NPM portfolio company Conclusion has published an AI test measuring how much Dutch consumers know about artificial intelligence. The average score was 5.1 out of 10. The test, conducted among 1,000 people in the Netherlands, found that only 18 percent know that AI does not reason like a human, while just 16 percent always check whether information provided by AI is correct.
A quarter of Dutch consumers believe that AI weighs arguments in the same way humans do, while 57 percent do not know whether a language model can reason independently. At the same time, there is awareness of certain risks. For example, 68 percent know that user conversations can be used to improve AI systems, while 62 percent realise that AI-generated text cannot always be identified as such.
There is also uncertainty about responsibility when AI makes mistakes. Almost half of Dutch consumers understand that it is not always clear who is responsible when AI provides incorrect advice. More than a quarter, however, believe responsibility is always clear. In practice, responsibility can depend on the application, the parties involved and the circumstances.
“We don’t need to train everyone to become an AI expert,” said Bastiaan Sjardin, Chief Technology Officer at Conclusion. “But the basics need to be in place. Understanding that AI does not reason. Knowing that an answer can sound convincing and still be wrong. And realising that when mistakes occur, it is not automatically clear who is responsible. That knowledge is still insufficient, and the more AI becomes involved in decisions about healthcare, money and work, the greater the risk becomes.”