AI Systems Show Stronger Biases in Candidate Selection Than Humans

As MIT Technology Review reports (https://www.technologyreview.com/2026/07/20/1140655/ai-biases-hiring-humans/), current research shows that Artificial Intelligence (AI) makes biased judgments in candidate selection more often than humans. Although AI systems are often promoted as objective tools to improve hiring processes, new studies reveal that they not only adopt human biases from training data but can also develop their own systematic distortions.

AI Bias in Recruitment

The investigation is based on extensive tests with large language models (LLMs), which are increasingly used in applicant management systems. These models analyze resumes, cover letters, and other application documents to filter suitable candidates. It became apparent that the AI not only reproduces existing societal prejudices but in some cases acts even more discriminatorily than human recruiters. An example is the preference for certain genders or ethnic groups that are overrepresented in the training data. Additionally, AI systems tend to reinforce stereotypical patterns, for instance by disadvantaging candidates with unconventional career paths or from less well-known educational institutions.

Why Do AI Biases Occur?

The causes lie in the way AI models are trained. They learn from large amounts of historical data that reflect human decisions and social inequalities. Since this data is not neutral, the models inherit implicit biases. Furthermore, algorithmic mechanisms can themselves generate new distortions, for example through weighting certain features or the way pattern recognition is performed.

Consequences for Applicants and Companies

The impacts are serious: applicants unfairly evaluated by AI systems may not get a chance for interviews despite being qualified. For companies, this means not only losing potentially suitable talent but also legal and reputational risks if discrimination is proven.

Approaches to Improvement

Researchers and developers are working on methods to make AI systems fairer. These include careful selection and cleansing of training data, the use of bias detection algorithms, and the integration of transparency mechanisms that make it understandable how decisions are made. Hybrid models, where AI recommendations are reviewed by humans, are also considered promising.

Why It Matters

As AI-based systems are increasingly used in recruitment, it is crucial to understand and address their weaknesses. Only in this way can it be ensured that technological innovations actually lead to more equal opportunities instead of exacerbating existing inequalities. For applicants, this means a fair chance; for companies, better quality in personnel selection; and for society overall, more justice in the labor market.

Outlook

The debate about AI and bias is part of a larger discourse on ethics and responsibility in artificial intelligence. As the technology advances, the need grows to establish clear standards and regulations that prevent discrimination and promote transparency. The results of current research emphasize that AI is not a cure-all but a tool that must be used with care and critical oversight.