Netherlands residents have little trust in ChatGPT; Trust in algorithms plummeting

Netherlands residents have little trust in algorithms, and that trust has been declining sharply for five years, according to a study by KPMG Netherlands. Trust in generative AI like ChatGPT is also low, with only 21 percent of Netherlands residents saying they believe the answers provided by the AI are correct.

The researchers found that more knowledge of algorithms leads to less trust. “We see that people who are familiar with algorithms are less likely to trust them than people who are not familiar with them or have not yet formed an opinion,” Frank van Praat of KPMG’s Responsible AI team said. About three-quarters of Dutch people know what an algorithm is, and young, theoretically trained people are most familiar with its application.

For example, 79 percent of people who are familiar with algorithms understand that the use of ChatGPT and other generative AIs entails risks. For people unfamiliar with algorithms, that was only 50 percent. Despite that, only 39 percent of Netherlands residents check the answers they get from ChatGPT.

Confidence in algorithms as a whole fell 22 percent this year compared to last year, which is in line with declines in previous years, KPMG Netherlands said. The organization has investigated Dutch trust in algorithms every year for the past five years. Only 7 percent of respondents said they have become more confident in the use of algorithms.

The distrust of government services and financial institutions’ use of algorithms is growing. Half of the respondents said government institutions are not honest and transparent about the algorithms they use. 53 percent said the same about financial institutions like banks and insurers.