Collaboration is often encouraged in science: it is viewed as an intrinsically good thing. Yet there are downsides to collaboration. The most obvious downside is that requirement to coordinate the efforts (e.g., hold meetings). But there are many other downsides to collaboration:
- Collaboration introduces delays: at a minimum, you must let others react. Not everyone can always work at the same time.
- The result of any collaboration is always a compromise. Compromises are often political.
- Collaboration allows a parasitic behavior called social loafing: lazy or incompetent people are often eager to collaborate as widely as possible so that they can hide their small contributions.
I like to distinguish vertical collaboration from horizontal collaboration. Though this may be sometimes unavoidable, I dislike having to outsource low-level tasks like programming or data analysis (vertical collaboration). I prefer to collaborate with equals who can check my analysis, criticize my ideas or bring unique expertise (horizontal collaboration).
Toomela (2007) describes similar concepts. He calls vertical collaboration unidirectional collaboration because there is an individual at the top setting the direction and collecting the final results. And he calls horizontal collaboration dialogical collaboration because it requires that a whole team shares the same ideas and goals.
But Toomela goes further: he remarks that for elaborative knowledge, that is research problems with well-defined questions (e.g., prove that P is equal to NP), any form of collaboration will work. However, he adds that emerging knowledge, that is open-ended research problems, is ill-suited for dialogical collaboration.
Groups work better on legible tasks. For example, suppose we ask a group to estimate the weight of the Earth. If one member makes a mistake, others are likely to be willing to intervene to correct their collaboration. After all, the error is objective and factual. Through logical arguments, you can demonstrate the error. I would put a lot more faith in an estimate of the weight of the Earth signed by 12 engineers, than by a single one. From an information theoretical point of view, we could say that information about the problem can be exchanged efficiently between the different brains involved, so that they can work efficiently as a single mind.
But what if you don’t know exactly what the problem is? You are unsatisfied by the current state-of-the-art, but your understanding is not deep enough to ask the right questions. In other words, you have an illegible problem. It cannot be expressed in a logical form based on facts. In such a case, dialogical collaboration is likely counterproductive. I am much more likely to like the design of a piece of software if it was designed by a single individual rather than a large team. It is difficult to exchange information so that several brains cannot effectively work as a single mind.
I find it interesting that funding agencies often expect clear objectives. If you want to be funded, your problem has to be clearly stated, and you must present precise goals together with a clear path. We can deduce that most emerging knowledge is not directly funded.
Emerging knowledge is fundamentally a private business. The next scientific paradigm will not be funded by your favorite government agency, it will not come up of a committee, it will probably be invented by a single overworked individual.