There have been many articles written about the danger of putting data science tools in the hands of analysts. Warnings have included concerns that reckless use of the “black box” of machine learning could lead to financial ruin or some other terrible consequence for the business.
So is the fear warranted? Have we really seen some incredible damage done by analysts using these modern techniques?
First, a Bit of History
Technological advancements that apply mathematics and provide automation to analyses are certainly not new. I can only imagine the conversation as calculators became available for people to use in business. Were there people suggesting that providing a calculator to someone who used an abacus, or who performed math with paper and pencil, would lead to devastating results? If someone used a calculator but didn’t understand the complexities of all the algorithms, would they perform better or worse than the trained expert with the abacus?
My experience in data science and watching technology applied to different domains would suggest that the use of these tools has reduced errors and improved the accuracy of the results, hence why adoption has been so widespread. Interestingly, the safety net to prevent errors has remained the same. The failsafe is not the math and science expert, but the domain expert who can look at the results and understand if they make sense in the context of the subject matter.
But I’ve Heard Stories of Huge Issues with Machine Learning and Data Science!
Certainly, mistakes have been made with bad models, incorrect conclusions and even typos causing impacts to businesses. The reality is that these types of mistakes have been occurring for decades. I have personally seen numerous examples where a complex spreadsheet was rebuilt using modern technology with an Alteryx workflow and the answers didn’t agree. Every time I have seen this, the newer technology provided the correct answer.
While I’m sure someone can find an example that went the other way, I’m certain that you are less likely to make a mistake using the more modern approach to solving complex analytic processes than by using technology from two decades ago.
Will mistakes be made? Certainly, the art and science of analytics is still human powered, and humans make mistakes. However, I’d want to equip my analysts with the best possible tools for the journey.
So What is Happening?
New technologies that allow analysts to learn and leverage Machine Learning (ML) and advanced modeling techniques have become easy to use and readily available. The challenge is quickly becoming cultural versus technological, and this obstacle is widely recognized.
NewVantage Partners recently released their 2021 Big Data and AI Executive Survey. Their results showed that 92% of executives reported the greatest challenge to becoming more data-driven continues to be cultural, not technologically driven.
Developing Data Science Confidence
Education is the key to fulfilling the incredible impact these technologies can provide to organizations around the world. That can include both traditional education at all levels, but it also includes other ways to gain knowledge and comfort with data technologies. Tools like Alteryx that offer an entry point for machine learning also include built-in informative features to help users understand what the tools are doing and how to interpret their results — adding a layer of learning to productive hands-on experience.
And, just as important, this approach lets the domain expert with critical subject-matter knowledge maximize their impact on the organization. Since these folks already serve as that failsafe for identifying any risks of machine learning, why not directly hand them the most powerful tools available and empower them to solve complex problems.
New technologies always bring about new fears. But if we’re going to make data-driven decisions, we have to admit there’s just no evidence to show there’s a true danger from putting data science tools in more people’s hands. And as these technologies advance and education continues, we’ll see more and more confidence in those newfangled calculators and less nostalgia for the days of the abacus.
Watch This Next.
Listen to Alan’s interview about Alteryx Machine Learning and the Alteryx open-source Python libraries on our Data Science Mixer podcast.
You can find show notes here.