Who is more trustworthy?: Selectuve Trust in children aged 6 to 11 in digital voice assistants

Abstract

Children's access to the internet and electronic devices has been growing every day. At the same time, we live in an age of constant exposure to fake news and, therefore, parents and teachers now worry about teaching their children to consume information in a critical way. In the past 20 years, several researchers have been investigating the development of selective trust, i.e., the ability to discriminate between good and bad informants in new learning situations. Some recent studies have shown that 4-year-old children are already capable of distinguishing a precise electronic informant from an imprecise electronic informant. Moreover, there is evidence that children use different strategies when they have to choose between two digital informants or between a digital informant and a human informant. More recently, digital voice assistants have gained popularity. On one hand, these devices resemble computers because of their capacity to provide information (since they use the internet as a database), but they also share technical features with social digital agents, such as robots. The main goal of the present study is to contribute to this research field by investigating children's selective trust when facing two potential informants: a real person and a digital voice assistant. Additionally, the study aims to explore which characteristics children assign to both informants (epistemic, biologic and psychologic). Sixty children from 6 to11 years of age will participate in the study. The procedure consists in presenting children with two characters (a human informant and a digital voice assistant) who vary in terms of trustworthiness (one of them always gives accurate information whereas the other one doesn't). Participants will be randomly assigned into two different conditions. In the first condition (C1), the human informant is trustworthy and in the second condition (C2), the digital voice assistant is more reliable. In an initial familiarization phase, participants must evaluate the informants based on their answers to six questions (the answers are known by children) three of them about factual knowledge (i.e., "What's the shape of a circle?"), and three of them are about personal knowledge (i.e., "How does a person feel when they are crying?"). In the second phase (test phase), they will choose one of the informants to answer another set of six questions (six attempts) to which they don't know the answer. Half of the questions are factuals (i.e., "What's the hottest planet in the solar system?") and half are personal questions (i.e., "How does a person who is learning mathematics feel?"). In line with previous studies, it is expected that, during familiarization, when children do not have information about the informants' previous reliability history, they will show a preference for the digital informant for factual questions and the human informant for personal questions. However, during test trials, when children already know about the informants' reliability status, we expect participants to always endorse the most trustworthy informant, regardless of the type of question being made (factual or personal), in both conditions. Finally, we expect that most children will not assign biological characteristics (i.e., ability to move autonomously) to voice assistants, but the younger ones (between 6 and 7 years old) will assign psychological characteristics (i.e., ability to feel curiosity). In contrast, older children will tend to attribute only epistemic characteristics (i.e., ability to access different types of knowledge) to the voice assistant, as U.S. studies have shown.