Ready to be interviewed by AI the next time you apply for a job?

Job interviews are increasingly being driven by AI. According to the World Economic Forum, 90 percent of employers already use some form of automated screening system to rank applications. A full 88 percent of companies also use some form of AI to screen applications. Larger international companies in particular say they are saving time and money this way.

But they miss the fact that technology also causes candidates to change their behavior and the way they write applications. A comprehensive international study shows that when candidates believe that AI is assessing them, they emphasize their analytical abilities, for example. At the same time, they downplay other important human qualities such as empathy, collaboration, or creativity. But when the candidate systematically gives a distorted image of himself in this way, the recruitment process is also distorted, which affects the algorithm that controls the AI. Because all data collection is based on statistical techniques, which are more or less relevant.

The problems are several: Algorithms are already often accused of systematically distorting the selection process, for example, women are often disadvantaged. In their CVs, they like to emphasize their interpersonal advantages over how analytical they are. For example, when the American company Amazon began using AI to scan applications, the algorithm was trained on mainly male applicants and historical data. The consequence was that female candidates were disadvantaged.

Here's the irony: Today's automated systems are marketed as a way to create equality in the selection process. Now, instead, there is a great risk that the algorithms will create a system that distorts selection - albeit in a different way. The problems can be summarized as follows:

1. Today's algorithms are developed by software engineers. The systems they develop make decisions based on probabilities. But they lack psychometric knowledge, which also takes into account legal and ethical aspects to evaluate a selection system.

   
   

2. The customer's needs and environment may change in an unforeseen way, which the algorithm does not take into account, since the models are based on historical data and therefore risk recreating old patterns and systematic errors.

   
   

3. The complexity of the algorithm makes it difficult to determine whether – and why – it makes mistakes. Automated selection systems are only as good as the data they use. The risk now is that the systems no longer measure the candidate’s genuine abilities. Instead, data is collected based on what candidates believe the automated systems value most. The consequence is that selection is unintentionally homogenized and the diversity that so many companies strive for disappears.

AI in itself does not automatically mean that the selection process will be fairer. Instead, AI can just as easily amplify existing problems. Today's AI solutions lack transparency and are difficult to understand. At the same time, test providers only refer to the "algorithm" that calculates a result. For the customer, it is difficult to understand why a candidate is ranked in a certain way.

AI is neither a miracle solution nor a scam. It is a tool. Used correctly, the technology can improve the recruitment process. Used incorrectly, AI risks making it worse – but faster. The Swedish AI regulation will be fully introduced in the fall of 2026. It will be interesting to see which companies will be the first to be accused of discrimination and how they handle candidates who say they have been disadvantaged due to the design of the algorithm!

Sources:

How AI assessment tools affect job candidates´ behavior. Harvard Business Review, March-April 2026.

AI in hiring: More research required. Monitor in Psychology. January/February 2023.

Guide to manage the risk when machine learning goes of the rails.Harvard Business Review, January-February 2021.