Using artificial intelligence

Explanation for matching Candidates with Client jobs by using artificial intelligence

Robert Half (“RH”) applies artificial intelligence (“AI”) to filter, also on personal aspects, and match Candidates with specifics of positions in the interest and for the purpose of finding suitable positions for you. For this, RH is using a service offered by Textkernel (“TK”), a company based in the Netherlands / EU.

In general AI describes computer systems that are able to perform tasks normally requiring human intelligence, such as self-driving-cars or translation between languages. For this, the involved data processing activities can be simplified as follows:

'Candidates' and 'Clients' job data gets passed to TK AI by RH

 

TK AI reads the data and determines which data is relevant for extraction.

 

Extracted data is classified by TK. No machine learning takes place.

 

Following extraction and classification, data is processed by an algorithm developed by TK to match candidates with the specifics of Client jobs. No machine learning takes place.

 

The result of the matching query get displayed by a 9 digit score to RH.

 

RH will exercise their professional judgement before passing candidates details to a Client.

For a more detailed overview regarding the processing activities involved, please read the following description:

(1) At first, RH pass a combination of candidates’ personal data which is already in the RH database and candidate resumes (CVs), together with the specifics of Clients’ job data to TK AI. The following personal data gets transferred to TK AI:

  • Recent Job titles
  • Recent Profession
  • Location (countrycode + city/postalcode)
  • Education Level
  • Language Skills
  • IT Skills
  • Total Years of Experience
  • Majors/Degrees
  • Fulltext keywords

(2) After TK AI has received the candidate personal data and the specifics of the Clients’ job data from RH, the AI reviews the data and determines which data is relevant for extracting, so that candidates can be matched with the specifics of Client jobs.

(3) The extracted data is then classified. This process takes into account that people use different expressions or words for things that have the same or similar meaning. Following extraction and classification the data is then processed by an algorithm developed by TK to match candidates with the specifics of Clients’ jobs. The involved logic is shaped by the following criteria:

Candidate data / personal data considered Job data considered Example of Condition applied (Conditions will vary according to the specifics of each Client job)
Recent Job titles Position Should Have
Recent Profession (normalised) Profession Nice to Have
Location (countrycode + city/postalcode) Location Must Have
Education Level (normalised) Education Level Nice to Have
Workfield Workfield Must Have
Language Skills (normalised) Language Skills Nice to Have
IT Skills IT Skills Nice to Have
Total Years of Experience Years of Experience Nice to Have
Relevant Keywords Relevant keywords from Client’s job that further improve match results e.g. Professional Qualification, Degree) Nice to Have
Fulltext keywords Fulltext Automatic (depending on keyword)

Explanation of Conditions applied:

The TK AI match logic includes certain Conditions, which is explained below:

  • Must Have
  • Is effectively a filter which will filter out every result that does not match. For example, a Must Have term of “Location” as a Location will filter out all candidates that do not have specific Location from the match results.

  • Should Have
  • Influences the ranking by placing matching candidates higher on the list of suitable candidates for a Client job. The ‘Should Have’ behaviour is identical to ‘Nice to Have’ (see below) except that ‘Should Have’ conditions carry more weight when determining the ranked order of the match results.

  • Nice to Have
  • Influences the ranking by placing matching candidates higher on the list of suitable candidates for a Client job, but without filtering non-matching results out of the list.

(4) The matched and ranked candidates are transferred back to RH database. The result of the matching query gets displayed by a 9 digit score that is assigned to a candidate in respect of their match with a specific Client job. This score determines the ranking of that candidate’s suitability/fit for the Client job in question. The degree of separation between candidates, facilitated by the 9 digit scoring, means that it is highly unlikely that more than one candidate would receive an identical score in respect of the match against a specific Client job.

  • Example 9 digit score: 0.64231317

(5) The following measures are in place to prevent/detect bias results:

  • The match is based only on key factual data (e.g. job title, location, skills, relevant keywords). No interpretations or other data influence the matches.
  • In particular, no discriminative data (e.g. gender, nationality) will influence the matching process. Terms/expressions that may make a client job more attractive to certain demographics are ignored.
  • Gender in job titles can be derived for certain languages (e.g. French, German). RH’s matches are not influenced by masculine/feminine forms. Gender is ignored when matching and ranking proposed candidates.
  • RH’s solution is additionally configured to reduce bias as follows:

    • Longer resumes (CVs) are as relevant as shorter ones.
    • The level of a skill is more important than how many times the skill is mentioned in the resume (CV).
    • Recent job titles and skills are more important than older ones.

(6) Based on the result of the matching query, RH recruitment consultants generally review the factual suitability of the proposed candidates suggested by TK AI and decide if some, all or none of the TK AI proposed candidates are suitable. The candidates that RH recruitment consultants decide are the best match for a specific Client’s job, may be included on the final shortlist of candidates (and CVs) submitted to the Client for the Client’s job.

(7) On receipt of the shortlisted candidates (and CVs), the Client (through its hiring team) selects the candidates they want to interview (if any). At the end of the Client’s interview process, the Client (through its hiring team) makes the final decision which may be that none of the candidates are suitable for the Client’s job, or to select the candidate that is suitable for the Client’s job and to hire/engage/employ the selected candidate.

The use of TK AI has no consequence or impact, especially not on your application or on finding suitable positions for you because RH recruitment consultants use their expertise, personal knowledge of you, the Client and its business, the specifics of the Client’s job, and the Client requirements for the role as well as the job market and candidate market at any time and the preparation of the candidate shortlist. The final decision to select candidates for interview and ultimately to hire/engage/employ a candidate for a specific job, is made by the Client (through its hiring team), not by the TK AI.