Sophia Bakhtiari
Research Mentor: Lauren Girouard
Mentor Department: Psychology, LSA
Author(s): Sophia Bakhtiari, Lauren Girouard
Session: Session 1 (9:00 AM – 9:50 AM)
Presentation Type: Poster 42
Abstract
The rise of artificial intelligence, and the chatbots that use it, have thrown into question where children will turn for health related information: adults, doctors, or chatbots. For one, will children be able to recognize when chatbots make errors? And how will this vary with age? Previous research has shown that children, regardless of age, trusted accurate humans at a similar level to accurate robots; however, the children between the ages of 3-5 years trusted an inaccurate human more than an inaccurate robot (Li & Yow 2023). Another study tested whether children 3-5 years of age preferred learning from a human or robot; Baumann et. al. found that 3 year old children trusted both informants equally, while 5 year old children preferred the robot. If children trust technology like chatbots, will 7-14 year old children be able to recognize when chatbots make mistakes? The goal of this study is to understand the relationship between a child’s error recognition and their age. In the error monitoring trials, children are asked to examine two statements made by a pair of adults, doctors, or chatbots, respectively, and identify which informant made an error in their statement. The child is then presented with three questions and asked which informant they would rather answer the question, a test to determine whether they trust the informant that made no errors or the informant with the error. This study is ongoing, having collected 149 out of the 288 target responses. With participants’ ages ranging from 7-14 years old, it’s important to consider how age will impact the ability of children in this error monitoring section. With more critical reading skills, having read more advanced books and literature in school, older children may have an advantage compared to younger children in correctly identifying the adult, human, or chatbot that made a grammatical error in their statement. There was a significant direct correlation between age and error monitoring, with younger children less frequently identifying the error. Young children especially may need help with increasing their AI literacy and learning to spot errors made by popular AI chatbots.
