When AI becomes routine: Which research skills need to be preserved?

As the use of AI in businesses grows, one question is becoming increasingly important: is it enough simply to ensure that AI systems meet legal and technical requirements? The EU research project AIOLIA argues that responsible AI use needs to go further. Alongside data protection and transparency, AI ethics also plays a crucial role.

One of the ethical challenges highlighted in the report is the impact of AI on human cognition, particularly in professional settings. The more frequently employees delegate tasks to AI systems, the more important it becomes to ask which skills they will continue to develop themselves over the long term (“non-skilling”) and which existing skills they may lose (“deskilling”). “Where AI systems work well, they shift the balance between human judgement and algorithmic outputs in ways that become more pronounced over time.” For example, AI can influence which lines of reasoning are considered worth pursuing and which conclusions appear convincing – an issue the AI Act was never designed to address, Euractiv writes.

Both the European Parliament and the European Commission explicitly identify deskilling as a risk in their respective AI papers and propose upskilling and reskilling as ways to respond to technological change.

At the same time, the AIOLIA report identifies a trade-off when it comes to checking AI-generated results. If every piece of information is fully verified manually, many of the efficiency gains are lost. If, on the other hand, results are accepted largely without scrutiny, the risk of errors increases and the underlying information becomes harder to trace. “The trade-off […] therefore depends on the specific context in which AI is used and on the nature of the decisions it supports.”

What does this mean for the use of AI in professional research?

AI should complement human research skills rather than gradually replace them. What matters, therefore, is not only whether an AI system produces accurate results, but also how sustained use affects source evaluation, subject expertise and independent judgement.

For companies, this means going beyond simply requiring human oversight as a formal safeguard. They need to define which tasks and decisions require particularly rigorous review. AI literacy therefore involves more than knowing how to write good prompts or use a particular tool. Employees need to be able to judge when AI is appropriate, when additional sources are needed and when human expertise should take precedence.

By Maria Kleiner