Today, the vast majority of the advancements and applications of Artificial Intelligence that we hear about refer to a category of algorithms known as Machine Learning. Self-learning algorithms use statistics to draw models from huge amounts of data. And data include: numbers, words, images, clicks, whatever. If it can be stored digitally, it can be incorporated into an automatic learning algorithm.
Automatic Learning is the process that drives many of the services that we use today: recommendation systems like those of Netflix, YouTube and Spotify; search engines like Google and Baidu; social networks like Facebook and Twitter; vocal assistants like Siri and Alexa. And the list goes on...
In all these cases, each platform collects as much information as possible about us: which TV genres we like watching, the links we click on, the states that provoke a reaction from us, and automatic learning is able to make very precise assumptions about what we do, about the next activity we might want to do, or as in the case of a vocal assistant, about which words better correspond to the funny sounds that come out of our mouths.
In fact, this process is pretty straightforward: find the model, apply the model. And it is practically present in many aspects of our lives. And this is largely due to an invention of 1986, courtesy of Geoffrey Hinton, today known as the father of Deep Learning.
Deep Learning is an area of automatic learning in which deep neural networks are studied. It uses a technique that gives machines a better ability to find, and amplify, even the smallest models. This technique is known as "Deep Neural Network": "deep" since it has many levels of simple computational nodes that work together to search for data and deliver an end result in the form of a prediction.
Neural networks are inspired by the inner workings of the human brain. The nodes are the neurons and the net represents the brain. However, Hinton published his discovery at a time when neural networks received little consideration. No one really knew how to use them and this didn't lead to good results. It took more than thirty years for the discovery to become established. But suddenly, it emerged from the abyss.
One last thing we should explain in this introduction is that automatic (and profound) learning is available in three categories: supervised, unsupervised and reinforcement learning.
- In supervised learning, the most frequent category, data are labeled to indicate exactly which models the machine should look for. Just like a tracking dog that pursues goals once it is able to recognize the tracks it is following. This is what happens when you press "play" on a Netflix program: you're telling the algorithm to find similar programs.
- In unsupervised learning, the data have no labels. The machine looks for any model that can be found. This is like letting a person control tons of different objects and classify them into groups with similar characteristics. Unsupervised techniques are not so popular because they have less obvious applications, but they have gained strength in the field of computer security.
- Finally, there is reinforcement learning, the ultimate frontier of automatic learning. A reinforcement algorithm learns through trial and error to reach a clear goal. Try many different things and you will be rewarded or penalized depending on whether your behaviors help you or prevent you from reaching your goal. As when a child is educated: thank him with praise and affection. Reinforcement learning is at the basis of AlphaGo from Google, the program that beats the best human beings in the complex game of Go.
What can Machine Learning contribute to human capital management?
Applied to human capital management, the current use of Machine Learning is limited (although the potential for growth is broad).
In the majority of cases today, machine learning is used to lend efficiency to recruitment processes, thanks to its ability to go beyond verifiable skills, such as, the level of studies, etc. Applying machine learning to recruitment processes, which are costly and inefficient, can enable better search and discovery of the best candidates among thousands of people.
Let's explore other use cases for Machine Learning in Human Capital Management:
Obtain more qualified job applicants and level the gender playing field by applying machine learning to the development of job descriptions.
Creating gender neutral job descriptions, that is job descriptions that are relevant to a role but gender neutral in terms of pronoun application, ensures that the best possible candidates, men and women, are applying for an available role.
It may sound trivial, but a recent study conducted by Total Jobs has concluded that using gender-neutral wording attracts 42% more responses than non gender-neutral job ads.
As a result, human resources has access to a larger, more qualified pool of applicants.
Develop a more highly-skilled workforce with improved employee training recommendations.
A second example of how machine learning can improve the functions of Human Capital Management is its application to employee training programs. In many companies, employees have access to a broad range of training options, but often fail to discover those which are most well-suited for them. Machine learning algorithms are able to present internal and external courses that best fit the employee's development goals based on many variables, including the skills that the employee intends to develop and the courses carried out by other employees with similar professional goals.
Software and Artificial Intelligence are transforming the role of Human Resources
And this transformation will only become more disruptive over time.
These two cases are clear examples of how Machine Learning can elevate the role of human resources from tactical to strategic. Other contributions to this transformation are the application of intelligent softwares, which allow for repetitive actions to become automated, enable better insights regarding the existing employee make-up and potential turnover.
By applying new-age, intelligent software and artificial intelligence, companies can react and pro-act in the right time with corrective policies that curb their deficiencies, while also attracting more of the "right" talent for their specific context.
As we move toward a future where human capital and programmatic machines (i.e. robots), must work together to create value for the business, the role and quality of the people who comprise the workforce will become an increasingly more strategic asset for the business. CHRO's must face the digital transformation today in order to be well-equipped for that reality.