It Seems Like Anyone Can Be a Data Scientist… but Is It True?
Yes, one can learn R and Hadoop and “claim” to be a data scientist, but that’s far from the truth. By comparison, one can also take a few medical classes and claim to be a doctor or watch a few courtroom TV shows and claim to be a lawyer. The difference is that the disciplines of medicine and law are “professionalized.” As a result, they are able to guard their gates by setting standards on who can call themselves a “doctor” or a “lawyer.” In data science, we cannot do that as of yet.
Insofar as R and Hadoop, they’re just part of the data science toolkit. They don’t constitute “data science” any more than a scalpel constitutes “surgery.” In the same way that physics relies upon mathematics, data science relies upon statistical tools for handling large and small data sets, structured and unstructured data, etc. But the mathematics of physics is not a substitute for scientific thinking, analysis, approach or method—and neither are Hadoop and R substitutes for understanding behaviour in data.
Statistics, specifically, is concerned largely with methods for testing hypotheses using data; thus, before one can constructively use Hadoop or R, one needs to know statistics and know it well. Because, unlike statistics—which is concerned largely with testing the hypotheses and stops there—data science focuses on the implications of systematic departures from hypotheses (as evidenced by statistical tests) and the bigger conclusions we can make as a result of those departures.
Moreover, apart from data science requiring a cumulative knowledge of numerous tools or sub-disciplines, like statistics, R, Hadoop, etc., one must be able to bring those tools to bear in answering important business questions and achieving business outcomes, neither of which initiative directly follows from knowledge of the tools. That ability, skill, experience or talent, is what’s brought to the table by the data scientist, enabling him or her to justifiably call himself or herself, a “data scientist.”
This leads me to believe that the real question here is, “Can anyone BE a data scientist?” And to that I would say no, not at all, for the very reasons I just mentioned. In my experience, not even your top CS or STEM majors from a top school can easily become good data scientists, without additional training in it, and some personal factors. Apart from its multidisciplinary nature, data science requires a deep love of the divergence between observed reality in data and the prediction of mathematical models. To do that, one needs something more than just a mastery of tools. One needs a love for imperfection.
I’ve been in this field almost 20 years, from back before the term “data science” existed, so I’ve seen many things. In fact I believe data science excellence requires a number of years in actually applying it before one can truly understand data, how it behaves, how different models work, backwards and forwards, etc. Yet, most importantly, excellence requires making mistakes and understanding mistakes, along with appreciating the variations between observed and predicted reality. Thus, I affectionately call data science the science for imperfect people, like myself.
I say that half-jokingly. In truth, I believe all good science is for imperfect people—people who become curious, not angry, when they see imperfection and variation. STEM majors that cannot stand imperfection and variation will never make good scientists or good data scientists, much like bigots cannot make for good neighbours. Why? Because the world we live in is imperfect and variable and its beauty lies in that imperfection and variability.
