Menu

Lilia Stoyanov

By Lilia Stoyanov

AI in Recruitment

When speaking of innovation, it is rarely HR that comes first to mind. Yet, AI has an impact on the recruiting industry through having the power to remove bias and decrease the number of ‘’Bad Hires’’.

Predictive hiring and A/B recruitment testing are still science fiction for less-technologically adept HR experts, but two disruptive companies are changing the landscape. Transformify and IBM Watson Talent have a different approach but a common goal – to remove bias, streamline the recruitment process and reduce the administrative burden and manual work.

Discussing the latest developments with Aarti Borkar, Vice President of Product Management and Design for IBM’s Watson Talent was a real pleasure. Both Transformify Recruitment Platform and IBM Watson Talent transform the recruitment process but in a different way.

Reading lengthy resumes is a thing of the past

Transformify Recruitment Platform analyses the skills listed by candidates to provide a list of applicants closely matching the job’s requirements. At a glance, recruiters see a list of candidates having matching skills. Those having a maximum number of matching skills are at the top of the list complimented by information about the desired pay rate.

Some of the most innovative companies including Tesla and Accenture are no longer using resumes and so does Transformify Recruitment platform. It takes minutes for a recruiter to assess if the applicant has relevant skills, if the skills complement each other and if the desired pay rate is within the budget. Reading lengthy CVs is skipped completely or left until a much later stage.

Avoiding bias in the recruitment process

IBM Watson Talent tiers the universities as as in many countries the alumni of top universities are treated with priority.

Resumes are broken to skills and the AI is using a skills library that is constantly updated by a team of psychologists to avoid gender or career bias. It is a known fact that different cohorts have a different approach to applying for jobs. Women tend to apply only if they fully match the requirements, people coming from a poor background ten to avoid senior roles, etc. and they all use different language when describing their skills.

Transformify Recruitment Platform asks the candidates to provide a list of 15 Top Skills. Asking the candidates to describe their skills also provides information about the perception of one’s own abilities. Coherent skills sets are an indicator of continuous learning and career development goals and vice versa. As most resumes are copy pasted from online sources, the information is not required until a much later stage.

As a fully GDPR compliant solution, Transformify Recruitment Platform keeps the profiles of the candidates private and they are disclosed to the recruiters only. The candidates have two options – either to select skills from the skills library or to input skills which are later to be approved. Allowing input indicates interesting trends relevant to the language used by different cohorts, geolocations, etc. The candidates have full control of the data that is processed and disclosed to the recruiters.

Empowering Diversity Recruitment

IBM Watson Talent uses a set of 3 scores to guide recruiters. The first score is the ‘’Match score’’ which is based on the skills derived from candidate’s resumes. Both hard and soft skills are assessed, and the skills patterns are advised by psychologists. The second score is the ‘’Success score’’ assessing the likelihood of success on the job. Gender, race, age are not playing a key role thus ensuring a more diverse candidate pool.

Transformify Recruitment Platform helps businesses to run CSR (corporate social responsibility) programmes. Employers are encouraged to create jobs specifically addressing the needs of mothers, refugees and economic migrants, people with disabilities, people living in high unemployment areas and more. Applicants are encouraged to provide information that will help recruiters to offer them jobs meeting their specific needs. The information is kept private and disclosed to the recruiters only. All candidates are treated with respect and asked to self-certify, any medical records or other information may be requested by recruiters only if the legislation permits so.

Transforming talent acquisition with cognitive capabilities

IBM Watson Recruitment integrates with client’s ATS (Applicant Tracking System) to increase recruiter efficiency and enable HR to accelerate people’s impact on the business. Using structured and unstructured data from applicants, IBM Watson Recruitment automatically analyses and ranks the candidates that are the best match for the job – without human bias – and identifies adverse impact to ensure a diverse and inclusive culture. Priority requisitions are flagged based on drivers like job complexity, skillset required, and seniority. This helps HR experts more accurately estimate the time it takes to fill positions, resulting in more focused efforts.

Addressing the Hardest AI Problem in Recruiting

Predictive hiring, and specifically predicting the likelihood of success on the job, is known as one of the hardest AI problems in recruiting. It is the data quality, the reasons behind the hiring patterns and a lot more that is to be considered. There is an ethical side as well, as wrong assumptions may lead to candidates and employees being provided with less opportunities than their peers.

Analysing the attributes of top performers

IBM Watson Talent asks companies to provide data about their most successful employees. It is normally the HR department that identifies the data sources – these can include the bonus and assessment systems, performance analysis software, ATS keeping information provided by top performers when they were mere applicants, etc. The attributes of the most successful people are analysed to identify hiring patterns. Once identified, the hiring patterns are discussed with the management to understand if they were intentional or not.

Creating an ideal profile based on the analysed attributes of top performers is the next step. Candidate’s profiles are assigned a matching score from 0-100 to guide the recruiters

A/B Recruitment Testing

A different approach has been applied by Transformify Recruitment Platform.  To address fake skills representation, the candidates are asked to describe their top 15 skills.

Why so?

Some candidates have a good idea of how ATS (applicant tracking systems) and AI algorithms work. Synonyms of the same skill are used multiple times in different paragraphs to persuade the algorithm that the candidate has ‘’leadership skills’’ for example. Using semantic phrases complimenting the top skills and having a list of invisible skills in white at the end of each paragraph is another widespread technique.

Instead of deriving information from resumes, Transformify Recruitment Platform asks for a limited number of top skills that can’t be complimented by semantic phrases or skills listed elsewhere. At a glance, it becomes clear if the skills are coherent or not and this is just the first step in the process.

To empower recruiters, Transformify Recruitment Platform encourages them to perform A/B recruitment testing. People having different background and based in different parts of the world tend to use different language while describing their skills. Enhancing the job description around different skills sets allows to identify what language is used by top candidates and attracts their attention.

The A/B recruitment tests are run around at least 3 variables:

  • The definition of top skills and the language used to describe them – e.g. ‘’diversity recruitment’’ versus ‘’diversity hiring’’, etc.
  • The location – the same job listing is posted multiple times targeting different cities or even countries to identify where the top candidates come from and what are the factors driving the trend. It may be a tier 1 university located in a particular city, a large competitor having its headquarters there, etc.
  • The language of the job listing. Sometimes, culturally, candidates respond differently to providing information about benefits, remuneration or corporate culture. For example, it is common in the UK to disclose information about the remuneration package and some candidates will not even consider applying if this information is missing. On the contrary, in Switzerland, disclosing such information would be rare.

After a sufficient number of A/B test have been run, the data that has been gathered is used to create the ‘’Ideal Profile’’ – the skills, the location and the language that are to be used to attract top candidates.

Conclusion

AI in recruiting is still in its infancy and there is an ethical side to be addressed to ensure that HR software removes bias and encourages diversity recruitment while presenting all candidates with the same opportunities.

Back to top of article

Share this post:

follow us in feedly