Artificial intelligence (AI) is applied everyday to automate and optimize processes in countless industries, recruitment included. The war for talent is now more competitive than ever, and finding suitable candidates is crucial. Using a technology like AI and its subset, Machine Learning (ML) technology for talent selection processes represents a significant competitive advantage.
How does it work?
The system has access to extensive amounts of data from multiple sources. It studies and recognizes the patterns within that data and uses them to base a statistical analysis. The evolution of the algorithm then executes tasks, without having to be programmed explicitly. ML technology gives computer systems the ability to learn ‘autonomously’, based on past statistical analyses of data. It is ideal for automating repetitive elements of a process. Applying ML to the repetitive selection processes of recruitment would allow workers to complete more tasks in less time, thus for the company to make economies of scale.
What different ways can ML be applied in selection processes?
- Job advertising
Selection processes are long, and for the most part repetitive. Manually writing and placing job adverts can a considerable amount of time that could be dedicated to other tasks.
Using ML technology to automate job advertising processes would be much more efficient. Based on the information provided by recruiters, the algorithm executes the program entirely. For example, it can find the best platform to advertise a job (job boards, social media, direct contact, recruitment firm…), and even phrase it to attract the targeted candidates, then post it.
Textio for example, is a company using data and machine learning to analyze language patterns in job adverts. Consequently, they are able to find out why some posts work where others don’t.
- CV screening
Although recruiters have access to vast networks, they lack a method to exploit them without having to invest more time and resources. A cost effective and time saving alternative to hiring more recruiters is using ML to recognize elements in data from applicants’ resumes. The program can narrow down the field of search to select the best candidates. This technology would identify keywords that correspond to experience, skills and traits corresponding to the job, automatically shortlisting applicants.
Harver is an ML software using data and science to predict a candidate’s quality. It uses factors such as culture, flexibility, soft skills and ability to succeed throughout their career to select the best fit.
- Candidate assessment and preselection
ML for assessment purposes evaluate how good of a match candidates are for companies. These AI tools are quick and easy for employers, and efficiently filter suitable candidates from others. They evaluate all kinds of aspects: skills, compatibility with company culture, personality, then shortlist the candidates checking the most boxes. Whether it is an aptitude test assessing skills or knowledge or a simple interview to learn more about the candidate, this kind of ML program can save all kinds of resources for recruiters.
Interview Mocha is an online ML skill testing platform for pre-hire screenings. Its catalogue contains assessments going from coding, personality, management skills, finance and a lot more.
- Predicting hire needs
As the baby boomer generation heads for retirement, the issue of leaving an immense skill gap arises. Therefore, there is a crucial need for companies to plan ahead on what this new generation needs if they want to remain competitive.
Ascendify is a company providing a machine learning infused talent platform. This enables hiring companies to estimate both time and cost to hire and form a candidate. This kind of platform enables companies to select the most time saving and cost-effective option to organize their selection processes and stay competitive.
Is ML a flawless technology?
Unfortunately, like most technologies, ML can malfunction and issues may arise from it. For example:
- Gender Bias
A recruiting engine used by Amazon had taught itself to favor male candidates over females. How can a computer system show bias against women? The problem actually arose from the data. The system’s data observed patterns in CVs over a 10-year period, most of which came from men. Thus, the algorithm favored men over women.
A research from Carnegie Mellon University studying Google’s ad targeting system, found serious flaws in the system. Indeed, when the engine believed a job seeker was male rather than female, it was very likely to offer ads for high-paying executive jobs.
- English first
An English company used an ML algorithm set to reject candidates with poor English, most of which were foreigners. As a result, the algorithm taught itself that common English names equaled acceptable qualifications, while ‘foreign’ names didn’t.
These biases are avoidable. The most important factor to check is the information processed by the system. One flaw in data can ruin an entire program, and result in a waste of resources, time and money.
Although revolutionary, ML systems use narrow AI, which lacks the nuances of human nature. Therefore, they are incomplete without a human recruiter’s complementarity. Such systems cannot complete the selection processes alone, but can be used in various very effective ways.