The Netherlands trusts AI, but 82% do not know that AI reasons differently from humans

Dutch people recognise the risks of AI, but overestimate what it can do and underestimate where it fails

Only 18 per cent of people in the Netherlands understand that AI cannot reason in the same way humans can. The rest either believe it can or simply do not know. That is one of the findings from our AI assessment among 1,000 Dutch citizens, on which the Netherlands scores an average of just 5.1 out of 10. People in the Netherlands are becoming increasingly aware of the risks associated with AI. Yet persistent misconceptions remain about what AI can and cannot do. For example, more than a quarter (26 per cent) believe it is always clear who is responsible when AI makes a mistake, while in practice this is rarely the case.

AI writes cover letters, answers medical questions and helps people make financial decisions. One in four Dutch citizens (25 per cent) believes that AI weighs arguments in the same way a human does. In reality, AI systems such as ChatGPT predict the most likely next word based on patterns in language. They do not understand what they are writing.Even so, 57 per cent of respondents do not know whether a large language model (LLM) can reason independently. People who believe AI can reason are more likely to accept its answers at face value. For everyday questions, this is rarely a problem. However, as AI is increasingly used to support decisions that carry real consequences, the risks grow.

It is not the case that people in the Netherlands fail to recognise the risks of AI. Some 68 per cent know that AI companies use user conversations to improve their systems. Sixty-two per cent understand that AI-generated text is not always identifiable as such. And 51 per cent recognise that AI can conduct a phone conversation without people realising they are speaking to a computer. Knowledge about what AI can do is increasing. Yet that knowledge does not automatically translate into more cautious behaviour: only 16 per cent of respondents always verify whether information provided by AI is correct.

Bastiaan Sjardin, Chief Technology Officer at Conclusion: "It is not surprising that a quarter of the Netherlands believes AI makes deliberate judgements in the same way humans do. But a language model does not weigh anything up. It learns from patterns in data and uses probabilities to predict the next word in a sentence. That is impressive, but it is not human reasoning. Once we recognise that reality, we can use the technology for what it truly is: a useful assistant, while we remain firmly in control of the decisions that matter most."

Imagine an AI system generates incorrect advice with financial or medical consequences. Who is responsible? Almost half of Dutch citizens (48 per cent) recognise that this is not always clear. More than a quarter (26 per cent), however, believe responsibility is always obvious. Determining accountability for AI-related errors can be complex and depends on the application, the parties involved and the specific circumstances. European legislation is providing an increasing number of frameworks in this area, although practical implementation is still evolving.

The research paints a picture of a country that is making AI an increasing part of everyday life without fully understanding how it works. "We do not need everyone to become an AI expert," says Bastiaan. "But people do need a solid grasp of the basics. They need to understand that AI does not reason. They need to know that an answer can sound convincing and still be wrong. And they need to realise that when mistakes happen, responsibility is not always immediately clear. That understanding is still lacking. And the more AI influences decisions about healthcare, money and work, the greater the risks become."