What Is Artificial Intelligence In HR? Getting Past The Buzz About AI In HR

By Perceptyx - July 23, 2019

For many, the mention of artificial intelligence (AI) sparks a number of misperceptions, thanks to movies and media sensationalism. Industry hype, overselling the extent to which AI can take the place of human effort, often compounds these misperceptions. Artificial intelligence in HR (as in other fields) refers not to sentient machines like the Terminator, but rather to computer code and mathematical models that perform the specific functions that they are programmed to perform.

As an industry term, in terms of working with data, artificial intelligence does not denote machine self-awareness. (Tweet this!) That notion springs from popular fiction, and the possibility of self-aware machines is even disputed among AI researchers. In a sense, a pocket calculator is an example of AI. It takes mathematical inputs and calculates outputs in the same way a person can make a calculation using pencil and paper. Unlike a person, however, a calculator has no agenda, future plans, or ability to incorporate wider knowledge.

In this article, we’ll explore artificial intelligence and machine learning (ML), how they are related, and applications for AI in HR.

Machine Learning: The Foundation Of Artificial Intelligence In HR

Machine learning is the basis of AI. The machine in ML refers to the code or set of instructions; learning refers to the process by which the relationships among the data inputs (such as customer characteristics and sales history in a spreadsheet) are detected. At its most basic, ML is a set of coded if-then-else instructions and mathematical formulas, such as those commonly used in Excel files and mathematics/statistics courses. The machine reads the entered data and makes calculations or sorts according to the coded instructions, or algorithms.

Algorithms are essentially a fancy name for a set of instructions for carrying out a task and making decisions. For example, the following is an algorithm for avoiding traffic tickets:

If Speed > Speed_Limit & Police_Car = Spotted, THEN Reduce_Speed, ELSE Continue

Machine learning algorithms are similar to the above if-then-else instructions.

Prediction Vs. Inference

Algorithms used in AI applications in HR are useful for both predictive and inferential purposes:

  • Prediction: Can we predict a level or outcome of something—for example, can we predict who is most likely to leave the organization?
  • Inference: What’s the relationship between a factor and its influence on an outcome? For example, what is the effect of pay level on the risk of an employee leaving the organization?

Prediction and inference both rely on a dependent variable, the dimension you want to understand change or variation in, such as employee attrition. Dependent variables are also referred to as outcome, response, or labeled variables.

Prediction and inference also require an additional variable, the predictor variable, which is something that might potentially explain change or variation in the dependent variable. For example, pay level might be a predictor variable for attrition. Predictor variables are also called independent, control, or sometimes, feature variables.

Prediction is the what of the outcome - it shows what is the likely outcome (e.g. who is likely to leave the organization), based on the set of predictors used. Inference is the why of the outcome - it shows the relationship between the predictors (such as importance, direction of effect, effect size, etc.) and the outcome (e.g. how big of an impact does pay have on risk of leaving the organization?).

Types Of Machine Learning

There are two types of machine learning:

  • Supervised learning relies on a dependent variable and predictors.
  • Unsupervised learning has predictors, but no dependent variable.

In supervised learning, we are interested in the level or outcome of a dependent variable and the algorithms examine the impact of the predictors on the dependent variable. An example would be a survey question asking employees “Do you plan to stay with the company for the next 12 months?” Responses to this can then be compared to HRIS attrition data, to see if this question may function as a predictor of future attrition.

In unsupervised learning, the machine looks for patterns and can uncover relationships that no one thought to ask about or code for—but the output requires a human “sanity check.” There may be correlations that are meaningless or useless. For example, an unsupervised learning model may reveal that there is a cluster of employees that all work at home on Tuesdays, have all been promoted within the last year, and all work in the same time zone. While this may be an interesting grouping, there’s not anything that can be done with it.

AI speeds up data analysis—but to get the answers you need, you have to ask the right questions. Learn what you should be asking with our free guide, Using Employee Survey Questions To Support A People Analytics Practice.

What can be done with AI in HR?

Both supervised and unsupervised learning models are useful in HR. Artificial intelligence can not only predict the rate of voluntary attrition; it can also reveal the factors associated with attrition, who’s most at risk of leaving, and how soon they are likely to leave. An unsupervised learning model can reveal patterns in data, such as a grouping of new managers from an underrepresented group, and who have not been given opportunities to attend the company’s internal management academy. AI can illustrate safety program impacts, to help businesses make the best decisions for their employees’ health and safety. It can also sort open-ended comments into themes, correlate them to high or low engagement, and detect sentiment or emotion of comments—as well as answer many other questions.

Most importantly, artificial intelligence in HR allows integration of insights and observations. A dashboard that integrates data for reporting results might show results pertaining to employee engagement trends.

Other dashboards can illustrate things like attrition risk, satisfaction with onboarding experiences, classifying the level of risk by job role, tenure, or other factors, and highlighting themes that correlate to employees who left or those who stayed. A dashboard displaying analysis results for open-ended comments can show the most common themes in a bar graph, or use a word cloud to illustrate relationships between themes.

The use of AI in HR not only allows more data to be collected; it allows data to be analyzed faster and can uncover insights that might otherwise go unnoticed. While artificial intelligence will probably never reach the level of sentience or self-awareness, the functions it can perform in sorting, classifying, and correlating data make it increasingly valuable to HR.

Want to see how artificial intelligence can benefit your company?

At Perceptyx, we use AI to help you uncover the insights you need to unleash your employees’ full productive potential. With custom surveys paired to our people analytics platform, we can help you capitalize on your most important resource—your employees. Get in touch and let us show you how.

Download Now: The People Analytics Playbook


We promise that we won't SPAM you.