Why a Professional Association for Data Scientists is a Bad Idea
This post presents the argument against having a professional association for data scientists.
By Anthony Tockar, Verge Labs.
It's no secret that this a profession with a very low barrier to entry - grab a certificate on Udemy or Coursera and you're set! It's also common knowledge that data science is hard - after all, a solid grounding in statistics and agnostic programming skills don't come easy - not to mention the need to round it out with a good dose of business acumen. It doesn't quite add up...
Enter the Association of Data Science Professionals. With guarantees on the credibility of chartered candidates, companies can now be safe in the knowledge that their data scientists are the real deal.
Except there's one problem. Few data scientists worth their salt would join such a thing.
Consider some of the things that make a good data scientist:
The focus in the majority of data science courses is on the final point. Granted, there are some courses that also focus on communication, but the fact remains that data science is an applied discipline, and is learnt best by doing. The above reasons are why aspiring data scientists are encouraged to share their work online (via blog, github, kaggle, etc) - it's a far stronger signal than a degree or an accreditation.
A focus on the technical also has other limitations:
Who will administer such a program? Even experts in this field have knowledge that is obsolete in many areas, and even then there are few agreed-upon "unicorns" around.
Can it keep up? The pace of innovation is frantic, and the best education programs are constantly revised.
Will it stifle innovation? Data science is a young field, consisting of practitioners with wildly varied backgrounds. Conforming to a homogeneous syllabus could limit the ability for certain areas and applications to reach their potential.
The assumption in the above paragraphs is that technical qualification is a requirement for such a body. Certainly, there are other industry groups that have no such requirement, but few could argue that membership of such a group provides much of a signal. The field is far too large for any operator to hope to license members with any objectivity apart from standardised testing, and thus we are drawn to this mechanism as a foundational plinth.
Original. Reposted with permission.
- How Do I Get My First Data Science Job?
- Introducing the Anaconda Data Science Certification Program
- Data Engineer vs Data Scientist: the evolution of aggressive species