http://orcid.org/0000-0001-8778-5370
Figures
Abstract
Keeping commitments to others can be difficult, and we know that people sometimes fail to keep them. How does a speaker’s ability to keep commitments affect children’s practical decisions to trust and their epistemic decisions to learn? An amassing body of research documents children’s trust in testimonial learning decisions, which can be moved in the face of epistemic and moral evidence about an agent. However, other bases for trust go largely unexplored in this literature, such as interpersonal reasons to trust. Here, we investigated how direct bids for interpersonal trust in the form of making commitments, or obligations to the listener, influence a range of decisions toward that agent. We found that 3- and 4-year-olds’ (N = 75) practical decisions to wait and to share were moved as a function of a person’s commitment-keeping ability, but epistemic decisions to learn were not. Keeping one’s commitments may provide children with interpersonal reasons to trust, reasons that may function in ways distinct from the considerations that bear on accepting a claim.
Citation: Pesch A, Koenig MA (2018) Varieties of trust in preschoolers’ learning and practical decisions. PLoS ONE 13(8): e0202506. https://doi.org/10.1371/journal.pone.0202506
Editor: Jordy Kaufman, Swinburne University of Technology, AUSTRALIA
Received: December 5, 2017; Accepted: August 3, 2018; Published: August 20, 2018
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: This material is based upon work supported by a Diversity of Views and Experiences Fellowship to AP and by the Office of Vice Provost for Research (Grant No. 22890) and the John Templeton Foundation (Grant No. 60502) to MAK. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Testimony provides children with a rich source of information about the world. Much work has examined how preschool-aged children evaluate testimony, showing that children prefer to learn from agents who are more accurate [1], knowledgeable[2], and well-informed [3]. Preschoolers also monitor probabilistic or relative evidence about accuracy [4, 5], evidence about perceptual access to relevant information [6, 7], and overt evidence about expertise [8–10]. In addition, preschoolers base learning decisions on moral information [11, 12], attractiveness [13], group membership [14, 15], and age [16, 17], suggesting that children consider not only epistemic competencies when learning from others, but a variety of moral and interpersonal characteristics as well.
What does such selectivity reflect? On one view, children’s selective trust in testimony is conceptualized as a rational inference based on a range of evidence about an agent’s knowledge and intent [18–20]. This ‘evidential’ approach has launched a set of important questions regarding children’s abilities to use evidence (both positive and negative) to mitigate testimonial risks (for reviews, see[21, 22]). However, a dimension lost by focusing on the evidentiary value of testimony are the various interpersonal reasons that testimony provides to addressees to trust an agent. Thus, another way we might understand the trust elicited in selective social learning is to consider that a hearer’s decision to trust a speaker can be based in interpersonal considerations [23]. Indeed, there may be interpersonal epistemic reasons to believe someone that are not solely based in a listener’s ability to predict a speaker’s knowledge or goodwill [24]. This leaves open another sense in which children can trust others, taking them at their word, without knowing very much about them or even having evidence against them [25–27].
This sense of trust can be seen in our responses to obligations or commitments that others make, where trust in testimony might similarly reflect sensitivity to the interpersonal obligations or commitments that speakers present in different speech acts. A set of recent studies have found that trust is implicated in commitments to retrieve a larger second reward for a child in a delay of gratification task. Kidd, Palmeri, and Aslin [28] presented preschoolers with an experimenter who either fulfilled or broke their promise to retrieve art supplies on two occasions. Later, when tested in a traditional delay of gratification task (e.g. [29]), children in the unreliable condition were significantly less likely to wait for a second reward from that agent [28]. Thus, when confronted with an agent who stated their intent to retrieve a desired item, children’s later extensions of trust were found to be contingent on their prior interaction with that agent. A related study found that preschoolers were also less likely to wait for a larger reward in a delay of gratification task when the experimenter lied about a transgression rather than tell the truth [30]. These findings suggest that stated commitments to produce a desired item prompts children’s decisions to trust. This is further supported by recent work showing that children’s general trust views (i.e. viewing others as trustworthy) predict performance on a delay of gratification task, despite prior interaction with the agent administering the task [31]. Unclear from this work is whether this type of trust–an invitation to trust an agent to uphold her commitment to retrieve a reward–is the same type of trust that is involved when children are asked to learn new information from an agent. To our knowledge, only one study has sought to disentangle intention from outcome in an epistemic learning task. Liu, Vanderbilt, and Heyman [32] showed children an agent who either intended to help or deceive a third-party find stickers with either positive or negative outcomes. They found that 5- and 6-year-olds reasoned more about the outcome of prior testimony in their learning decisions, but took into account intention when assessing trait-relevant behavior. While this particular experiment does not answer whether the trust invited by the agent to learn was similar to the interpersonal trust invited in the delay of gratification task, it does suggest that children may be differentiating the information they use in their assessments of others.
The idea that children might base different decisions on distinct considerations is further supported by a recent body of work suggesting a set of dissociated judgments in children’s learning and practical decisions. For example, Hetherington, Hendrickson, and Koenig [33] presented preschoolers with either an antisocial in-group or prosocial out-group member. Importantly, the antisocial behavior of the in-group member reduced children’s liking of her and willingness to share with her compared to a neutral out-group member; however, it did not move decisions to seek new information from her. Likewise, when 18-month-old infants encountered an agent who incorrectly labeled a set of objects, their willingness to learn novel information was reduced but their willingness to share with that person was not [34]. When introduced to characters who differed in niceness and expertise, children based epistemic inferences on evidence of both expertise and niceness but based social inferences only on niceness [35]. Such findings suggest that children’s interpersonal reasons to trust the agent may dissociate from their epistemic reasons to believe their claims.
Do children extend different variants or forms of trust depending on the decision they are asked to make? In order to distinguish between epistemic and interpersonal extensions of trust, we presented children with learning decisions, as well as two types of practical decisions. As discussed above, delaying gratification is one type of practical decision that can be influenced by a speaker’s promise-keeping behavior. Another practical decision children make is how much and with whom to share [36, 37]. Research in behavioral economics has found that adults typically allocate slightly more than half of their money to strangers when playing a standard “trust game” [38, 39]. Children, too, share with others without any prior evidence of that person’s inclination to reciprocate [40, 41], and when asked to share a limited resource, share more with friends and familiar agents over strangers [42], suggesting that such decisions are not based in evidence about the recipient, and often moved by interpersonal considerations.
In the current study, we used commitments to elicit interpersonal trust in our target agent. Children interacted with an agent in an art project task, similar in design to Kidd et al.’s [28]. To produce violations of interpersonal trust, we presented two types of ‘bad interpersonal agents’ (one who failed to have good intentions, and one who failed to deliver good outcomes) in contrast to a good baseline agent (with good intentions, and who delivered good outcomes). We thus explored the extent to which interpersonal violations of trust–both in terms of intention and outcome–would influence children’s practical decisions (to delay gratification, to share a limited resource) and their epistemic learning decisions (to ask her for information and to endorse her claims). If interpersonal violations of trust are simply treated as inductive evidence about the reliability or goodwill of the agent, then we expect such information to broadly move children’s learning and practical decisions (consistent with [11, 12, 43, 44]). But, if interpersonal violations of trust constitute a violation of children’s interpersonal decisions to trust a speaker, then such violations might influence children’s interpersonal decisions—decisions to wait for the agent and to share with her—rather than their epistemic decisions to learn.
Method
Ethics statement
The Institutional Review Board at the University of Minnesota approved this research (#1603P85562). Written consent was obtained from the parents or legal guardians of the participants. Each participant verbally agreed to take part in the study.
Participants
Seventy-five 3- and 4-year-olds participated in the study (38 Female, Mage = 4.05; range = 3.07–4.96). This sample size was based on having 80% power to detect a moderate effect size of f = .36. Children were randomly selected from a database of children who are primarily Caucasian, native English speakers from middle to high SES homes. Children were assigned to one of three between-subject conditions: Baseline (positive intention-positive outcome) (n = 25, 12 females, Mage = 4.11; range = 3.13–4.97), Failed Commitments—Negative Outcomes (n = 25, 11 females, Mage = 4.16; range = 3.08–4.88), or Failed Commitments—Positive Outcomes (n = 25, 12 females, Mage = 4.01; range = 3.24–4.88). Eight additional children were excluded due to experimenter error (n = 4) or failure to complete the experiment (n = 4).
Procedure
Across all conditions, Experimenter 1 (E1) first asked participants for a mood rating. Participants were then invited to the testing room to participate in an art project task. Participants’ interaction with E1 during the art project differed depending on condition. E1 then administered the delay of gratification task. Following the delay of gratification task, E1 left and Experimenter 2 (E2) entered the testing room and administered the explicit liking, resource allocation, and selective learning tasks. These tasks were a series of video clips involving two agents: E1 and an unfamiliar, neutral agent (E3). Across participants, we counterbalanced which agent was presented first, which agent spoke first, and which side of the screen the agent appeared on. After completing these tasks, the second mood rating was administered. (See Table 1 for an overview of our procedure supplemented also by Fig 1).
Mood rating (adapted from [30]).
In line with the procedures used in Kidd et al. [28] and Michaelson and Munakata [30], the pre- and post-test mood ratings served to investigate whether or not the experimental manipulation affected participants’ mood. If so, this would potentially explain performance on our dependent measures. Participants were shown a six-point smiley face scale and asked to circle the face that matched how happy they were.
Art project task (adapted from [28]).
Participants were given two opportunities to wait for a larger reward from E1. For both choices, all participants were told that they could choose to use the current supplies on the table or wait for better supplies from the other room. Following Kidd, Palmeri, and Aslin [28], participants were presented with a small set of well-used crayons (e.g. dull, wrapper peeled and dirty) in a tightly sealed jar for Choice 1 and a small star sticker inside a plastic envelope for Choice 2. The larger reward was a large art set containing new crayons, colored pencils, and markers for Choice 1 and a sleeve of attractive stickers for Choice 2.
Intention information: In the Baseline and Failed Commitments—Negative Outcomes conditions, E1 stated her positive intentions across choices 1 and 2, “I really want you to have the best set of art supplies/stickers,” before she left the room. In the Failed Commitments—Positive Outcomes condition, E1 stated her negative intentions across choices 1 and 2, “Actually, I really don’t want you to have the best set of art supplies/stickers because I saved the best set for myself,” before she left the room. The experimenter left the participant in the room alone for 2.5 minutes. All participants chose to wait. (See S2 File. Experimental Script: Art Project and Delay of Gratification in Supporting Information for full scripted dialog).
Outcome information: In the Baseline condition, the experimenter returned with the promised supplies and stated, “Here is the best set of art supplies/stickers!” In the Failed Commitments—Negative Outcomes condition, the experimenter returned without the promised supplies and stated, “I’m sorry, but we don’t have any other art supplies/stickers after all, why don’t you use these [points to crayons/sticker on table] instead?” In the Failed Commitments—Positive Outcomes condition, the experimenter returned with supplies and stated, “Actually, here is another set of art supplies/stickers. I saved the very best for myself but you can use these.” All participants were encouraged to use the supplies for approximately 3 minutes.
Delay of gratification task (adapted from [28]).
Immediately following the art project task, E1 produced a marshmallow on a plate and gave the child the choice to eat one marshmallow right away, or wait for her to return with a second marshmallow. E1 then gathered up the materials from the art project task and left the participant alone in the room. The experimenter covertly observed the participant through a one-way window and waited until the participant had either eaten the marshmallow, became visibly upset, or 15 minutes had elapsed. In all scenarios, the experimenter returned to the room and delivered a second marshmallow to the participant, which ensured that participants received the same intention and outcome information about the experimenter within conditions before engaging in the explicit liking, resource allocation, and selective learning trials.
Explicit liking.
Two explicit liking trials were given: once before the selective learning and resource allocation trials, and the second after (Refer to Table 1). Participants were separately shown an image of E1 and E3 and asked to rate how much they liked that agent on a six-point smiley face scale. Across participants, we counterbalanced which agent’s image they were shown first.
Selective learning.
Participants were shown an image of an object (e.g. key) and told, “The first thing that’s hiding is this [object]. I wonder where it is!” Participants then saw a still image of the two agents with a different colored box in front of each one and were asked to indicate which agent they would like to ask to find out where the object was (ask trials). The agents then made conflicting claims about the location of the object (e.g. “I think it’s in the red box” vs. “I think it’s in the blue box”) and participants were asked to indicate where they thought the object was hidden (endorse trials). Participants completed six ask and six endorse trials. Each trial included a different object and a different set of colored boxes.
Resource allocation.
Participants completed three resource allocation trials to assess their willingness to share resources with the agents. Participants were presented with two cups, each with an image of one of the agents and five paper coins. The experimenter explained to participants, “Here are five coins for you to share. How many coins should this person [point to one agent] get and how many coins should this person [point to second agent] get?” Participants were encouraged to place all five coins into the cups. These trials were dispersed throughout the selective learning trials: the first appearing after selective learning trial two, the second after trial four, and the third after trial six.
Results
Preliminary analyses found no effect of gender or order, so these were collapsed within conditions for the following analyses to allow for greater statistical power.
Mood rating
We ran a three way mixed analysis of variance (ANOVA) with condition (Baseline vs. Failed Commitments—Negative Outcomes vs. Failed Commitments—Positive Outcomes) and age (3 vs. 4) as the between-subjects factors, trial (trial 1 vs. trial 2) as the within-subjects factor, and mood rating as the dependent variable. There were no main effects of condition (F(2, 67) = 0.795, ns) or age (F(1, 67) = 2.572, ns). There was a significant effect of trial (F (1, 67) = 5.058, p = .02): children’s initial mood ratings (M = 5.26) were higher than their second mood rating (M = 4.82). The absence of condition differences suggest that our measures of interest are unlikely to have been influenced by changes in mood.
Explicit liking
Preliminary analyses found no difference between trial 1 and trial 2 for explicit liking, so these were collapsed within agent. We ran a three way mixed ANOVA with condition (Baseline vs. Failed Commitments—Negative Outcomes vs. Failed Commitments—Positive Outcomes) and age (3 vs. 4) as the between-subjects factors, agent (E1 vs. E3) as the within-subjects factor, and liking score as the dependent variable. There was no effect of condition (F(2, 68) = 1.44, ns) or age (F(1,68) = 0.16, ns). There was a significant effect of agent (F(1,68) = 14.32, p < .001): Children rated E1 (M = 4.93) as more likeable than E3 (M = 4.12).
Practical decisions (delay of gratification and resource allocation)
Delay of gratification.
In the Baseline condition, children waited without eating the marshmallow for a mean duration of 733.64s whereas those in the Failed Commitments—Negative Outcomes condition waited 534.68s and those in the Failed Commitments—Positive Outcomes condition waited 393.80s (See Fig 2). Since these data were not normally distributed, a Kruskal-Wallis rank sum test was performed to test the null hypothesis that the conditions had the same distribution of wait times. The Kruskal-Wallis test showed that there was a significant difference in wait time between the three conditions, X2(2) = 9.31, p = 0.009. Post-hoc comparisons using Wilcoxon rank sum test revealed that children in the Failed Commitments–Negative Outcomes and Failed Commitments–Positive Outcomes conditions waited less compared to those in the Baseline condition (p = 0.03 and p = 0.003, respectively).
Error bars represent standard errors of the mean.
A linear regression analysis was run to test for the effect of age, controlling for condition. Age was significant, (B = 186.94, t = 2.26, p = .02), indicating that 4-year-olds (M = 643.8s) waited longer than 3-year-olds (M = 456.8s). In addition, we examined whether differences existed in the proportion of children waiting the entire 15 min without tasting the marshmallow. A chi-square test confirmed that the proportion of children waiting the full 15 minutes was significantly different between conditions (X2(2) = 10.39, p = 0.005). In the Baseline condition, 72% of children (18 out of 25) waited the full 15 minutes. The proportion of children waiting was significantly less compared to those in the Baseline condition in the Failed Commitments—Negative Outcomes condition (40% or 10 out of 25, p = .04) as well as in the Failed Commitments—Positive Outcomes condition (28% or 7 out of 25, p = .004).
Resource allocation.
A sharing score was calculated by summing the amount of trials (max = 3) on which children shared more coins (i.e. 3 or more) with E1. This score was converted to a percentage for ease of interpretation. An omnibus ANOVA was run with age (3 vs. 4) and condition (Baseline vs. Failed Commitments—Negative Outcomes vs. Failed Commitments—Positive Outcomes) as between-subjects factors, and sharing score as the dependent measure. We found no effect of age (F(1,69) = 1.87, p = .17). We found a main effect of condition (F(2,69) = 5.11, p = .009, ηp2 = .13). Post-hoc Tukey’s comparisons (all ps < .05) revealed that children shared more with E1 in the Baseline condition (M = 72.00) compared to both the Failed Commitments—Negative Outcomes (M = 50.64) and Failed Commitments—Positive Outcomes conditions (M = 49.40), (See Fig 3).
Error bars represent standard errors of the mean.
Epistemic decisions (ask and endorse trials)
Ask trials.
An ask score was calculated by summing the amount of trials out of six that children chose E1 as the person they would like to ask to find out where an object was hidden. This score was converted to a percentage for ease of interpretation. An omnibus ANOVA was run with age (3 vs. 4) and condition (Baseline vs. Failed Commitments—Negative Outcomes vs. Failed Commitments—Positive Outcomes) as between-subjects factors, and ask score as the dependent measure. We found a non-significant, but trending effect of age (F(1,68) = 3.18, p = 0.07, ηp2 = .04), with 4-year-olds somewhat more likely to ask E1 for information (M = .58) compared to 3-year-olds (M = .48). We found no effect of condition (F(2,68) = 1.75, p = 0.18) and no interaction between age and condition (F(2,68) = 1.21, p = 0.31) (See Fig 4).
Dashed line represents chance. Error bars represent standard errors of the mean.
Endorse trials.
An endorse score was calculated by summing the amount of trials out of six that children chose to endorse the location offered by E1. An omnibus ANOVA was run with age (3 vs. 4) and condition (Baseline vs. Failed Commitments—Negative Outcomes vs. Failed Commitments—Positive Outcomes) as between-subjects factors, and endorse score as the dependent measure. There were no significant differences between conditions (F(2,69) = 0.12, p = 0.88) or between age groups (F(1,69) = 1.43, p = 0.23) and no interaction between age and condition (F(2,69) = 0.61, p = 0.54) (See Fig 4).
Discussion
Much interest in testimonial learning has focused on children’s selective trust in testimony; however, the precise nature of this trust remains unknown. In the current study, we presented children with one kind of interpersonally bad agent–one who invoked an interpersonal violation of trust by failing to deliver a desired item or one who failed to withhold a desired item. These failures both reflect failed commitments that were stated directly to participants. We found that this manipulation reduced 3- and 4-year-olds’ decisions to wait for and share with E1, but did not similarly reduce their decisions to seek or endorse information offered by her. To account for this dissociation, we suggest distinguishing between cases of ‘epistemic or evidential trust’ from ‘interpersonal trust.’ In cases of epistemic trust, decisions to accept what someone claims are responsive to different types of inductive evidence that bear on the claim (e.g. such as the listener’s prior knowledge and/or the speaker’s certainty, prior reliability, their expertise, moral behavior, etc.). In cases of interpersonal trust, when we trust someone who made a commitment or promise to us, we do so not because we have collected the evidence (after all, whether they keep their commitment is up to them, not us) [24, 27]. Rather, when a speaker makes a commitment to act in a specific way to a listener, she offers an answer to the question of what she will do. If the listener accepts her answer in the spirit in which it is given, the listener has decided to trust the speaker interpersonally, taking her at her word.
Our findings argue for identifying epistemic reasons to trust in testimony that are based on weighing the evidence one has to come to a conclusion (evidence about the world, the agent, etc.) as distinct from interpersonal decisions to trust, which can be based on equally compelling reasons to trust that person (e.g., speech acts like promises, relations between speaker and addressee). This idea is supported by other studies which have manipulated agent characteristics across multiple domains. For example, when introduced to characters who differed in niceness and expertise, children based epistemic inferences on evidence of both expertise and niceness but based social inferences only on niceness [35]. Similarly, when introduced to in-group members who refused to share, children’s practical decisions to share were reduced but their epistemic decisions to learn from that agent were not [33]. These findings suggest that children appreciate that an agent can be untrustworthy for interpersonal reasons (e.g., relationship-building) but still have knowledge to share. We build upon this work in the present study by examining how a speaker’s interpersonal commitments influence a range of decisions, both practical and epistemic. Given recent evidence that a delay of gratification task measures children’s trust [31], our manipulation suggests that in our study, and in other similar experiments (e.g. [33]), a distinct type of trust was reduced, one that may not be implicated in all testimonial learning decisions. Trusting someone does not consist solely in predicting their reliability in light of the evidence we have collected. In fact, reasons to trust a person can be quite distinct from evidential reasons to favor someone’s testimony [27, 45]. Thus, it is important to present children with a range of decisions involving multi-dimensional agents in order to better understand when and how reasons to trust a person might dissociate from reasons to believe a claim.
Consistent with these ideas is the possibility that certain speech acts (promising, requesting, resolving) instigate children’s decisions to trust, while other speech acts (arguments, explanations, demonstrations) do not. As discussed above, when we accept an agent’s spoken commitments on trust as the answer to the question of what she will do, we accept her answer as our conclusion [27]. Explanations or arguments, in contrast, obviate the need for trust by placing children in a position to evaluate a set of considerations and draw conclusions based on them. In the current study, if children treated an agent’s commitments as direct bids for trust, then children’s distrust on delay tasks and sharing tasks can be seen as direct, natural responses to an experimenter who made a commitment, but who failed to uphold that commitment, thereby violating the trust invoked by it. Moreover, the current study highlights that failed commitments, even when they produce positive outcomes, can reduce children’s trust in that speaker. This suggests that children register not only when others let them down or disappoint them, but also when others commit to a negative action, punishment, or “threat” that they did not execute. The inconsistencies between one’s commitments and one’s actions may produce breaches of interpersonal trust.
Our learning trials, which presented speakers’ divergent claims about object locations, were not responsive to the manipulation. In contrast to delay decisions and resource allocations, children’s performance did not differ from chance in any of the three conditions. This may reflect the absence of interpersonal trust in the testimonial learning decisions that we presented, trials for which other kinds of evidence (e.g., perceptual access) may have been privileged [46, 6] However, there are other explanations for this null finding. One possibility is that because our learning trials did not present speakers who made promises or stated commitments, and did not present interpersonal bids for trust per se, children did not treat their prior experiences with the experimenter as relevant to the decisions posed by the learning trials. Along these lines, perhaps children simply treated the experimenter’s variable commitment-keeping behavior as evidence for or against her reliability. If true, however, it goes unexplained why this evidence about reliability was not registered in their evaluations of her testimony. In other words, a strictly evidential picture of children’s trust leaves it unclear why children were selective in their practical decisions (to wait, and to share) but not their epistemic decisions (to ask, and to endorse). After all, the experimenter showcased both inaccurate claims (negative outcomes) as well as lack of goodwill (negative intentions), both of which are forms of evidence known to attenuate children’s learning decisions [11, 12, 32]. If children reduced an agent’s failed commitments to evidence, it will need to be explained (a) why the agent’s unreliability was registered in certain practical decisions and not epistemic ones, and (b) why the interpersonal reasons to trust invoked by commitments played no role.
Another possibility is that children extend a form of interpersonal trust when considering a broad range of communication, including the claims made by speakers on our learning trials, and our manipulation was not sufficient to disrupt the default levels of trust that children regularly extend [47]. Indeed, our experimental manipulation ensured that children did not receive information about the differential knowledge or access of our two agents, suggesting that in the absence of conflicting epistemic information, children may have found both agents worth trusting for new information. Consistent with this is the possibility that the mechanisms underlying children’s decisions to trust in testimony may not be exclusively responsive to epistemic or evidential considerations, but interpersonal reasons as well. By taking into account when and how interpersonal bases for trust affect learning decisions, we can begin to expand current evidential models of selective trust and better discriminate the kinds of reasons children have to trust others. While children’s lack of selective learning in the present study could be due to the absence or presence of interpersonal trust, it nonetheless contrasts with the selectivity produced by the practical decisions children made. Future work will be necessary to understand how interpersonal trust functions in epistemic and practical decisions.
Current research in children’s testimonial learning seeks to take seriously the varieties of trust that feature in children’s cognitive development [48]. Recent work has begun to shift attention towards the various social goals [49] or inferences [18] that might feature in children’s selective learning from others (see also [50, 51]). Our findings demonstrate a dissociation between children’s practical and epistemic decisions and call for closer examination of the forms of trust that feature in such decisions. In addition, it will be important for future work to distinguish between epistemic and interpersonal bases for trust in samples from more diverse populations. For example, children growing up in less reliable environments may be less trusting in general, or more uniquely vigilant of certain violations of trust. We also think it important for future work to consider the possibility that when children accept information from deceptive informants (e.g. [47, 52]), it may not reflect irrational decisions that neglect the evidence, but rather children’s robust decisions to trust another person for interpersonal reasons, despite evidence against them.
Supporting information
S2 File. Experimental script: Art project and delay of gratification.
https://doi.org/10.1371/journal.pone.0202506.s002
(DOCX)
Acknowledgments
We thank Natalie Low, Amy Yi, and Victoria Fritz for their lab assistance as well as the parents and children who participated in this research. This research was supported by a Diversity of Views and Experiences Fellowship to AP, and by the John Templeton Foundation (Grant No. 60502) and the Office of Vice Provost for Research (Grant No. 22890) to MAK.
References
- 1. Koenig MA, Harris PL. Preschoolers mistrust ignorant and inaccurate speakers. Child Dev. 2005;76(6):1261–77. pmid:16274439
- 2. Einav S, Robinson EJ. When being right is not enough: four-year-olds distinguish knowledgeable informants from merely accurate informants. Psychol Sci. 2011;22(10):1250–3. pmid:21881060
- 3. Kushnir T, Wellman HM, Gelman SA. The role of preschoolers' social understanding in evaluating the informativeness of causal interventions. Cognition. 2008;107(3):1084–92. pmid:18039543
- 4. Einav S, Robinson EJ. Children's sensitivity to error magnitude when evaluating informants. Cognitive Development. 2010;25(3):218–32.
- 5. Pasquini ES, Corriveau KH, Koenig M, Harris PL. Preschoolers monitor the relative accuracy of informants. Dev Psychol. 2007;43(5):1216–26. pmid:17723046
- 6. Nurmsoo E, Robinson EJ. Identifying unreliable informants: do children excuse past inaccuracy? Dev Sci. 2009;12(1):41–7. pmid:19120411
- 7. Povinelli DJ, deBlois S. Young children's (Homo sapiens) understanding of knowledge formation in themselves and others. J Comp Psychol. 1992;106(3):228–38. pmid:1395492
- 8. Koenig MA, Jaswal VK. Characterizing children's expectations about expertise and incompetence: halo or pitchfork effects? Child Dev. 2011;82(5):1634–47. pmid:21790541
- 9. Kushnir T, Vredenburgh C, Schneider LA. "Who can help me fix this toy?" The distinction between causal knowledge and word knowledge guides preschoolers' selective requests for information. Dev Psychol. 2013;49(3):446–53. pmid:23339590
- 10. Lutz DJ, Keil FC. Early understanding of the division of cognitive labor. Child Dev. 2002;73(4):1073–84. pmid:12146734
- 11. Doebel S, Koenig MA. Children's use of moral behavior in selective trust: discrimination versus learning. Dev Psychol. 2013;49(3):462–9. pmid:23356525
- 12. Mascaro O, Sperber D. The moral, epistemic, and mindreading components of children's vigilance towards deception. Cognition. 2009;112(3):367–80. pmid:19540473
- 13. Bascandziev I, Harris PL. In beauty we trust: children prefer information from more attractive informants. Br J Dev Psychol. 2014;32(1):94–9. pmid:24164592
- 14. Elashi FB, Mills CM. Do children trust based on group membership or prior accuracy? The role of novel group membership in children's trust decisions. J Exp Child Psychol. 2014;128:88–104. pmid:25108696
- 15. Kinzler KD, Corriveau KH, Harris PL. Children's selective trust in native-accented speakers. Dev Sci. 2011;14(1):106–11. pmid:21159092
- 16. Jaswal VK, Neely LA. Adults don't always know best: preschoolers use past reliability over age when learning new words. Psychol Sci. 2006;17(9):757–8. pmid:16984291
- 17. Rakoczy H, Hamann K, Warneken F, Tomasello M. Bigger knows better: Young children selectively learn rule games from adults rather than from peers. British Journal of Developmental Psychology. 2010;28(4):785–98.
- 18. Landrum AR, Eaves BS Jr., Shafto P. Learning to trust and trusting to learn: a theoretical framework. Trends Cogn Sci. 2015;19(3):109–11. pmid:25563822
- 19. Shafto P, Eaves B, Navarro DJ, Perfors A. Epistemic trust: modeling children's reasoning about others' knowledge and intent. Dev Sci. 2012;15(3):436–47. pmid:22490183
- 20. Sobel DM, Kushnir T. Knowledge matters: how children evaluate the reliability of testimony as a process of rational inference. Psychol Rev. 2013;120(4):779–97. pmid:24015954
- 21. Harris PL, Koenig MA, Corriveau KH, Jaswal VK. Cognitive foundations of learning from testimony. Annual Review of Psychology. 2018;69;1.1–1.23.
- 22. Stephens E, Suarez S, Koenig M. Early testimonial learning: monitoring speech acts and speakers. Adv Child Dev Behav. 2015;48:151–83. pmid:25735944
- 23. Hawley K. Partiality and prejudice in trusting. Synthese. 2012;191(9):2029–45.
- 24.
Anscombe GEM, Geach M, Gormally L. Faith in a hard ground: essays on religion, philosophy, and ethics. Exeter, U.K.; Charlottesville, Va.: Imprint Academic; 2008. xxvi, 273 p. p.
- 25. Baier A. Trust and Antitrust. Ethics. 1986;96(2):231–60.
- 26. Holton R. Deciding to trust, coming to believe. Australian Journal of Philosophy. 1994;72:63–76.
- 27.
Marušić B. Evidence and agency: norms of belief for promising and resolving. Oxford: Oxford University Press; 2015. xi, 239 pages p.
- 28. Kidd C, Palmeri H, Aslin RN. Rational snacking: young children's decision-making on the marshmallow task is moderated by beliefs about environmental reliability. Cognition. 2013;126(1):109–14. pmid:23063236
- 29. Mischel W, Ebbesen EB, Zeiss AR. Cognitive and attentional mechanisms in delay of gratification. Journal of Personality and Social Psychology. 1972;21(2):204–18. pmid:5010404
- 30. Michaelson LE, Munakata Y. Trust Matters: Seeing how an adult treats another person influences preschoolers' willingness to delay gratification. Dev Sci. 2016;1–9.
- 31. Ma F, Chen B, Xu F, Lee K, Heyman GD. Generalized trust predicts young children's willingness to delay gratification. Journal of Experimental Child Psychology. 2018;169;118–125. pmid:29357990
- 32. Liu D, Vanderbilt KE, Heyman GD. Selective trust: children's use of intention and outcome of past testimony. Dev Psychol. 2013;49(3):439–45. pmid:23339589
- 33. Hetherington C, Hendrickson C, Koenig M. Reducing an in-group bias in preschool children: the impact of moral behavior. Dev Sci. 2014;17(6):1042–9. pmid:24836151
- 34. Brooker I, Poulin-Dubois D. Is a Bird an Apple? The Effect of Speaker Labeling Accuracy on Infants' Word Learning, Imitation, and Helping Behaviors. Infancy. 2013;18(Suppl 1):E46–E68. pmid:26807045
- 35. Landrum AR, Pflaum AD, Mills CM. Inducing Knowledgeability From Niceness: Children Use Social Features for Making Epistemic Inferences. Journal of Cognition and Development. 2016;17(5):699–717.
- 36. Blake PR, Corbit J, Callaghan TC, Warneken F. Give as I give: Adult influence on children's giving in two cultures. J Exp Child Psychol. 2016;152:149–60. pmid:27552298
- 37. Olson KR, Spelke ES. Foundations of cooperation in young children. Cognition. 2008;108(1):222–31. pmid:18226808
- 38. Berg J, Dickhaut J, McCabe K. Trust, Reciprocity, and Social History. Games and Economic Behavior. 1995;10:122–42.
- 39. van den Bos W, van Dijk E, Westenberg M, Rombouts SA, Crone EA. Changing brains, changing perspectives: the neurocognitive development of reciprocity. Psychol Sci. 2011;22(1):60–70. pmid:21164174
- 40. Dunfield K, Kuhlmeier VA, O’Connell L, Kelley E. Examining the Diversity of Prosocial Behavior: Helping, Sharing, and Comforting in Infancy. Infancy. 2011;16(3):227–47.
- 41. Warneken F, Tomasello M. The emergence of contingent reciprocity in young children. J Exp Child Psychol. 2013;116(2):338–50. pmid:23917162
- 42. Moore C. Fairness in children's resource allocation depends on the recipient. Psychol Sci. 2009;20(8):944–8. pmid:19515118
- 43. Landrum AR, Mills CM, Johnston AM. When do children trust the expert? Benevolence information influences children's trust more than expertise. Dev Sci. 2013;16(4):622–38. pmid:23786479
- 44. Vanderbilt KE, Liu D, Heyman GD. The development of distrust. Child Dev. 2011;82(5):1372–80. pmid:21824130
- 45. Moran R. Getting told and being believed. Philosopher’s Imprint. 2005;5(5):1–29.
- 46. Brosseau-Liard P.E., & Birch S.A. Epistemic states and traits: Preschoolers appreciate the differential informativeness of situation-specific and person-specific cues to knowledge. Child Development. 2011;82(6);1788–1796. pmid:22004452
- 47. Jaswal VK, Croft AC, Setia AR, Cole CA. Young children have a specific, highly robust bias to trust testimony. Psychol Sci. 2010;21(10):1541–7. pmid:20855905
- 48.
Koenig MA, McMyler B. Testimonial knowledge: understanding the evidential, uncovering the interpersonal. The Routledge Handbook of Social Epistemology Taylor & Francis: New York, NY. Forthcoming.
- 49. Jaswal VK, Kondrad RL. Why Children Are Not Always Epistemically Vigilant: Cognitive Limits and Social Considerations. Child Development Perspectives. 2016;10(4):240–4.
- 50. Nielsen M. Copying actions and copying outcomes: social learning through the second year. Dev Psychol. 2006;42(3):555–65. pmid:16756445
- 51. Uzgiris IC. Two functions of imitation during infancy. International Journal of Behavioral Development. 1981;4:1–12.
- 52. Heyman GD, Sritanyaratana L, Vanderbilt KE. Young children's trust in overtly misleading advice. Cogn Sci. 2013;37(4):646–67. pmid:23294130