Job application algorithms

Research commissioned by the Netherlands Institute for Human Rights shows that almost all employers use job application algorithms (see the figures below).

To give an idea of how those algorithms influence your chances, we will follow the story of Salma and Richard.

To handle the amount of job applications for the position of shop assistant, the retail store uses various digital means to facilitate the recruitment process.

The store places the job application online. A computer system helps with the first selection of candidates. Moreover, candidates are asked to play games and undergo assessments online, so that the company can assess their competences.

The company aims to find suitable candidates for the new position in the most efficient way. However, these digital means can also lead to discrimination or exclusion, even if unintentionally.

Swipe further through Salma and Richard’s job application process to see how this can happen.

Figures

• 96% of employers post job vacancies via social media or online job-posting websites
• 71% of employers look for candidates on social media

Selection and assessment

• 12% of employers use algorithms to select candidates
• 12% of employers use algorithms to assess candidates

‘There are 0 positions’

Job vacancies don’t always find you

The algorithms used by online platforms such as LinkedIn, Google and Facebook can influence your opportunities. Algorithms can contain (unintentional) bias, which means that they prefer candidates from certain target groups. On the basis of your behaviour on a website (such as clickstream data), as well as the behaviour of other people like you, algorithms make an estimate of which job vacancies would interest you and, therefore, you will be exposed to these targeted advertisements more frequently.

In February 2025, the Netherlands Institute for Human Rights ruled that Facebook discriminates when offering job advertisements. Facebook shows stereotypical male jobs such as mechanic mainly to men. And stereotypical female jobs such as receptionist mainly to women.

You don’t always find job vacancies

Not only do job vacancies not always find you, the opposite may also be the case: you don’t always find all the job vacancies. Research commissioned by the Netherlands Institute for Human Rights shows that algorithms sometimes do not match a search with gendered key words to non-gender-neutral job titles on online job-boards. In Dutch for instance, professions often have a gender such as ‘lerares’ (a female teacher) and ‘timmerman’ (a male woodworker). Think for instance of actor and actress in English. If you type in a gendered search-term, you will not always find the vacancies with gender-neutral descriptions, even though gender is not relevant for the job. In our story, Salma types ‘winkelmedewerkster’ (female shop assistant) in the search bar, but does not find the positions for 'winkelmedewerker' (the gender neutral term). ‘Unfortunately’, the website indicates that 'there are 0 positions that match your search'. Luckily Richard has tipped Salma about his search results.

Game over

The language and requirements used in a vacancy may discourage some groups more than others to apply. Research shows, for example, that on average men are more likely to find themselves qualified for a role on a higher level than women, even though they have the same qualifications. This can lead to women being less likely to apply for vacancies with strict requirements.

Furthermore, certain use of language such as ‘a true leader’ mainly attracts men, while older jobseekers tend to give up when they see terms that seem to point to young candidates like ‘Join our enthusiastic young team!’.

In addition, it is important that the procedure is easy to understand for everyone. Increasingly, companies are testing candidates’ competences with online assessments or games. But these tests are not well suited for everyone. Due to his dyslexia, Richard found the assessment and the game more challenging. Also, people who have impaired hearing or vision, or those who have difficulties with using digital means for other reasons, may find such games or assessments to be a hurdle.

Fitting the picture

Salma has been asked to present herself in a video. She thinks that a person will assess the video, but that is not the case. A computer analyses Salma’s expressions and then provides a score. This score is based on all kinds of information Salma is not aware off, such as micro-expressions in her face. The position of her eyebrows may give an indication of the level to which a person is extravert.

In the end, the computer is unable to assess Salma properly. That may be caused by her dark skin or because she has a slight accent.

Automatic image and text analysis may work less well for some target groups because the data with which the software is trained is not representative enough. The computer then has “seen and heard” too few examples of people with a dark skin, women and/or people with a slight accent to be able to recognise them properly. This can lead to certain groups being (unintentionally) excluded by the system.

Unintentional exclusion

This company’s computer is looking for "the best candidates” by scanning the CV, the motivational letter, and by looking at the results of the online game assessment and video job application. This software compares the CV and the motivational letter with those of both the successful and the non-successful candidates who applied for jobs with the company in the past.

The computer gives Salma fewer points on account of her CV, because Salma has two gaps in her CV. The first time because she stopped work temporarily when she gave birth to a child and the second time when she needed to recover from a car accident. But the computer does not know that.

A gap in someone’s CV seems to be relevant when assessing candidates, but in practice the computer will more often indirectly reject people who have had a child or who have been ill.

The computer knows best

People tend to believe what computers say. We call this “automation bias”. But computers may have it wrong, as we have seen above. Moreover, what makes things even more challenging is that mistakes by computers can be very difficult to uncover, particularly for you as a jobseeker. How could Salma and Richard possibly have known what happened behind the scenes? Moreover, how could a recruiter have known?

Nevertheless, an employer should always be able to explain to you why you were rejected. When an employer is unable to explain how the selection process of job applicants is carried out, then this can add to the legal assumption that discrimination took place. Even if that selection process is done by the computer.