The Perils of Predictive Analytics
For centuries people have been captivated by the idea of predicting the future. Crystal ball gazers and fortune tellers all promised to be able to do this. They played on our biases, weaknesses, and gullibility and counted on us attributing chance occurrences to their predictive powers.
The rise of predictive analytics gives us the ability to reduce uncertainty by applying statistics and determining the probabilities that future patterns will emerge in the behavior of people and systems.
The Internet provides a platform for us to communicate, share, buy, play, and learn. And because people are largely creatures of habit and tend to repeat behaviors, our online activities when combined with today’s computing power and statistical knowledge, tell a lot about what we are likely to do. We can give odds, based on science, about what will most likely occur. To do this has required access to mountains of data about what we do, when we do it, how often we do it, and where we do it.
By tracking things such as our location, Facebook likes, retweets, where we check-in, what and when we buy, what we search for and so on, analysts are able to make reliable predictions on our future behavior. This data is often called “data exhaust” by analysts as in and of itself it has no real meaning or value. However, when aggregated, correlated, and combined and then analyzed with the tools of statistics this data becomes not only relevant but commercially valuable.
Privacy Invasion We are being monitored and watched every time we log into any electronic device whether it is a computer, a mobile phone, a tablet, or a game. And everything we do is collected without us being aware. We do not give permission for it to be collected, in most cases, nor do we have any control over what is collected. And we have no way to turn off the monitoring.
For example, when we buy something, it is not hard to predict that we might buy more of it. It is even possible to narrow this down to specific types of items, the amounts we spend and the frequency we buy them. Or when we do something as simple as check into a restaurant or hotel, we leave a location trail as well as an economic trail. Combined with a profession, easily derived from a LinkedIn or Facebook profile, this data can predict with a high degree of certainty where we are likely to be at a given time. It can also predict how often we will be there, what kind of hotels we prefer, perhaps even the type of room we prefer, our income, and much more. And all of this can be sold to an hotelier or retailer, for example, without our knowledge or permission.
Commercialization that Plays on Our Predilections Predictive analytics has had tremendous commercial benefits. Firms such as Amazon are built on predictive analytics that help them predict what we will buy, how much of it and when so that they can stock warehouses and order products before they are needed. There has never been so much demand for statisticians and analysts.
Much of the work in developing predictive analytics has been paid for by Madison Avenue, Wall Street, and the retail world. We are marketed to heavily based on our location, age, socio-economic status, and past behavior. Products are recommended to us based on a prediction about what we are likely to buy.
Shoshana Zuboff, a Harvard professor and no fan of predictive analytics, has focused her research on the study of the rise of the digital world, its individual, organizational, and social consequences, and its relationship to the history and future of capitalism. She is concerned that we are applying analytics to making money and toward turning us all into “slaves” of the commercial world.
She says, in her article entitled “A Digital Declaration,”
“Now the focus has quietly shifted to the commercial monetization of knowledge about current behavior as well as influencing and shaping emerging behavior for future revenue streams. The opportunity is to analyze, predict, and shape while profiting from each point in the value chain.”
Biases that impede truth All humans have biases, and many that tend to impact human resource professionals and recruiters.
The selection and hiring of people are fraught with bias and subjectivity. Psychologists have assembled long lists of these biases which include our tendency to reject new evidence that contradicts something we believe to be true. Or the tendency to search for and remember information in a way that confirms our preconceptions.
For example, if we believe that people with high GPAs, for example, are better workers, then we will seek evidence to prove that and dismiss any that contradicts it.
Recruiters also often rely too heavily on one trait or piece of information when making decisions -often the first piece of information acquired or the information obtained from a trusted source. If someone recommends a candidate, for example, that recommendation may outweigh any facts that contradict or suggest that the person is not so good.
Many recruiters and hiring managers also suffer from what is called the “Hothand effect” which is the fallacious belief that someone who has experienced success doing something has a greater chance of further success in additional trys.
For a thorough discussion of how bias affects all of us and everything we do, grab a copy of Cathy O'Neil's Weapons of Math Destruction. This is an insightful analysis of bias in algorithms by a Harvard Ph.D. and ex-hedgefund quant.
Analytics can help dispel many of these, but only if we believe the results of the analytics and act on them. There are many instances where we build our biases unconsciously into the algorithms that analyze our data.
Analytics can offer insight and help make sense of mountains of data that have been beyond our reach. Analytics can help us make choices based on facts. They can provide us insights and reduce uncertainty. But, as with everything, there are dangers. We need to troll the waters of data with care, ethics, and human judgment.
------------------------------
I will be speaking about predictive analytics at the Human Resources Executive Talent Acquisition Tech Conference in Austin, Texas in November. Hope to see some of you there.
Follow me on Twitter @kwheeler. If you like this, you might like to read my other articles and visit www.futureoftalent.org for more ideas and white papers. Contact me at [email protected] if you'd like to inquire about having me speak at an event or to your team or leadership.
About the Author: Kevin is a well-known consultant, keynote speaker, and futurist. He focuses on the future of work, recruitment, human resources, learning, education, and leadership development. He is a sought after speaker at events globally. He writes extensively about the future, ways to improve and streamline current work practices, and about emerging technology.
He is also the Founder and President of The Future of Talent Institute, a think tank dedicated to seeking out trends and issues that will have a direct effect on talent, how people will work, and how they will be employed in the future.
Kevin a great essay on predictive analytics and it use on humans. We on the other hand are using it to predict the behavior of machines and when they will fail. Having been through the HR cycle of no face to face meetings before a resume is churned up by the grist mill, I completely 100% agree with the analysis. Nothing beats the face to face, real life analytics of the human brain ;D PS I am now a follower of yours. Keep up the posts
Global Talent Executive | Board Member | Talent Acquisition Thought Leader and Innovator | Led teams to hire >1,000,000 people | Tech / AI / Automation Enthusiast | Host of ‘Growing Your Business with People’
8 年Kevin, Hope you are well! I really liked your article as it really explores some of the tensions of predictive analytics. I find the ethical tensions seem disturbingly unclear. Only thing I may have built on a bit is when you talked about the fortune tellers who played on our fears and desires - and how there is a new group of these people using the cloak of "predictive analytics" as their "crystal ball" (pulling together random correlations in an attempt to draw causation). This is particularly dangerous in the area of hiring/talent acquisition- as often times unforeseen biases are baked into the randomly hobbled together "algorithm" - not even mentioning that there is low likelihood of any measurable increase in the quality of hires (outside of what expectancy theory may afford). I also particularly liked the discussion on the balance of privacy and consumer benefit (e.g. Amazon). It is a very complex area - especially as you move from consumer products to areas like healthcare. We want everything seamless, simple and easy, yet want to feel as though we have high levels of personal privacy. Tough balance to achieve. Thank you for a thoughtful article. Jeff
This post raises some very good points, but must admit I have a different take: At its core, predictive analytics enables informed, data-driven decisions to improve your chances of success. It should turn a complex mound of disparate data into easily consumable insights, and deliver quantifiable ROI to the business. By no means does it guarantee outcomes, but rather like the "Vegas edge," you'll win more than you lose over time. It is not about snake oil or fortune telling. It is about dispelling myths and letting the data speak for itself. Yes, bias influences decision making at organizations, leading to significantly less optimal outcomes. This is exactly why predictive analytics is so critical, and should all but eliminate that bias if organizations are prepared to act on the findings. With respect to recruiting, what organizations really need, and struggle to measure, is quality of hire. Are we hiring people that perform better and stay longer? Time to fill and cost per hire, or butts in seats fast and cheap, without considering performance and retention, isn't a recipe for success. There is only so much recruiting can do to influence cost per hire and time to fill anyway - the biggest driver is mix of jobs hired. So, what factors should we really care about... test scores, personality assessments, education or industry experience? And, what's the financial impact of better recruiting profiles that, on average, result in 25% more productive hires that stay with the organization 1.5 years longer? That is the beauty of analytics. Just add data, and you can answer all those questions and more. For now anyway, I'll stay away from data privacy, ethics and swimming upstream against big data. Thanks Kevin for sparking this thought provoking discussion!
President of Regions Beyond-USA
8 年Twenty years ago this article would have been entitled, 'The Perils of Statistics' . . . not much has changed, only the pop culture of new buzz words.
Former CFO and Fortune 100 Global Talent Acquisition Leader
8 年Agree Kevin, and HR often seems to be particularly susceptible to the false promise of "manufactured answers". Truly gaining value from analysis requires thinking harder, not less.