Predictive evaluations: Using AI to reward (or punish) employees for future performance

Human resources professionals are (or soon will be) using artificial intelligence (AI) to recruit and interview job candidates, to measure an employee's value, and to conduct performance assessments.

The next step - which some companies are already taking - is to use AI to assess not past, but expected future performance. From Bloomberg 's Rebecca Greenfield:

Using artificial intelligence, Watson Analytics looks at an employee's experiences and projects to infer the potential skills and qualities each person might have to serve IBM in the future. Watson also scours IBM's internal training system to see if an employee has gained new skills. Managers then take Watson's assessment rating into account as they make bonus, pay and promotion decisions.

IBM says Watson has a 96% accuracy rate, Bloomberg reports, though it's hard to know what constitutes 'accurate.'

Further, the notion of your compensation being based on predicted future performance is a bit unsettling. For starters, future performance can be dependent on a number of factors beyond an employee's control, such as quality of training, workload, the dynamics of the workplace, and even whether their boss is a jerk.

Making personnel decisions based on future performance is not new; professional sports teams do it all the time. Yet athletic performance is a very specific thing, and there are decades of studies and analyses showing that athletic skills deteriorate at a remarkably consistent rate. For example, only a desperate or foolhardy NFL team would offer a large, multiyear contract to a running back who has passed age 30. That's because most running backs peak in their early to mid-20s, and experience a sharp drop-off in ability by age 29. No wonder; thousands of high-speed hits by large defenders eventually wear down their legs.

A worn-out halfback can't become 24 again or magically undo the damage from years of physical beatings on the field. Enterprise employees, on the other hand, can learn new skills or perform better under the right supervisor. I'm not sure how Watson takes those variables into account when performing predictive employee reviews. It may have a method - let's face it, Watson is pretty darned smart! - but I don't know what it is.

Should an employee's compensation (or even employment status) be tied to predicted future performance, or is that rewarding (or punishing) someone for something that hasn't happened yet? What do you think? Let us know in the comments section below.