How might we improve the recruiting process using Artificial Intelligence? (Intro)
If you have no network, if you get out of prison, if you have no referral, if you are a veteran, if you are a freshly graduated student, if you are not cis-male, if you are old, if you are not white, if you do not belong to a specialized category, job seeking gives a hard time. The right to work and the protection against unemployment is considered as a fundamental human right (Universal Declaration of Human Rights, article 23). Unemployment does not only has an economic dimension, but it also has a psychological one. In the collective unconsciousness, unemployment is synonymous with poverty, social and psychological exclusion (Sen, 1997) and sometimes even psychological disorders (Jahoda, Lazarsfeld et Zeisel, 1933; Eisenberg et Lazarsfeld, 1938; Bakke, 1940; Hill, 1977). Unemployment harms motivation which often leads to resignation (Sen, 1997).
Two definitions of artificial intelligence
‘Artificial intelligence’ covers two meanings: first, it refers to a machine’s ability to reproduce human cognitive processes. Second, ‘artificial intelligence’ refers to the family of algorithms whose processes and results are comparable to those found in human cognition. In other words, there is a difference between being intelligent and seeming intelligent. Coming back to the sources of theoretical writings on AI help us realize the deep understanding of the meanings of the two words taken and thought differently: ‘artificial’ on the one side, really denotes something un-natural, and may be false, and ‘intelligence’ on the other side was a concept much detailed by psychology and sociology, but also taken in the sense of “data-driven knowledge” as in the military. Today, I believe the public hears and fears AI in its first meaning. It is either perceived as a magical solution or a massive destruction weapon. I asked myself: what if AI was the magic solution for solving my recruiting problem? But, knowing that data scientists do not understand AI in its original meaning anymore (the ability to reproduce human cognitive processes), what is the reality of AI in the second meaning and how can AI help make recruiting processes more efficient?
Designing Seqy: a prospective product
I started designing a platform named “Seqy,” where candidates would finally have their dreams fulfilled.
Personae and User Journeys
Personae profiles are informed summary of the mindset, needs and goals typically held by key stakeholders (Luma Institute, 2012). Using the Design Thinking frameworks of “Personae” (fig.1) and “User journeys,” we imagined five user journeys: one for four candidates (fig.2) and one for a recruiter (fig.3). Because the “recruiting process” is broad and complex, I built them alone but relying on a 212-companies benchmark.
It is essential to keep these two paths in mind while reading the rest of this essay, as a grid for evaluating the success or the failure of the product designed.
Personae help us realize what it takes for each application with a personalized cover letter. Personae are flesh and bones; they need to buy food, work to pay the rent. They work 35–40 hours per week, and after one or two hours a day commuting, they come home, where they can eventually connect to a computer. Maybe they have food. And maybe they have one hour left for applying during their free time during which they would have wanted to relax because the job is exhausting. One application takes between thirty minutes and one hour, depending on the web platform. In this context, recruiters expect candidates to find their job offer, write a letter, and modify their CV.
Simplicity is poor design; complexity is key
Instead of asking the candidate twice quite the same info: where does the person live? Where does he want to live? Where would he accept to relocate? Most interfaces assume that collecting the entity “location” is enough to cover the three cases mentioned above. As a result, the information is most often only asked once: “Where do you live?”, And the candidate has nowhere else but the cover letter to explain a will for relocation. However, no one reads this cover letter, and no one hires the candidate — not to mention the vicious circle where before being able to relocate, by getting a place to live, every real-estate agency will ask for payslips.
Interpretation framework: About portraits and imagination
Recruiting someone consists in selecting someone and rejecting someone else, based on the interpretation gotten from reading an autoportrait (CV, resumes, cover letters). In other words, it is a matter of what the recruiter projects in his/her imagination about a specific truncated vision of the person. Can User Experience Design be transparent enough to explain this aspect both to the applicant and the recruiter? How to avoid the side effect of the candidates feeling judged for who they are when they do not get the position? How do we design an AI-based solution where the difference between causality and co-occurrence is clear enough to avoid misinterpretations?
It’s a match!
An analogy can be made with online dating. Writing an application to someone you don’t know, is pretty similar to online dating: when descriptions are too general, they give no chance to draw a mental portrait; when the text is too long, depending on the reader’s energy, it is not going to be read; and when descriptions are too specific, the picture is so clear you might not want to meet the person, fearing there would be nothing more. The parallel between recruiting and online dating is worthwhile, especially for User Experience Design (UX). Indeed, matchmaking applications (on mobile and websites) benefit from much more investments than job boards. This is the reason why we took a closer look at online dating platforms as a source of inspiration. Indeed, successful platforms from MatchGroup, like OKCupid, are very good at enriching existing profiles with fresh data, with statistical methods used after data collecting.
Can we quantify qualitative data?
When thinking about what artificial intelligence can do (e.g., calculus) and cannot do (e.g., understand innuendos, be self-conscious, remember things), I realized that the biggest question is: how can we quantify qualitative info without altering it? In other words, applied to the recruitment problem, can we write a CV that is as good as a cover letter and can we encode it so that the computer could portrait the person accurately? Language seems to be a hot topic when it comes to interpreting natural language of short descriptions, cover letters, publications, social media profiles, job descriptions, etc. so NLP will be at the core of our exploration. Candidates want to be considered and read, not based on their CV alone, but also their story, their sense of humor, their ability to answer problems, their agility, their strengths and weaknesses, their voice, etc. In an enterprise to build an AI in charge of the recruiting process, it is the whole dimensions of candidates that has to be taken into account.
Leverage public data to enrich profiles
The problem of the truncated portrait is complex. First, recruiters do not have the time to read more information; second, the candidates do not have much time to allocate to provide more details either. A recurring candidate pain point is to think: “With all the time I spent on the Internet, creating profiles, writing blogs, codes, etc. How could a company still ask me to spend the time to describe myself? My abilities? etc. ?” How can we optimize the use of candidates’ data to offer them the ideal job? The recruiting field is a big data issue because, in exchange of an economic situation, people are readily willing to share facts about themselves in their CVs. They share their professional history, their interests, the places where they have been, the languages they know, their hard skills and sometimes their soft skills. Some of them are also very transparent on other platforms throughout the internet in their social media profiles, blogs, pages, websites, publications. Crossing those data should allow anyone to have a complete first sketch of anyone.
The problem
How could AI be an enabler to solve the difficulties in recruitment? “Seqy” is the name of our solution to: “How might we improve the recruiting process using Artificial Intelligence?”
To answer this question, we will follow the steps described hereafter:
Aim 1
Understand what goes wrong for candidates and for recruiters to understand what they would perceive as a solution to their problems, a way to improve the process, in the sense of making it a better experience. We will explore issues such as time lost, frustration, oversimplification, bureaucracy, inhuman tech.
Aim 2
Understand existing solutions on the market and in the academic world.
Aim 3
Detail the functionalities of the online platform “Seqy” that follows these guidelines: 1. save time and boosts productivity; 2. better quality results; 3. Less discrimination; 4. Less money spent on data collecting. Offer an alternative to the “CV” linear way of thinking which leads to an absurd hyper-specialization. Capitalize on the concept of “Learning curves.”
Aim 4
Study feasibility, desirability and business viability of the retained solutions.
