Daniel Lemire's blog

, 2 min read

Bricolage by smart people

Scientific research is fundamentally about learning, about trial and error. Luck and unplanned interactions are a central part of it. Thus research cannot be planned and managed like, say, teaching duties or a Walmart store. If you could manage it, then it would not be research.

Research is usually greedy, in the sense of a greedy algorithm. At each point in time, you try to take the next best move, without knowing anything about the future. Maybe working on this new cloud computing algorithm will open the door to fame and fortune. Maybe you will meet a brilliant student next week who has great ideas on how to advance the field. Or maybe it is a dead end, maybe the problem has already been solved by a famous California professor last year. Typically, you do not know.

If you work hard and you are extremely clever, you might be able to make better guesses. But very, very few researchers can foresee the future 5 or 10 years ahead. Even Einstein got stuck in dead ends. Fernando Pereira is a leading computer scientist and an ACM Fellow who works at Google. He describes his view of research in similar terms:

Most successful projects I know, and certainly all that I have been involved in (…) started bottom-up, with zero to minor management involvement, and grew through repeated successful interactions with their environment. Pretty much like all the successful projects I’ve been involved in both in academia and industry.

However, much of research today is supposedly based on 5-year plans and funded by the government. The sort of plans that Soviet Russia liked so much. It should come to no surprise as Soviet Russia literally invented our government-sponsored research model. It is worth repeating that it is an absurd model, as Pereira puts it:

(…) in government-funded work I had to go through the ritual of pretending to know where the proposed work would get to in several years before doing the actual experiments. Which over time, as competition for funding increased, became the self-contradictory process of claiming the work was novel and required new funding to carry out while having already enough results to convince reviewers that the project was a sure thing.

It is fascinating how we have a hard time dealing with the fact that R&D is in fact nothing else but bricolage done by smart people.