Andart

Isn't it Obvious?

New Scientist: Europe revokes controversial gene patent

The European Patent office has overturned Myriad Genetic's patent on the nreast cancer genes BRCA1 and 2. This enables European breast cancer tests that do not pay licence to Myriad, and mirrors the legal troubles the company has with Canada.

Also this week, the European Council has approved software patents, something that has been fiercely opposed by many software developers.

(This is a rambling railyard of thoughts; your mileage may vary on their philosophical, legal and intellectual quality)

What is at stake here? If we ignore paranoia about companies forcing us to licence our metabolism, general arguments about the unpatentability of life or code, the core problem seems to be obviousness.

The patent system is intended to ensure that creators are rewarded for their work and publish it, in turn leading to greater creativity and competition. However, today patents today can often be used to block competitors rather than turn profit through their own value. This is especially true for very broad patents or patents on obvious, efficient ways of achieving something.

Detecting breast cancer by looking at a marker gene is obvious once we become awareness that such a gene exists. There are many examples of similarly obvious software patents, such as using XOR to draw cursors (I invented the XOR method independently as a kid; I even used it in a Sinclair ZX-81 drawing program I sold to a friend: had I lived in the US at the time I would have been liable for patent infringement). The problem with such basic and obvious patents is that if they are enforced they either require programmers to use non-obvious or inefficient methods, or to pay licencing fees. In both cases the overhead of programming increases enormously. Inefficient, non-obvious software is the bane of maintenance and further development, and even if the licencing fee is reasonable it can increase the cost of a piece of software from zero to a finite value - what would have been a contribution to the cultural commons of humanity now becomes limited. Even worse, given the complexity and size of the patent corpus finding if software infringes requires sophisticated legal help which is even more expensive, and requires lawyers to look over the shoulders of programmers and biologists. Planning a software project or experiment becomes both a technical and legal challenge, where both sides get in the way of each other.

One can argue that software should be unpatentable, but one still has to consider how to reward creators. A situation where most commercial code is closed and deliberately obfuscated might not be the best. It would slow development and reduce maintainability, although reverse engineering would thrive.

"Info-socialists" might suggest that the government uses tax money to reward useful creations: creators register what they create, making it available to everyone and getting paid for its value. As a libertarian I have many philosophical problems with this idea, and there are also many serious practical problems. The worst problem is how to measure how much to reward a creator: there is no market pricing, usage measurements can both be manipulated and be irrelevant (an emergency evacuation system might be extremely valuable even when it is almost never used) and centrally decided rewards run into the usual information problem, corruption and biases of governments.

Shorter patents for software and other fast-moving areas might reduce some of the problems, but it only reduces them and increases the incentive to sue the hell out of everyone before the patent expires, as well as to rush patents through the overworked sieves of patent offices.

If we need patents, then we must make sure the patents only cover non-obvious solutions. And that requires a reform of the patent offices and how they handle applications. It might be just as dramatic as abolishing them in the first place, because the changes required are profound. Somehow prior art spread across the scientific literature and infosphere must be detected. Even a skilled patent engineer is unlikely to succeed. Clearly we need to change the information flow so that prior art information or whether a suggested approach is obvious reaches the office very early in the process.

One could do this by using broader panels of reviewers (anonymous relative to the party seeking the patent), perhaps borrowing from the scientific peer review process. Setting up such a panel is nontrivial since it must not be a leak while at the same time likely involving people involved in closely related areas. This might be doable through non-disclosure agreements and funding from both the patent office, the patent party and perhaps stakeholders in the public commons (governments, NGOs etc). Perhaps a good place for engineers to earn some extra money, keep an eye on what is happening in the field and do a civic-technological duty.

A clever way of making patent offices more efficient and reliable would help a lot, both by making shorter length patents feasible and removing the worst patents. But it is not obvious how to do it, or how to get the political impetus to implement any change.

A deeper and more philosophical problem is defining obvious. Given that humans have different skills, outlooks and intelligence what is blindingly clear to one is a stroke of genius to another. Legally the solution seems to have been in most jurisdictions to establish a praxis. This can create enormous confusion internationally, as they reach different conclusions. It is unlikely that this will change. But what would make a reasonable definition of "too obvious to be patented"?

Maybe one approach would be contingency. If something is unlikely to be discovered/invented/constructed the same way without any communication or other information transfer between two people, then it is not obvious. It is contingent on many different factors, and involves some choices that may be arbitrary, aesthetic or due to complex balancing. This is close to how artistic intellectual property is handled.

If contingency is the basis of patentability, then fundamental discoveries will not be as rewarded as implementations of them. I'm not certain this is a good thing, since the natural reaction would likely be to keep the basic discovery secret or unpublished and spring off a number of contingent patents from them.

Maybe core discoveries is an area where it would be beneficial to use a public commons funding. If part of the licencing income from patents based on the discovery was channeled to the discovering party useful discoveries (those producing lots of wealth through their applications) would be rewarded, while useless discoveries would not. Here we get away from the info-socialist approach of centrally decided rewards (often unjust, and easily subject to corruption and public choice problems) although the solution of course still requires centralized patent licence taxation systems (ouch!) and could still get free software development into trouble. More work needed on that idea too.

The interesting thing about the intellectual property debate (once one gets away from the "piracy is theft"/"software should be free as in beer" shouting match) is that it is so non-obvious. It is a meeting ground between technology, philosophy, law and politics. We need to respect that. Simple solutions are unlikely to be workable since we seek to achieve a multitude of goals, many of which contradict each other and are ill-defined, and "we" actually are a large number of stakeholders from different cultures, economies and legal systems. At the same time whatever practices we develop need to be formulated in simple ways so that they can be explained to everyone: intellectual property is by its nature a cultural phenomenon, and as long as it is isolated to a small subgroup the theory will not mirror the practice.

So what we need to look for is an obvious (at least in retrospect) basic idea to construct a non-obvious praxis on.

Posted by Anders at May 21, 2004 11:41 AM