Machine Learning Depends on Good Data—Here's How to Clean It Up
11 de septiembre de 2018
Artificial intelligence (AI) and machine learning (ML) no longer just the stuff of science fiction—the technologies are here to stay. But after forty years of seeing the next best thing in HR technology, I have learned that implementing it is never as easy as it looks.
In the 1980s we "went paperless." In reality, we simply automated paper-based processes. We justified the expense by saying it would cut down personnel costs—it didn't. We pushed data entry out to the end user and did away with the human positions that audited our data for logic and accuracy. But then we realized we couldn't trust the technology, so we duplicated processes. Today, the promise of paperless remains somewhat unfulfilled.
A recent article in HBR online caught my eye. It discusses the dangers that may come with machine learning if we don't do a better job of making sure the data we input—be it candidate data for recruiting purposes or employee data used to determine appropriate learning and development opportunities—is accurate and timely. Putting in bad data means you'll ultimately be making bad decisions.
As we think about how machine learning will shape the future of hiring, talent management and learning and development, we simply cannot afford to have algorithms use bad data—the future is too exciting. Here are three ways to prepare your data for the future.
1) Practice Intentional Data Entry
HR data is entered by employees, managers and HR teams. What do they all have in common? They're prone to human error, especially when it comes to inputting data such as hours, dates and termination codes.
Combat potential errors with a data integrity campaign. Let employees and managers know the importance of data accuracy. Help managers understand that the termination code that they use informs metrics that dictate employment and business decisions. Help employees recognize that their time contributes to the accuracy of their paycheck.
2) Set up a Data Quality Process
Managers and employees may initially be skeptical about your data integrity campaign, so take it one step further and establish a quality control process within the HR department. Those HR specialists that are most invested in the data accuracy can serve as quality control.
HR professionals who partner with line managers can keep data front and center through reports and, more importantly, through personal interaction and coaching. Reports are so prolific these days that just presenting data is not nearly enough. HR can build a partnership with line managers by working through the implications of the data and then drawing and testing conclusions.
3) Edit Inaccurate Data at the Source
I once watched an HR professional make a data change on a spreadsheet after a manager had called attention to the inaccuracy. After we left the meeting, I asked if she was going to change the inaccurate data in the Human Resources Information System (HRIS). She replied, "Oh, I guess that's a good idea." Forgive me, but duh.
1) The termination code for a former employee is wrong. Why?
2) Because the manager used the wrong code list. Why?
3)Because an old code list is still on the intranet. Why?
4) Because it has not been audited and updated for a year. Why?
5) Because Susie, who was tasked with managing it, retired last year, and no one else knew to fix it.
Ah! Now you have a root cause that can be fixed and future data errors can be avoided.
Data and technology often get relegated to an administrative role in HR, but that is a grievous error. Any people decisions that are made based on bad data present significant risks to the organization. The future of HR technology is exciting. Let's get ready for it with better data now.
Photo: Creative Commons