Testing the capability of generative artificial intelligence for parent and caregiver information seeking

Abstract

Objective

This study explored the quality of generative artificial intelligence (AI) responses to common parenting questions across diverse sources of digitally available information.

Background

The recent rise of generative AI, such as ChatGPT and other large language models (LLMs), which generate answers by synthesizing publicly available information, raises questions about the quality of digital responses and the effect on parenting and outcomes for children.

Method

We hypothesized that querying a professionally prepared parenting newsletter would have higher quality responses than an LLM. We explored this by running 11 tests with five common parenting and caregiving topics about young children across controlled and open data sources. We analyzed three Cs (correctness, clarity, and connection), reliability (artificiality, credibility, and citation quality), and readability to assess the quality of LLM responses.

Results

ChatGPT largely provided correct and clear answers although citations were frequently absent and inaccurate. LLM responses often lacked emphasis on parent–child connection and developmental context, and reading level difficulty increased steeply.

Conclusion

Generative AI offers reasonably good answers to general parenting questions. However, parents and caregivers need to contextualize the information.

Implications

Topical experts may help meet nuanced parenting needs with cultural relevance and plain language, but AI can be useful for summarizing open-access content.