, but never mind.) Why do you seem uninterested in whether such benefits actually outweigh the social costs of granting a twenty-year monopoly? Never mind the social costs of patent litigation?
The reason I said "show me just one" deserving software patent is that I think you will have a hard time finding any. If you can agree with that, then you will have conceded that the vast majority of software patents are a net drain on society and should be abolished.
If the benefits that you think might flow from software patents actually exist, you should be able to find just one example where such things occur.
Posted by: nartreb | Dec 12, 2012 11:54:04 PM
nartreb - like I said, I think there are some, and probably many. They are the ones that don't make the news. But I also note that I'm skeptical I'll get agreement, so I'm not spending the time.
TJ - I think there are plenty of empirical studies (not necessarily for software), and people debate what they mean. We could always use more. The problem with the reductionist argument is that for every successful patent there are many unsuccessful ones. The same is true for copyright. But the underlying activity for ALL of them is desired. This is why you can't measure the system on a patent by patent basis.
I find it interesting that when I say many or even most patents are bad, and we should weed them out, you say that I am being too granular, and we should just categorize the ones that don't work and cut them out of the system. But then when I say we should measure the system by looking at the entire sum of activities, you say, no: let's consider on a patent by patent basis. It's obviously a difference between administrative costs v. incentivization. I favor broad incentivization where we will incur the administrative costs anyway. You seem to think we can reduce administration without affecting incentives too much - but I don't think so.
Posted by: Michael Risch | Dec 13, 2012 1:23:23 AM
You "think there are plenty of empirical studies"?? You keep saying that, yet you fail to cite any.
"(Not necessarily for software)" Oh? That's a significant admission. See how much progress we're making since your previous post on prawfsblawg?
"when I say many or most patents are bad" - hm, that's interesting. Obviously I'm not following everything you write, but I sure haven't noticed you saying that before. Do you actually believe that??
If I understand your last couple of sentences (and I may not), you consider litigation an "administrative cost". That's weird.
In any event, you have completely missed TJ's point. It's all well and good to say you'll "measure the system by looking at the entire sum of activities", but you still actually have to *look*. All you're doing is waving your arms and asserting that because some benefits might exist, then they do exist. (Your copyright example is telling. Are you *seriously* citing the Sonny Bono Copyright Extension Act as an example of good utilitarian policy???) If these benefits exist, you should be able to find a few examples.
"I'm skeptical I'll get agreement" is a cop-out. Either you can find a decent example or you can't.
Posted by: nartreb | Dec 13, 2012 2:01:52 AM
This was a blog post, and a short one at that. It was not intended to solve the patent problems of the world. Yes there are empirical studies. Petra Moser discusses many of them in a recent essay: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2180847 - as you can see, they are all over the map, but most find a link between patenting and innovation. How much and the mechanism is hotly debated.
As for examples, I can, and I did. I pointed to the Swype patent. The responses were drawings on cave walls, gestures, the fact that shape matching was known, and Swype is a joke. What I didn't get was a single response with hard evidence that a) anyone else was working on it, b) anyone else (even today) is working on alternate ways to do it that is effective (there are swype alternatives, but they stink), or c) that the literature at the time suggested that querty swiping might be implemented in the manner claimed. The patent may well be obvious, and I'm willing to concede that, but, I am just not willing to continue the exercise based on naked assertions. Everyone thinks everything was obvious.
Posted by: Michael Risch | Dec 13, 2012 6:04:12 AM
You never gave an example of a patent you actually thought was valuable. You pointed to Swype as an example of something that ought to be patentable, *assuming* it was novel and non-obvious. Your position on this hasn't changed: "the patent may well be obvious".
Isn't there *any* patent that you can actually point to and say, "this was not obvious at the time it was filed"? If you can't do that, don't you think it's a waste of time to be talking about "granularity"?
Posted by: nartreb | Dec 13, 2012 10:04:02 AM
I guess I was unclear. I don't think Swype was obvious when filed. I'm willing to be proven wrong (which is what I meant by "may well be"). I was just being deferential because I don't want to argue about it anymore.
Here's some debate about it, by the way. http://news.ycombinator.com/item?id=4093264
Posted by: Michael Risch | Dec 13, 2012 10:15:10 AM
The swype patent is actually an interesting one. It's outside my field of expertise, so I'll need to do some research on novelty/obviousness, but it seems clever. There's something counter-intuitive about the idea of swiping around on an interface that looks like a keyboard. I've heard of a few other swiping interfaces, but their interfaces do not resemble QWERTY keyboards.
It's also a patent where the idea is likely to be the hardest part. The implementation looks to be mostly straightforward: some parameter-tuning, but no real difficulty. For once, I don't much care whether the patent gives details of the code.
Of course, I don't care about seeing the code for an online shopping cart, either. Screening for obviousness is important to do, and the PTO is failing miserably (nor is there any reason to think this will change). You won't be surprised to read that the existence of one (possibly) non-obvious patent doesn't change my evaluation of the overall balance of harm due to software patents.
But let's assume for a moment that Swype was non-obvious, and pretend that the PTO is capable of weeding out obvious patents. The question remains, is the Swype invention deserving of a twenty-year monopoly? Who exactly does that benefit? How much R&D did Swype require?
Empirical test: find a competent software engineer. Give them a copy of the Swype video tutorial, and tell them "code this." How many man-hours will it take?
In other words, does the size of the incentive (monopoly) have any relation to the effort put forth? If not,how do you justify the existence of this patent? Swype may be a rare example of a software patent that's not plainly obvious to me, but it still makes a good case study in the costs of our patent system.
This comment is getting too long. I'll take up the Swype patent in detail elsewhere (though probably not until after a vacation). Meanwhile, your turn. Using Swype as an example, can you point to any of your expected benefits, eg "ending duplicative (wasteful) races carried out in secret" from this patent?
Posted by: nartreb | Dec 13, 2012 12:24:27 PM
Yes, I hear you. I also said that terms are too long, and at my Santa Clara presentation one of my key points is that with software we may have to rethink something that we've generally accepted in other fields - that we might reward the good idea that is easy to implement, or even the accidental idea. And this consideration depends on a lot of factors (I argue over the clear disagreement of some) that are not at the granular level. Maybe the ideas just come too easily and the software is made too easily, so we just can't have such a system.
The question is whether we want the system to incentivize the ideas earlier, whether patents generate investment to pay for that time given that so many ideas fall flat, etc. There's debate about whether any of this bears out in the end (and it's not looking strong for software), but my point - indeed the ONLY point I was trying to make - is that you can't measure that on a patent by patent basis. Look at the pharma industry - only tiny fraction or R&D becomes a profitable, FDA approved drug. The patent system has to support returns on that ONE DRUG PATENT (or few, really) that are sufficient to pay for the entire rest of the system. I'm not saying that software should be like that - it isn't and it shouldn't. I'm just saying that the picture is bigger.
Posted by: Michael Risch | Dec 13, 2012 12:35:24 PM
OK, I'm flying away tonight, so no more comments from me for a while. But I wanted to correct something I said above.
The swype *software* is interesting and possibly novel. The swype *patent* is a different beast altogether. There are 51 claims; the only mention of a QWERTY keyboard is in claim 17. At a quick glance, the vast majority of claims seem to read on a vast ocean of previously existing input schemes.
Even if claim 17 and perhaps a few others ought to survive, a patent like this is an abomination.
Posted by: nartreb | Dec 15, 2012 7:00:51 PM
No, I am happy to do make the assessment on a systematic basis as well as a patent-by-patent basis. But the question is what the assessment comprises. Your argument isn't really about granularity versus systematic but about what constitutes "progress." That is, you dispute that showing THE ENTIRE SYSTEM fails to induce a single invention would establish that it fails to promote progress. And that is what I am criticizing you for here.
Posted by: TJ | Dec 15, 2012 8:48:28 PM
Or, in other words, the commentator you criticize was totally making a systematic argument, not a patent-by-patent assessment. He was arguing that the system as a whole was broken because it produces no benefits, and it produces no benefits because it induces no inventions. Your response was not--and cannot be--that he was considering things patent-by-patent. Your response was that whether the patent system as a whole induces any inventions is not the right yardstick to measure the performance of the patent system. That response has nothing to do with granularity and everything to do with what constitutes the right measure of the performance of the patent system. You've made clear that you don't think "whether the system induces inventions" is it; but you have simply waved your hands on what you would replace it with. And it is a cop-out, because if you don't specify a measure, you can dismiss every empirical study you dislike by saying it uses a different yardstick compared to the secret yardstick you have in your head.
I hate to make a I-know-your-argument-better-than-you-do argument, but in this case it is true.
Posted by: TJ | Dec 15, 2012 9:38:49 PM
TJ - I just don't think so. I'm familiar with the "It takes a theory to beat a theory" point of view, but I'm not a believer. After all, you can have a crummy explanation for something even if you don't know what the right explanation is.
Even so, you act as if I've made up this blog post out of whole cloth, as if I was noodling one day and came up with these theories on my own with hand waving. Would that it were true! There's been lot's of theoretical literature on all of the points I made, and to the extent that the assertions of those theories can be tested, we can examine empirical work. Does this mean I am skeptical of empirical studies that seem to show there is no value to the system? Of course. Does it also mean that I am skeptical of empirical studies that show there is a lot of value to the system? Well, that's true, too. It's not a matter of discarding whatever I want - it's a matter of critically looking at the evidence and seeing where it fits and what questions it answers.
As for the systematic part - yes, if the theory if progress is that a patent system must induce inventions that would not have come in 20 years, that was a systematic request. But I still maintain it is not the right question we should be asking. As I note in point 2 of the main post - of course we want inducement of inventions. I'm not saying we don't. I'm saying that no matter what patent I pick, I will hear - as I did from nartreb - that it is an abomination. You are right that this is a systematic argument - that no patent is "worth it." But to me it is still a granularity issue - I'm quite willing to sift through the gobs and gobs of bad patents out there while looking at a bigger picture for the system. I realize many believe that this number approaches 100%, but I'm willing to consider a system where everyone invests and applies for a patent and almost no one gets a patent. That might actually have the incentive effect we want, and you can't get there if you look patent by patent.
Posted by: Michael Risch | Dec 17, 2012 2:55:31 PM
I made that last comment unnecessarily personal. I don't mean that YOU in particular are opportunistically picking through empirical studies. I mean that your methodology allows that option, and that is problematic. Because one cannot "critically look[ ] at the evidence and see[ ] . . . what questions it answers" without knowing first what the questions are. Knowing that the answer to life, the universe and everything is forty-two gets us nowhere.
As for systematic versus granular, I get that you are willing to sift through gobs and gobs of patents to find the few that are worth it (and I even agree that there are a few that are "worth it," though I judge worth it from an inducement perspective). I was responding only to your accusation that I have been shifting gears on you, in making an argument based on systematic evaluation one day and individualized evaluation the next. My response was that nothing about my argument, or the commentator's argument, contradicted my prior criticism of your approach on the basis that I view finding needles in haystacks as an efficient way to administer the patent system. I have been making systematic arguments all the way through.
Posted by: TJ | Dec 17, 2012 5:01:44 PM
Typo: obviously I meant "I view finding needles in haystacks as an INefficient way to administer the patent system."
Posted by: TJ | Dec 17, 2012 5:03:21 PM