Do parents trust ChatGPT to answer health questions about their children? The answer is yes, according to a recent study by the Life Span Institute at the University of Kansas.
In the past, Internet users tended to turn to search engines to try to resolve medical concerns. They now complement this habit by looking at language patterns.
A worrying study
The study notably made it possible to recruit 116 parents who were given written documents related to children’s health problems. Half were generated by AI (often ChatGPT), the rest by experts.
Scientists found that parents were unable to distinguish between the two. Additionally, most of them opted for AI-generated texts and considered them more reliable.
The authors conclude:
Results indicate that ChatGPT is capable of influencing behavioral intentions regarding medication, sleep, and diet. Furthermore, there was little difference between ChatGPT and content experts in perceived morality, reliability, expertise, accuracy, and trust.
They add: “Notably, when differences were present, ChatGPT was rated as more trustworthy and accurate, and participants indicated that they would be more likely to trust the information presented by ChatGPT over that of the expert”.
These results are quite shocking. Indeed, while humans are not infallible, we know that these language models regularly make mistakes and often produce erroneous responses. We must hope that parents will be made aware of the risks and will take care to check and cross-reference their sources when using these technologies. In any case, this illustrates the ability of ChatGPT and its rivals to seduce us via often very (too?) clear texts.
When AI analyzes baby’s cries
The use of AI in child health is operating at full speed these days. Last April, the site AI Business mentioned in particular the case of the Nanni AI application developed by the startup Ubenwa Health. The latter would be able to translate a baby’s cries to help parents better understand their child’s needs.
This technology was trained on thousands of clinically labeled recordings of infant cries. It would be able to detect certain medical problems, for example signs of birth asphyxia, a respiratory disease in infants, in crying with an accuracy of 92.5%.
