Citizen data scientists: WE NEED YOU!

First let me ask you a question. Did you know that Miss America’s age is closely correlated with the number of murders by steam and other hot items? Or that the stork population is related to the birth rate?

If your immediate reaction to this was that it was coincidence, then congratulations. You can now print out your certification as a data scientist. Why? Well, put simply: only a well-trained and experienced data scientist would immediately recognize these examples as false correlations, or so Gregory Piatetsky of KDnuggets, the online platform for data mining, asserts.

In his article, “The Mirage of a Citizen Data Scientist”, he gives examples of why citizen data scientists could be considered a curse rather than a blessing.

 

He cites, for example, the idea of a plane flown by a combination of an untrained pilot and a reliable autopilot. Most of the time, he suggests, the autopilot would be fine. If anything went wrong, however, the autopilot would immediately hand over to the untrained pilot, just when experience was most needed. Piatetsky believes this is similar to citizen data scientists. They are, in effect, largely untrained operators dabbling in data science.

I think, though, that Piatetsky is wrong.

 

These citizen data scientists are usually people with deep insights into their respective business fields. They are familiar with their data and what they say is often backed by other evidence. This does not sound to me like ‘dabbling’. Nor does Gartner's definition of citizen data scientists:

“People on the business side that may have some data skills”, or those “from a math or even a social science degree - … putting [their skills]to work exploring and analyzing data”.

Discarding the idea of citizen data scientists altogether risks throwing the baby out with the bathwater. But critics like Piatetsky are right to say that there are some things that can only be done by qualified and experienced data scientists. These, include, for example, answering questions like:

• Is there any reference to business process analysis?
• Do we have the right data, and is it clean?
• Can we clearly interpret the results?
• Have we used the correct method?

Here is another example. Your company wants to open a new branch. As the data scientist involved, you do a lot of analysis. You look at the competition, the visitor flow, and the purchasing power in the region and then make a forecast of profitability for each possible location. Finally, you make a recommendation. You have done everything right. But in the end, managers make a different decision.

Wasted work? Possibly, but perhaps you just asked the wrong question. Instead, you needed to ask, “How are decisions made around here?" The answer would have been "By managers, based on gut instinct", and you would have saved a lot of effort. This type of knowledge is only really available to citizen data scientists, rather than ‘pure’ data scientists.

Of course, we need data scientists, those who know and understand the detail of data analysis, with their experience and their knowledge of scientific and statistical methods. But these people are scarce. So why not go the other way and bring the business expert closer to the analysis, providing training to help them develop their skills? SAS has recently started providing a training program for citizen data scientists to achieve this very thing.

For me, citizen data scientists are an obvious answer to a problem. I expect them to play an important role in future, bringing their own expertise and curiosity to the task. I believe we are seeing the dawn of a new era in data science.