Terrible Advice From a Great Scientist 276
Shipud writes "E.O. Wilson is the renowned father of sociobiology, a professor (emeritus) at Harvard, two time pulitzer prize winner, and a popularizer of science. In a recent article in the Wall Street Journal, Wilson provides controversial advice to aspiring young scientists. Wilson claims that math literacy is not essential, and that scientific models in biology, intuitively generated, can later be formalized by a specialized statistician. One blogger calls out Wilson on his article, arguing that knowing mathematics is essential to generating models, and that lacking what Darwin called the "extra sense" is essentially limiting to any scientist."
He's right (Score:5, Insightful)
Math, intuition, and insight are all important. But they don't all have to come from the same person. I have worked on plenty of teams where the creative work and number crunching tasks were delegated to different people. I am currently working on a 3D educational game, using OpenGL. It involves lots of gnarly trig and vector math, which I am good at. It also involves lots of creative scene design and character development, which I am not good at. So I work with an artsy chick, and we make a good team. I don't see why splitting creativity and implementation shouldn't work for biology as well.
Re:He's right (Score:4, Insightful)
Science doesn't work like that.
Re:He's right (Score:5, Insightful)
Intuition and the part of math that involves being good at grinding through lengthy, dense calculations without making sign errors don't have to be the same person. However, a strong and intuitive sense of what math is capable of (which requires advanced mathematical education) do need to go together for scientific productivity. Otherwise, it's just like the techno-incompetent manager asking engineers to implement his "brilliant" physically impossible designs.
Re:He's right (Score:5, Insightful)
I know of certain articles in highly recognised journals which passed the review process, pushed by the editors who liked the message so much.
I also know that their data was largely noise, because the main authors clearly are math illiterate. Of course not everyone needs to be a mathematician, but every scientist should know the basics of statistics and be able to recognise a binomial or Poisson process after a cursory glance at the data.
Likewise not everyone should be some über-coder, but every scientist should be able to write small programmes in MATLAB, R, numpy, or whatever is appropriate for their field. These are basic qualifications which prevent you from churning out bullshit.
Re:WSJ article title is somewhat misleading. (Score:3, Insightful)
Ssh, we're busy crafting a strawman here, and you're just trying to blow it down!
Re:He's right (Score:5, Insightful)
But it should be obvious that for some other problems this won't work. For example, it doesn't make sense to try and split the coding into a "creative coder" (who knows nothing about programming) and an "implementation coder" who turns the creative's ideas into actual code. The creative would toss out nonsensical ideas (like "instead of using vectors, why not use genetic algorithms?"), and then the implementer would have to explain why all those ideas are silly... or else they would just have to ignore the creative type and simply code something that makes sense.
In other words, generating good source code requires someone who knows enough about programming that they can see creative solutions. Their intuition is not separate from their programming talent: their intuition is based upon years of training and experience with source code, math, engineering, and so forth. That's where the good ideas come from.
Coming up with good scientific ideas is similar. Analysing scientific data even moreso. It's only once you have a deep, subliminal understanding of the important concepts that you're going to make substantive progress. Whether a deep understanding of math counts as an "important concept" depends on the field, of course... but I would argue that for science generally, the more mathematical know-how you have, the more informed and powerful your ideas will be.
"literacy" is not "skill". (Score:5, Insightful)
Sure, the roles do require "math literacy" which is a lower standard than "sufficient mathematical and comptuational capability to independently produce results for a research journal."
Just like natural language literacy is a lower standard than powerful, skilled writing.
Fascinating insight (Score:5, Insightful)
My 'aunt', who still works for a pharmaceutical firm analyzing statistical analyses by researchers, would snort tea out her nose reading this. Doing the research, finding a useful drug, doing minimal testing, and then concocting the analysis to fit the very limited empirical model is not uncommon in the drug industry. Her job was and is to study that 'analysis', identify any problems, send it back for improvement, and repeat until either the researchers give up and move on to something they can demonstrate is effective AND safe enough for the market, or succeed and are able to show provable, reliable results.
Wilson would not like herm, and for good reason - she would call his methods little more than guessing. She has proven repeatedly that well-meaning researchers can find some statistician to lend unwarranted credence to imaginary results.
Kinda sad that this passes as science at all. Wilson seems, to me, to be stating that research need not be proven, merely justified.
math comes second (Score:2, Insightful)
Re:He's not right (Score:5, Insightful)
Collecting data without having a darn good grasp of how the data analysis works is a great way to waste a huge amount of time and money collecting mostly useless data. It may not be the same person doing both, but the data-collector definitely needs to be intimately "in the loop" about how their experimental work impacts uncertainties in the final analysis.
Re:He's right (Score:4, Insightful)
Re:He's right (Score:4, Insightful)
I'd rather train a mathematician to be a biologist than the other way around.
With sarcastic apologies to Alex Belits, it doesn't work like that. I mean, it might, but there's issues of both interest and aptitude. Personally, I think you're both right. The best situation is to have both mathematics ability and some other kind of ability concentrated in a single human. Barring that, you can still get things done, perhaps just not as quickly. Thus, it is still preferable but not essential for everyone to have strong mathematics stills.
Re:He's not right (Score:3, Insightful)
Without understanding the measurements and statistics involved, the experiment design will most certainly turn out to be crap.
Here, fixed that for ya.
Re:He's right (Score:4, Insightful)
This would be funny if it were not true... Beware the biologists who tells you they found a number N of categories of each suspiciously smaller than the previous one by the same ratio. And never checked for fear that their result might not be publishable after all.
As in : "This a ground-breaking, paradigm shifting result: these identical individuals are not: some are short-lived, some long-lived, and we found an intermediate category, too" -- "oh, so your mortality curve follows an exponential law".
Re:He's right (Score:4, Insightful)
Bullshit. Any scientist needs to understand basic maths, notably statistics. Not advanced calculus or complex algebra. But statistics and understanding what a model is is paramount. If you cannot recognised the patterns produced by common types of random processes, you may well start to believe you have found something.
And in fact just measured experimental noise.
Re:"literacy" is not "skill". (Score:4, Insightful)
But the math proved what they wanted to show, therefore it was "good enough"
Re:He's not right (Score:4, Insightful)
I was a natural sciences major in college and what you're talking about is one or maybe 2 classes worth of math. You don't need calculus or anything beyond that in most cases to design an experiment, obviously depending upon the particular field of study. Statistics itself is heavily derived from a set of formulas that you can look up in a book and the reasoning behind it requires at most intermediate algebra to understand.
I definitely agree that you need an understanding of statistics to design your experiments, but really, the amount of math you really need is surprisingly small given that you're going to want to bring in an expert that's experienced in the specific area you're working anyways. Now, were we to go back in time to days when there wasn't a huge team, that would presumably be a different matter. But, understanding doesn't really require that much math.
TL:DR, you're going to want an expert in dealing with modelling and data of the type you're looking at. It makes more sense than reinventing the wheel every time you do an experiment and forcing people to master not just one specialty, but several of them, and ultimately it's unlikely that they'll achieve a level high enough to compete with the best in both fields.
Re:He's right (Score:4, Insightful)
True, but statistics usually requires intermediate algebra and could probably be taught in highschool without too much trouble. And the bottom line is that the formal theory is neither necessary nor sufficient for somebody to look at the data and see meaningful patterns. As long as you have somebody on the team that can whip up a model to fit the data that you can then test against future experiments, you're fine. There's no particular reason why any particular scientist needs that specialty.
And anyways, it's not just the statistics, it's the experience of having crunched many numbers and found many errors in the past. Realistically a specialist is much more likely to find the problem in an efficient manner than the other team members. Do enough math and eventually you can pretty much see the errors without even trying.
That being said, everybody really should have that class as part of their education as it's so helpful during the 99.9999% of your life where you don't have a statistician on hand to analyze your data for you.
Re:He's not right (Score:5, Insightful)
Comment removed (Score:4, Insightful)
Re:from the father of handwaving (Score:4, Insightful)
Don't delude yourself: This is anti-intellectualism. Sociobiology has issues, but that's not because it's got "socio" in the name.
It's because evolutionary explanations have extremely high status, meaning they are often reflexively believed, even when they can't be backed up. It has become (has always been, really) a refuge for the kind of people who would rather make "bold statements" than work incrementally to increase our understanding. Wilson's statements sums it up all too accurately: make the statements now, leave to others to test it mathematically later.
On the contrary, social sciences have extremely LOW status, as your prejudicial comment sums up. Have you heard about the Cochrane collaboration, evidence-based medicine? You probably have. Why did it take so long to appear? Because medicine and molecular biology has high status, whereas the "social" population studies of epidemiologists had low status. So if the high-status people said, "from our understanding of molecular biology, this should work", for a long time that would be tried, even though from a 10.000 feet view it would have been obvious it did NOT work.
You need both kinds. You need people who take the bottom-up approach, building bricks of what we know, and assemble it into bigger things. Then you need to have people who take the top-down approach, because no matter how well the pieces fit, it's no good if the larger building can't actually stand. In some fields, like physics, these are closely intertwined. In others, they are tragically separated. For that to change, the white-coat status prejudices of people like you need to be broken down.