Do Children with Autism Spectrum Disorders Show Selective Trust in Social Robots?

PURPOSE: Previous researches suggest that social robots can facilitate the learning of children with Autism Spectrum Disorder (ASD) by enhancing their interests, engagement, and attention. However, there is limited understanding regarding whether children with ASD can learn directly from the testimony of social robots and whether they can remain vigilant based on the perceived accuracy of these robots. Therefore, the present study was conducted to examine whether children with ASD demonstrated selective trust towards social robots. METHODS: Twenty-nine children with ASD between ages of 4-7 years, and 38 typically-developing (TD) age and IQ-matched peers participated in classic selective trust tasks. During the tasks, they learned the names of novel objects from either a pair of social robots or a pair of human informants, where one informant had previously been established as accurate and the other inaccurate. RESULTS: Children with ASD trusted information from an accurate social robot over an inaccurate one, similar to their performance with human informants. However, compared to TD children, children with ASD exhibited lower levels of selective trust regardless of the type of informants they learned from. CONCLUSIONS: Our study suggests that children with ASD can selectively trust and acquire knowledge from social robots, shedding light on the potential use of social robots in supporting individuals with ASD.

School-age children are more skeptical of inaccurate robots than adults.

We expect children to learn new words, skills, and ideas from various technologies. When learning from humans, children prefer people who are reliable and trustworthy, yet children also forgive people's occasional mistakes. Are the dynamics of children learning from technologies, which can also be unreliable, similar to learning from humans? We tackle this question by focusing on early childhood, an age at which children are expected to master foundational academic skills. In this project, 168 4-7-year-old children (Study 1) and 168 adults (Study 2) played a word-guessing game with either a human or robot. The partner first gave a sequence of correct answers, but then followed this with a sequence of wrong answers, with a reaction following each one. Reactions varied by condition, either expressing an accident, an accident marked with an apology, or an unhelpful intention. We found that older children were less trusting than both younger children and adults and were even more skeptical after errors. Trust decreased most rapidly when errors were intentional, but only children (and especially older children) outright rejected help from intentionally unhelpful partners. As an exception to this general trend, older children maintained their trust for longer when a robot (but not a human) apologized for its mistake. Our work suggests that educational technology design cannot be one size fits all but rather must account for developmental changes in children's learning goals.

Detecting lies through others' eyes: Children use perceptual access cues to evaluate listeners' beliefs about informants' deception.

Children are often third-party observers of conversations between informants and receivers. Although 5- and 6-year-olds can identify and reject informants' false testimony, it remains unclear whether they expect others to do the same. Accurately assessing others' impressions of informants and their testimony in a conversational setting is essential for children's navigation of the social world. Using a novel second-order lie detection task, the current study examined whether 4- to 7-year-olds (N = 74; Mage = 69 months) take receivers' epistemic states into account when predicting whether a receiver would think an informant is truthful or deceptive. We pitted children's firsthand observations of reality against informants' false testimony while manipulating receivers' perceptual access to a sticker-hiding event. Results showed that when the receiver had perceptual access and was knowledgeable, children predicted that the receiver would think the informant is lying. Critically, when the receiver lacked perceptual access and was ignorant, children were significantly more likely to predict that the receiver would think the informant is telling the truth. Second-order theory of mind and executive function strengthened this effect. Findings are interpreted using a dual-process framework and provide new insights into children's understanding of others' selective trust and susceptibility to deception.

Children's susceptibility to online misinformation.

Children have a reputation for credulity that is undeserved; even preschoolers have proven adept at identifying implausible claims and unreliable informants. Still, the strategies children use to identify and reject dubious information are often superficial, which leaves them vulnerable to accepting such information if conveyed through seemingly authoritative channels or formatted in seemingly authentic ways. Indeed, children of all ages have difficulty differentiating legitimate websites and news stories from illegitimate ones, as they are misled by the inclusion of outwardly professional features such as graphs, statistics, and journalistic layout. Children may not be inherently credulous, but their skepticism toward dubious information is often shallow enough to be overridden by the deceptive trappings of online misinformation.

Do preschoolers trust a competent robot pointer?

How young children learn from different informants has been widely studied. However, most studies investigate how children learn verbally conveyed information. Furthermore, most studies investigate how children learn from humans. This study sought to investigate how 3-year-old children learn from, and come to trust, a competent robot versus an incompetent human when competency is established using a pointing paradigm. During an induction phase, a robot informant pointed at a toy inside a transparent box, whereas a human pointed at an empty box. During the test phase, both agents pointed at opaque boxes. We found that young children asked the robot for help to locate a hidden toy more than the human (ask questions) and correctly identified the robot to be accurate (judgment questions). However, children equally endorsed the locations pointed at by both the robot and the human (endorse questions). This suggests that 3-year-olds are sensitive to the epistemic characteristics of the informant even when its displayed social properties are minimal.

In the presence and absence of conflicting testimony, children's selective trust in the in-group informant in moral judgment and knowledge access.

In this study, we assessed whether the trust model formed by children in a moral judgment context with an inaccurate in-group informant affected their corresponding trust model in the knowledge access context and whether conditions (the presence of conflicting testimony: an inaccurate in-group informant paired with an accurate out-group informant; the absence of conflicting testimony: only an inaccurate in-group informant) influenced the trust model. Children aged 3 to 6 years (N = 215; 108 girls) in blue T-shirts as in-group members completed selective trust tasks in the moral judgment and knowledge access contexts. Results for moral judgment showed that children under both conditions were more likely to trust informants based on accurate judgments and gave less consideration to group identity. Results for knowledge access showed that in the presence of conflicting testimony, 3- and 4-year-olds trusted the in-group informant at chance, but 5- and 6-year-olds trusted the accurate informant. In the absence of conflicting testimony, 3- and 4-year-olds agreed more with the inaccurate in-group informant, but 5- and 6-year-olds trusted the in-group informant at chance. The results indicated that older children considered the accuracy of the informant's previous moral judgment for selective trust in the context of knowledge access while ignoring group identity, but that younger children were affected by in-group identity. The study found that 3- to 6-year-olds' trust in inaccurate in-group informants was conditional and that their trust choices appeared to be experimentally conditioned, domain specific, and age differentiated.

Relations between fantasy and transfer of learning from storybooks.

Prior research presents a mixed picture regarding the circumstances under which children transfer learning of problem solutions from fantastical stories to real-world problems. Two experiments examined 3- to 5-year-old children's transfer of learning from fantastical storybooks that systematically varied in the fantastical abilities of storybook characters. In both experiments, participants heard stories about a character solving physical problems, and then participants attempted to solve analogous real-world problems. In Experiment 1, children heard stories that varied the fantastical abilities and practices of the protagonist; characters either did or did not have the ability to violate physical laws and did or did not use magic to help in solving a problem. Children were more likely to transfer problem solutions from the stories in which characters were presented as having the ability to violate real-world physical laws. In Experiment 2, the fantastical abilities of the characters varied by whether the characters were described as real, as pretend but living in a world where no physical laws could be violated, as pretend and living in a world where some physical laws could be violated, or as pretend and living in a world where many physical laws could be violated. Other than varying the characters' abilities, all characters used realistic solutions to solve the problem. Again, transfer was higher for children who heard about characters with the ability to violate real-world laws. The findings suggest that fantastical stories in which characters have the ability to do impossible things but use realistic solutions to problems can be effective in teaching children how to solve physical problems.

The roles of group membership and social exclusion in children's testimonial learning.

Here, we used high- and low-stakes testimonial learning tasks to better understand two important types of social influence on children's learning decisions: group membership and social ostracism. Children (4- and 5-year-olds; N = 100) were either included or excluded by in-group or outgroup members in an online ball tossing game. Then, children were asked to selectively learn new information from either an in-group or out-group member. They also received counterintuitive information from an in-group or out-group member that was in conflict with their own intuitions. When learning new information, children who were excluded were more likely to selectively trust information from their in-group member. In contrast, when accepting counterintuitive information, children relied only on group membership regardless of their exclusion status. Together, these findings demonstrate ways in which different forms of testimonial learning are guided not only by epistemic motivations but also by social motivations of affiliation and maintaining relationships with others.

Social cognition and trust: Exploring the role of theory of mind and hostile attribution bias in children's skepticism of inaccurate informants.

A growing body of research has examined the role of individual differences in children's selective trust. The current study was designed to explore how individual differences in theory of mind and hostile attribution bias affect children's trust. Four- and five-year-old children took part in a standard selective trust paradigm in which they had the choice between a previously inaccurate informant and an unfamiliar informant. They were also asked to interpret why the previously inaccurate informant had provided incorrect information in the past. Finally, children completed a hostile attribution bias task and a theory of mind task. Children with better theory of mind ability were more likely to defer to the unfamiliar informant on the selective trust task. Children with greater hostile attribution bias were more likely to interpret previous inaccuracy as a result of "being tricky" rather than having "made a mistake." However, these interpretations did not influence children's choices on the selective trust task. Therefore, although there is reason to believe that establishing selective trust involves both cognitive and social processes, the current study raises questions about the nature of this relationship and how children draw on different sociocognitive skills when establishing epistemic trust.

Knowing How You Know: Toddlers Reevaluate Words Learned From an Unreliable Speaker.

There has been little investigation of the way source monitoring, the ability to track the source of one's knowledge, may be involved in lexical acquisition. In two experiments, we tested whether toddlers (mean age 30 months) can monitor the source of their lexical knowledge and reevaluate their implicit belief about a word mapping when this source is proven to be unreliable. Experiment 1 replicated previous research (Koenig & Woodward, 2010): children displayed better performance in a word learning test when they learned words from a speaker who has previously revealed themself as reliable (correctly labeling familiar objects) as opposed to an unreliable labeler (incorrectly labeling familiar objects). Experiment 2 then provided the critical test for source monitoring: children first learned novel words from a speaker before watching that speaker labeling familiar objects correctly or incorrectly. Children who were exposed to the reliable speaker were significantly more likely to endorse the word mappings taught by the speaker than children who were exposed to a speaker who they later discovered was an unreliable labeler. Thus, young children can reevaluate recently learned word mappings upon discovering that the source of their knowledge is unreliable. This suggests that children can monitor the source of their knowledge in order to decide whether that knowledge is justified, even at an age where they are not credited with the ability to verbally report how they have come to know what they know.

Unsafe to eat? How familiar cartoon characters affect children's learning about potentially harmful foods.

Young children learn about the properties of foods, such as taste and healthiness, from others. By using selective trust tasks in which a familiar cartoon character and an unfamiliar informant provided different testimony about food safety, this study examined how an informant's familiarity affected 4- to 6-year-old children's selective social learning about food safety. In Experiment 1, when judging the safety of foods from the familiar cartoon character and the unfamiliar character, children across all age groups showed a preference for asking the familiar character for information. For endorse questions, 4- and 5-year-olds did not consistently accept or reject either character's statements, while 6-year-olds endorsed the unfamiliar cartoon character's statements more often than the familiar character's statements. In Experiment 2, when the unfamiliar informant was a real adult instead of a fictional cartoon character, children sought out information from the familiar character more often than from the adult, and they did not differentially endorse statements by either informant. Moreover, children who had less advanced theory of mind skills and who viewed cartoon characters as more real were more likely to ask the cartoon character. These results suggest that although children prefer to obtain information from familiar characters, they accept information about food safety from multiple kinds of sources and their social-cognitive skills play a role in their decisions.

Susceptibility to Being Lured Away by a Stranger: A Real-World Field Test of Selective Trust in Early Childhood.

In this preregistered field study, we examined preschool children's selective trust in a real-life situation. We investigated whether 3- to 6-year-old children (total N = 240) could be lured to a new location within their school grounds by an unfamiliar adult confederate. In a between-subjects manipulation, the confederate established either a high or a low level of personal credibility by providing information that the child knew to be either true or false. In Experiment 1, in which the confederate was female, children showed sensitivity to informational accuracy by being less willing to leave with an uninformed confederate, and this effect increased with age. In Experiment 2, in which the confederate was male, children were reluctant to leave regardless of informational accuracy. These findings point to real-world implications of epistemic-trust research and provide the first evidence regarding the early development of selective trust in a high-stakes naturalistic context.

(In)Sensitivity to Accuracy? Children's and Adults' Decisions About Who to Trust: The Teacher or the Internet.

Recent research has challenged the extended idea that when presented with conflicting information provided by different sources, children, as do adults, make epistemic judgments based on the past accuracy of each source. Instead, individuals may use relatively simple, but adaptive non-epistemic strategies. Here we examined how primary-school children (N = 114) and undergraduate students (N = 57) deal with conflicting information provided by two key sources of information in their day-to-day lives: their teacher and the Internet. In order to study whether the inaccuracy of a source generated a decline in trust, we manipulated this variable between participants: teacher-wrong and Internet-wrong conditions. For this, we first presented two baseline trials, followed by the accuracy manipulation, and finally, two post-test trials. Analyses were performed on group performance as well as on individual performance, to explore the individual patterns of responses. Results revealed that most participants showed no preference for any source during baseline, with no age differences in their overall choices. Crucially, when a given source provided inaccurate information about a familiar issue, most children and adults did not lose trust on this source. We propose tentative explanations for these findings considering potential differences in the participants' strategies to approach the task, whether or not epistemic.

In sickness and in filth: Developing a disdain for dirty people.

Cleanliness is universally valued, and people who are dirty are routinely marginalized. In this research, we measured the roots of negative attitudes toward physically unclean individuals and examined the differences that exist in these attitudes between childhood and adulthood. We presented 5- to 9-year-old children and adults (total N = 260) with paired photographs of a dirty person and a clean person, and we measured biases with a selective trust task and an explicit evaluation task. In Study 1, where images of adults were evaluated, both children and adults demonstrated clear biases, but adults were more likely to selectively trust the clean informant. Study 2 instead used images of children and included several additional tasks measuring implicit attitudes (e.g., an implicit association task) and overt behaviors (a resource distribution task) and also manipulated the cause of dirtiness to include illness, enjoyment of filth, and accidental spillage. Children and adults again revealed strong biases regardless of the cause of dirtiness, but only children exhibited a bias on the explicit evaluation task. Study 3 replicated these findings in India, a country that has historically endorsed strong purity norms. Overall, this research indicates that dirty people are targets of discrimination from early in development, that this is not merely a Western phenomenon, and that this pervasive bias is most strongly directed at individuals of similar ages.

Epistemic vigilance online: Textual inaccuracy and children's selective trust in webpages.

In this age of 'fake news', it is crucial that children are equipped with the skills to identify unreliable information online. Our study is the first to examine whether children are influenced by the presence of inaccuracies contained in webpages when deciding which sources to trust. Forty-eight 8- to 10-year-olds viewed three pairs of webpages, relating to the same topics, where one webpage per pair contained three obvious inaccuracies (factual, typographical, or exaggerations, according to condition). The paired webpages offered conflicting claims about two novel facts. We asked participants questions pertaining to the novel facts to assess whether they systematically selected answers from the accurate sources. Selective trust in the accurate webpage was found in the typos condition only. This study highlights the limitations of 8- to 10-year-olds in critically evaluating the accuracy of webpage content and indicates a potential focus for educational intervention. Statement of contribution What is already known on this subject? Children display early epistemic vigilance towards spoken testimony. They use speakers' past accuracy when deciding whom to trust regarding novel information. Little is known about children's selective trust towards web-based sources. What does this study add? This study is the first to examine whether textual inaccuracy affects children's trust in webpages. Typos but not semantic errors led to reduced trust in a webpage compared to an accurate source. Children aged 8-10 years show limited evaluation of the accuracy of online content.

Retrospective inferences in selective trust.

Young children learn selectively from others based on the speakers' prior accuracy. This indicates that they recognize the models' (in)competence and use it to predict who will provide the most accurate and useful information in the future. Here, we investigated whether 5-year-old children are also able to use speaker reliability retrospectively, once they have more information regarding their competence. They first experienced two previously unknown speakers who provided conflicting information about the referent of a novel label, with each speaker using the same novel label to refer exclusively to a different novel object. Following this, children learned about the speakers' differing labelling accuracy. Subsequently, children selectively endorsed the object-label link initially provided by the speaker who turned out to be reliable significantly above chance. Crucially, more than half of these children justified their object selection with reference to speaker reliability, indicating the ability to explicitly reason about their selective trust in others based on the informants' individual competences. Findings further corroborate the notion that young children are able to use advanced, metacognitive strategies (trait reasoning) to learn selectively. By contrast, since learning preceded reliability exposure and gaze data showed no preferential looking toward the more reliable speaker, findings cannot be accounted for by attentional bias accounts of selective social learning.

The role of epistemic and social characteristics in children's selective trust: Three meta-analyses.

Over the last 15 years, researchers have been increasingly interested in understanding the nature and development of children's selective trust. Three meta-analyses were conducted on a total of 51 unique studies (88 experiments) to provide a quantitative overview of 3- to 6-year-old children's selective trust in an informant based on the informant's epistemic or social characteristics, and to examine the relation between age and children's selective trust decisions. The first and second meta-analyses found that children displayed medium-to-large pooled effects in favor of trusting the informant who was knowledgeable or the informant with positive social characteristics. Moderator analyses revealed that 4-year-olds were more likely to endorse knowledgeable informants than 3-year-olds. The third meta-analysis examined cases where two informants simultaneously differed in their epistemic and social characteristics. The results revealed that 3-year-old children did not selectively endorse informants who were more knowledgeable but had negative social characteristics over informants who were less knowledgeable but had positive social characteristics. However, 4- to 6-year-olds consistently prioritized epistemic cues over social characteristics when deciding who to trust. Together, these meta-analyses suggest that epistemic and social characteristics are both valuable to children when they evaluate the reliability of informants. Moreover, with age, children place greater value on epistemic characteristics when deciding whether to endorse an informant's testimony. Implications for the development of epistemic trust and the design of studies of children's selective trust are discussed.

Looking smart: Preschoolers' judgements about knowledge based on facial appearance.

Preschoolers use others' behaviours to make inferences about what traits they possess (Harris et al., 2018, Ann. Rev. Psychol., 69, 251). The current study examined whether 4- and 5-year-olds also associate others' behaviour with how they appear on the surface. Specifically, we asked whether children's sensitivity to different face-traits (e.g., Cogsdill et al., 2014, Psychol. Sci., 25, 1132) would bias them to associate knowledgeable behaviours with faces that adults rate as highly competent- or trustworthy-looking. We find that preschoolers expect puppets with trustworthy-looking faces to be knowledgeable about the functions of familiar objects. In contrast, children did not match a puppet's knowledge to facial features that adults rate as varying in competence. These data suggest that children, like adults, are biased to associate facial appearance and behaviour. Furthermore, this bias appears to be rooted in a response to the same facial features that have been found to govern judgements of trustworthiness across development (e.g., Jessen & Grossmann, 2016, J. Cogn. Neurosci., 28, 1728). Statement of contribution What is already known on this subject? Preschoolers selectively trust others using epistemic and non-epistemic cues (Harris et al., 2018). Preschoolers associate specific faces with trustworthiness and competence (Cogsdill et al., 2014). What the present study adds? Preschoolers infer that trustworthy-looking characters will behave knowledgeably. Preschoolers do not infer that competent-looking characters will behave knowledgeably. Children's reliability judgements are influenced by others' appearance.

An eye-tracking study of selective trust development in children with and without autism spectrum disorder.

The purpose of this study was to explore whether children with autism display selectivity in social learning. We investigated the processing of word mappings provided by speakers who differed on previously demonstrated accuracy and on potential degree of reliability in three groups of children (children with autism spectrum disorder, children with developmental language disorder, and typically developing children) aged 4-9 years. In Task 1, one speaker consistently misnamed familiar objects and the second speaker consistently gave correct names. In Task 2, both speakers provided correct information but differed on how they could achieve this accuracy. We analyzed how the speakers' profiles influenced children's decisions to rely on them in order to learn novel words. We also examined how children attended to the speakers' testimony by tracking their eye movements and comparing children' gaze distribution across speakers' faces and objects of their choice. Results show that children rely on associative trait attribution heuristics to selectively learn from accurate speakers. In Task 1, children in all groups preferred the novel object selected by accurate speakers and directly avoided information provided by previously inaccurate speakers, as revealed by the eye-tracking data. In Task 2, where more sophisticated reasoning about speakers' reliability was required, only children in the typically developing group performed above chance. Nonverbal intelligence score emerged as a predictor of children's preference for more reliable informational sources. In addition, children with autism exhibited reduced attention to speakers' faces compared with children in the comparison groups.

Knowing who knows: Metacognitive and causal learning abilities guide infants' selective social learning.

Given the widespread interest in the development of children's selective social learning, there is mounting evidence suggesting that infants prefer to learn from competent informants (Poulin-Dubois & Brosseau-Liard, Current Directions in Psychological Science, 2016, 25). However, little research has been dedicated to understanding how this selectivity develops. The present study investigated whether causal learning and precursor metacognitive abilities govern discriminant learning in a classic word-learning paradigm. Infants were exposed to a speaker who accurately (reliable condition) or inaccurately (unreliable condition) labeled familiar objects and were subsequently tested on their ability to learn a novel word from the informant. The predictive power of causal learning skills and precursor metacognition (as measured through decision confidence) on infants' word learning was examined across both reliable and unreliable conditions. Results suggest that infants are more inclined to accept an unreliable speaker's testimony on a word learning task when they also lack confidence in their own knowledge on a task measuring their metacognitive ability. Additionally, when uncertain, infants draw on causal learning abilities to better learn the association between a label and a novel toy. This study is the first to shed light on the role of causal learning and precursor metacognitive judgments in infants' abilities to engage in selective trust.