Machine Learning Applied to Online Recruiting
The article Application ofMachine Learning Algorithms to an online Recruitment System is an interesting paper that describes a Machine Learning (ML) methodology to make the hiring process more effective. I thought it could be a good example to talk about ML and reflect on the massive selection processes we face today.
In a previous post I discussed the opportunities and challenges of applying Machine Learning (ML) to different areas within Human Resources.
The article is in line with this idea. I will use quotation marks to highlight when I quote the paper.
Let us start with the objective: “The paper describes a novel approach for evaluating job applicants in online recruitment systems leveraging Machine Learning.”
It is evident that E-recruitment systems that automate the candidate screening process can be useful in the internet era where companies receive a massive number of CVs.
In short, “The proposed system extracts a set of objective criteria from the applicants’ LinkedIn profile, and infers their personality characteristics using linguistic analysis on their blog posts.”
So, the system gives the candidates the option to use the LinkedIn account so it can automatically extract the user´s information of their profile. Then, it applies linguistic analysis to the candidate´s blog. The text analysis in simple terms uses a dictionary of words classified in categories and counts the relative frequencies of words that fall into eac category. Finally, the system combines the two former criteria to derive the candidate´s relevance score for the position through supervised learning algorithms.
Supervised ML is what I expected in this context. ML helps build the selection model that would otherwise be very complex to design. But of course, “This approach requires sufficient training data as an input, which consist of previous candidate selection decisions.”
The system was tested in a scenario with 100 job applicants and “Our system was found to perform consistently compared to human recruiters.”
First of all, I would like to congratulate the authors of the paper. I think it is clear and relevant. It shows the gains in efficiency that ML can bring while being realistic in the scope where it can be used. After finalizing the process, the top candidates are interviewed. So, it is aimed at facilitating the job of recruiters, not replacing them as we mentioned before.
As food for thought, you may think how important social media and LinkedIn content are becoming to be included in candidate pools.
Looking at the detail of the study, it focuses on 4 selection criteria, namely: Education, Work Experience, Loyalty (average number of years spent per job) and Extraversion.
The system outputs a ranking of the candidates from the selection criteria. It is interesting to note that the variables are defined as continuous (e.g. Number of years of experience) or Yes/no type so they can be input in the machine.
Let us analyze one of the positions the article suggests, a junior programmer position that required programming skills in C++ or Java development languages. Junior programmers were mainly judged by education and loyalty (because a company would not invest in training an individual prone to changing positions frequently). Loyalty is mainly measured by the average number of years spent in a job.
This is a good example of supervised ML where the design phase is critical. Experts with functional knowledge need to work with the AI specialist to build a model aligned to the business objectives. Once the model is defined, the machine will objectively apply the criteria.
I thought the above-mentioned paper would be an interesting example to share. AI will bring to the surface the criteria senior recruiters use to screen résumés. In this case, the way loyalty is measured and formalized into the model reinforces the idea that the number of years you stay in a job is meaningful in terms of building up your professional profile.
The Career Cycles argument seems to be supported by the paper. We are starting to put numbers behind theories and HR criteria. Talent management is becoming more analytical, indeed.