I promised you all a final rant. Well, here it is: university science education doesn’t produce scientists.
What’s a scientist anyway? Someone who pipettes all day, or stares through a telescope? That could just as well be a technician, and often is, even if hidden behind a title of "postdoc" or "professor". Let’s take some folks who are scientists by any estimation: Newton, Linnaeus, Mendele’ev, Darwin, Boyle, Einstein. We could easily add an astoudingly diverse range of names. On the flip side we can name many who were appalling scientists, such as Watson and Crick, Thomas Aquinas, Lysenko, or any of the proponents of "intelligent design" that plague us today.
What separates these two groups? It’s not that one worked on a particular thing or in a particular way. The day to day methods of Darwin were much closer to Lysenko than to Newton. It’s not intelligence. Newton was indubitably a genius, but so was Thomas Aquinas.
It’s a question of virtue.1 We utter "Lysenko contaminated his science with ideology" and "Watson and Crick stole data" and "Intelligent design beggars the question" in tones of moral outrage. These men are epistemically evil, just as Boyle and Darwin are epistemically virtuous. The virtues vary — objectivity was not one of Mendel’s great virtues, nor generosity with ideas one of Newton’s — but all acted in a manner which embodied some range of epistemic virtues.
So: a scientist is an epistemically virtuous individual.
Now look at the graduates of science departments around the world. They aren’t particularly epistemically evil, nor very epistemically virtuous either. Virtue is a question of habit. Decisions make ruts in our minds and repeated action, virtuous or evil, digs ruts a man can’t easily escape from. There are no such ruts impressed into the minds of most of the children emerging from university with their degrees. They have memorized some facts, perhaps learned some technical skills, but we cannot call them scientists.
Some of them go on to graduate school, where they may be shaped by an advisor, but it is just as likely that they will be ignored, and who knows what ruts will develop? In some fields, such a molecular biology, this has gone on for generations, and it is only by chance that you may happen upon a professor who is a scientist.
This is all very depressing. If lecturing the children and making them do laboratory experiments does not produce scientists, how can it be managed? We must drive some ruts through their minds in desirable directions, which can only be accomplished my making them do some science, and do it in such a way that they are not merely technicians but actively make and take responsibility for decisions with epistemic consequences.
They still need technical skills, or they won’t be in a position to act at all, virtuously or not. Some degree of the following are needed for a scientist:
Mastery of both her native language and English to the point where she doesn’t create impressions in her audience by accident. Her palette of intentionally created impressions may be limited, but she will not inadvertantly lead her audience astray.
Technical fluency in some means of experimenting, whether carpentry and metalwork, manipulating liquids, mathematics, or programming. Whatever the medium, the scientist must turn to her tools without mental hesitation.
Explicit thought. Anyone who has taught programming knows most humans have never had anything but vague thoughts. Faced with the computer, they struggle. Much epistemic vice is due to an inability to recognize vagueness. Programming is the best way to learn this explicitness.
Organizing definitions and categories. Careful choice of definitions and their arrangement vastly changes the effectiveness of a scientist’s thoughts2. The best way to learn this is to solve problems in a dependently typed programming language like Agda in such a way that errors are semantically impossible.
Expressing ideas directly in mathematics, without passing through English or any other spoken language. Spoken languages are poor vehicles for new thoughts. There have been great scientists who couldn’t express their ideas directly in mathematics, and they have, one and all, lamented their inability to do so. There is no such thing as a non-mathematical science, just fields which awkwardly embed their mathematics in English. It is also impossible to understand statistics in any useful way without this fluency.
Break the human habit of thinking in terms of "A acts on B". A mind that has not replaced this with basic notions of relations and mappings is crippled. Essentially, every scientist must internalize the essentials of category theory.
Once we have a student with these prerequisites, how do we drive ruts through her mind? We must give her role models and guidance, and then put her in situations where she must apply epistemic virtues.
The role models needn’t all be present, or even alive. A scientist who hasn’t read the founders and great investigators of her field, and understood how and why they approached the problems as they did, is at best a dilettante. A physicist who has not read Newton and a neuroscientist who has not studied Cajal are both pitiful creatures.
Someone does have to guide the student, though. Someone has to choose problems big enough to challenge her but not so large as to swamp her or make it too difficult to act virtuously (remember, we must be sure we lay ruts in the right places). Someone has to be a colleague to the student as she tackles the problem, and critic when she has finished. The written works of dead men don’t suffice.
And what kind of problems should the student get? I’ll offer some ideas which cover epistemic issues I think all scientists should address:
Measure something out in the wild that cannot be found via controlled experiment in the laboratory. It could be the population of fish in local waterways, or the incidence of a disease in the area, or anything in astronomy. Observational studies offer scope for epistemic problems which the laboratory scientist can blissfully ignore.
Plan an experiment limited by available resources. It could be a small clinical trial or an industrial project, but the outcome must matter and the resources to run the experiment must be sufficiently hard to come by that the student must do the experimental design and analysis correctly. All scientists today must come to terms with the foundations and practice of statistics.
Measure a parameter with a precise, fixed value, such as the speed of light or the fracturing stress of a material. The student needs to know the worry of hunting for systematic errors in an apparatus while trying to produce a correct value.
Choose the objects of study in an area without an established theory. Scientists who have always had clearly delineated definitions — the mass of an object is important, its color is irrelevant, etc. — are often at a loss when faced with nature in her raw form. The student should be faced with raw nature with no guiding theory and forced to impose mental order on some corner of the world.
Explore a solid theory, such as classical mechanics or chemical thermodynamics. This is a chance to know what the outcome of successful science looks like: a solid, predictive theory on which you can base a discipline of engineering. The student also learns how much there is yet to be done in even the best established fields.
Engineer a tool. The student must learn the difference between hacking something together for herself and producing a tool for others. There is only one way: she designs the tool, then is forced to watch silently while someone else tries to use it. Then she redesigns it, watches again, and on and on.
This list is by no means comprehensive, but it provides the student with a chance to exercise many epistemic virtues, and to find those which suit her best.
Unfortunately, this kind of education would require restructing the universities and firing a large number of professors. Since that is unlikely to happen, however desirable it might be, I’m afraid it’s just going to be a studiously ignored ideal.