Anchoring Effect of Performance Feedback on Accuracy of Metacognitive Monitoring in Preschool Children

Overconfidence at the Preschool Age

From the developmental perspective, preschoolers are overconfident when predict their future performance (Lipko et al., 2009) and also when they judge their past performance (Destan et al., 2014; Urban & Urban, 2020). However, the overall overconfidence decreases with increasing age (O’Leary & Sloutsky, 2017; Roebers, 2002).

Preschoolers, unlike older children and adults, do not become underconfident with practice and remain overconfident over several exam sessions (Lipko et al., 2009, 2012). Underconfidence with practice is a phenomenon in which people modify previously overconfident judgments so they become underconfident after a first study-test trial (Koriat, Sheffer, & Ma’ayan, 2002). There are, however, several factors affecting overconfidence rates even at preschool age.

Present Study

Because of the domain specificity of metacognitive monitoring at preschool age, we still know little about children’s predictions and postdictions on solving different kind of tasks. Most research on preschoolers’ monitoring ability has used memory tasks, such as images (Destan et al., 2017; van Loon et al., 2017) or vocabulary learning (Geurten & Meulemans, 2017; Lipowski et al., 2013). In the present research, we aim to explore the effect of performance feedback using two analogical reasoning tasks. Experiment 1 focuses on numerical orders (number patterns), the symbolic arithmetic task, where children make number sequences by arranging numbers in a specific order following an example. Experiment 2 concerns patterns of objects, the non-symbolic task. The objects are arranged in a particular order and the child has to identify the pattern and then predict what follows. Numerical orders and object patterns belong to basic mathematical skills on which other mathematical concepts are based and are therefore important in later education (see Anderson & Cordes, 2013; Hawes, Nosworthy, Archibald, & Ansari, 2019; Mundy & Gilmore, 2009; Seo & Ginsburg, 2004; Xenidou-Dervou, Molenaar, Ansari, van der Schoot, & van Lieshout, 2017).

The present study has three main objectives. The first is to investigate whether preschoolers’ predictive judgments become more accurate following performance feedback and having solved several items in subsequent learning trials (Finn & Metcalfe, 2008).

The second objective concerns the effect of performance feedback on monitoring of incorrect answers. Preschool children who do not receive feedback tend not to remember incorrectly solved items (Finn & Metcalfe, 2014) and are unable to utilize their recent experience to produce more adequate judgments on subsequent performance (Lipko et al., 2009). The aim is to examine whether preschool children are able to reduce certainty about future performance (predictive judgments) after incorrectly solved items and receiving negative performance feedback.

The third objective concerns the anchoring effect of performance feedback. In research by Destan and colleagues (2017), the postdictive judgment was made following performance feedback with performance feedback having an immediate effect on the judgment. Children did not rely on their own task experience, but followed the externally administered feedback. Furthermore, van Loon et al. (2017) found that 5-year-old children did not rely on item difficulty when making judgments after receiving performance feedback. This finding was further tested by Geurten and Meulemans (2017) with similar results. Our aim is to test whether performance feedback has a stronger effect on the accuracy of postdictive judgments than task experience does. If the link between performance feedback and postdictive judgment is stronger than the link between judgment and task experience that would support the hypothesis that performance feedback acts as an anchor.

Method

Procedure

The trained experimenter tested the children individually in a quiet room in their preschool. Children were randomly assigned to two groups: Performance Feedback (PF); and No Feedback (NF). The detailed description of procedure can be found in Supplementary Materials (Appendix B).

The children indicated their answers by pointing to the answer sheet. After each item had been completed, the experimenter asked the child to make a postdictive judgment about his or her performance on a 5-point scale from “I am very unsure that I solved the item correctly” to “I am very sure that I solved the item correctly”.

If the child had been assigned to the PF group then after making postdictive judgment, a child received feedback about whether the answer was correct. The experimenter then asked the child to make a predictive judgment about how he or she would perform on the next item from “I am very unsure I will solve the next item correctly” to “I am very sure I will solve the next item correctly” (see Figure 1). If the child had been assigned to the NF group, once the child had made postdictive judgment the experimenter immediately asked on the predictive judgment.

Click to enlarge

Figure 1

Research Procedure for NF and PF Groups

Note. The arrows indicate the hypothesized effect of the performance feedback investigated in the research.

Results

Descriptive statistics for the mean values of Performance, Predictive Judgments and Postdictive Judgments can be found in Table 1.

Discussion on Experiment 1

Preschool age children are often inaccurate when judging their learning performance. When making predictions about future performance, preschoolers do not become underconfident with practice and are overconfident across several trials (Lipko et al., 2009). The current study supports these findings: the preschoolers who received no feedback did not become more accurate in their predictions across items. However, children who received performance feedback became more accurate in their subsequent predictions (Lipowski et al., 2013). Moreover, children in the feedback group lowered their future expectations after they had received performance feedback and were less sure in their subsequent judgment. Our research supports the previous conclusion that children usually predict they will get the answer right, but if they receive negative information about their performance (negative performance feedback following an incorrect answer), they are less certain than children who do not receive feedback (Kvavilashvili & Ford, 2014).

Moreover, preschool age children can use the external information about their performance as an anchor not just for the next prediction but also for subsequent postdictive judgments. They utilized the performance feedback from the previous item more than they utilized their own task experience in the subsequent item. This finding is in line with Finn and Metcalfe (2007, 2008) who showed that children rely on the explicit information given by the experimenter and not on their previous performance. In addition to Geurten and Meulemans (2017) who studied the anchoring effect of feedback on predictive judgments, results of the current study support the anchoring effect of performance feedback on postdictive judgments of preschool children.

Task difficulty could affect the accuracy of metacognitive monitoring at this age. Preschool children seem to find it difficult to monitor more cognitively demanding tasks (Destan et al., 2017; Kvavilashvili & Ford, 2014). In the case of difficult tasks children from the age of 4 use a perceived task difficulty as a guide in their judgments instead of actual task experience (Geurten, Willems, Germain, & Meulemans, 2015). We therefore conducted Experiment 2 based on object pattern tasks. Object patterns are also part of basic math skills but different to the number patterns in nature (Resnik, 2000; see also Dehaene et al., 1998). Ability to solve a nonsymbolic task (Experiment 2) does not depend on the ability required to perform the symbolic arithmetic task (Experiment 1). Solving the arithmetic task depends on understanding of abstract representations of numbers, which tends to develop from school age, whereas nonsymbolic knowledge develops earlier, at preschool age, influencing the ability of preschoolers to accurately monitor performance in these tasks (Barth, La Mont, Lipton, & Spelke, 2005; Rubinsten, Henik, Berger, & Shahar-Shalev, 2002).

Method

Results

Descriptive statistics for the mean values of Performance, Predictive Judgments and Postdictive Judgments can be found in Table 2.

Discussion on Experiment 2

In Experiment 2 the children solved reasoning task involving patterns of objects. Similarly as in Experiment 1 the preschool children made significantly more accurate predictions about their future performance after receiving performance feedback (Geurten & Meulemans, 2017). Performance feedback also had an effect on uncertainty monitoring, with children who received feedback after an incorrect answer lowering their future expectations. Overall improvement in monitoring accuracy following feedback provides further evidence of the importance of external cues in monitoring development at preschool age (Urban & Urban, 2018, 2020).

Similarly as in Experiment 1, the examination of the postdictive judgments in Experiment 2 provided further support that performance feedback acts as an anchor. The performance feedback effect was transferred to the judgment regardless of task experience. In other words, children relied on the external information rather than internal cues (van Loon et al., 2017).

Moreover, the children that received no feedback in Experiment 2 were able to utilize the internal cues from task experience; however, once they had received performance feedback, they relied on an external cue. This result can be related to the nature of the task since object pattern task employed in Experiment 2 is developmentally more appropriate than the number pattern task used in Experiment 1 (Barth et al., 2005). This result can also provide further support for the conclusion that preschool children find highly cognitively demanding tasks more difficult to monitor (Destan et al., 2017; Kvavilashvili & Ford, 2014) and that it is important to maintain a level of task difficulty that is developmentally adequate so the children remain motivated to proceed (Geurten et al., 2015; Schneider, 1998; Shin, Bjorklund, & Beck, 2007).

Effect of Performance Feedback on Monitoring Accuracy of Predictive Judgments

From the age of 4, children understand that their past performance predicts their future performance and therefore children's second predictions for the same set of tasks are more accurate than the firsts. Moreover, children use their past performance as a cue for subsequent predictions. However, they have no valid cues for making predictions for new set of tasks (Lipko-Speed, 2013). For this reason, in order to foster the accuracy of predictive judgments, Geurten and Meulemans (2017) decided to provide an external cue to the preschoolers. They adapted calibration feedback consisting of information about performance as well as about the monitoring accuracy. However, young children that are not yet able to process ternary relations (Halford & Andrews, 2014) or the low-performers (Urban & Urban, 2018, 2020) may have difficulties to take into account this rather complex external cue, but they are able to improve their monitoring accuracy when performance feedback is provided.

In both experiments conducted in this study, the effect of performance feedback on the overall accuracy of predictive judgments was large ( ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .140 and ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .139). The finding that preschool children ameliorated their predictions in line with feedback is consistent with previous research findings (Lipowski et al., 2013). More importantly, in both experiments the preschoolers were able to utilize performance feedback to reduce certainty about future performance after incorrectly solving an item. The ability to monitor uncertainty is important for subsequent regulation of actions, for example modifying learning strategy, like skipping an item or seeking help and plays an important role in self-regulation of learning in primary education (Coughlin et al., 2015; Destan et al., 2017; van Loon et al., 2017).

Anchoring Effect of Performance Feedback Using Postdictive Judgments

The second aim of the study was to test whether performance feedback acts as an anchor on postdictive judgments. The anchoring effect is the adjustment of estimation based on previously presented external information, an anchor (Bahník, Englich, & Strack, 2017). Geurten and Meulemans (2017) conducted a series of experiments and found that children who were more sensitive to feedback effect were also more sensitive to the anchoring effect. More importantly, they concluded that feedback acts as an anchor on predictive judgments regarding memory tasks, because the children in their research relied on performance feedback rather than on task difficulty. This result is in line with research by van Loon and colleagues (2017) who studied the utilization of internal and external cues in making postdictive judgments in memory tasks. They found preschoolers who received no feedback used task difficulty as an internal cue when making postdictive judgments; however, when feedback was received, the external information became more important than task experience.

Both the experiments conducted in this study support these conclusions: the preschool children making postdictive judgments relied more on performance feedback than on task experience and thus, by definition (Tversky & Kahneman, 1974), performance feedback acted as an external cue serving as an anchor for the children’s judgments. The cue utilization hypothesis (Koriat, 1997) assumes that there is always an interplay between several internal, external or mnemonic cues which accounts for the accuracy of judgments. Roebers et al. (2019) stated that cue utilization is a factor driving recognition improvements over time. However, this theory cannot explain why an individual making judgments decides to rely on one particular cue and not another (Koriat, 2016; Winne, 2011).

Therefore to understand why the preschoolers chose to use the externally administered performance feedback instead of task experience we can turn to theories related to the anchoring effect, such as conversational inferences theory (Grice, 1975). This theory states that people making judgments transfer social rules from everyday life (e.g., trust in the expertise of scientists or superior knowledge of teachers) to the test situation, and so they tend to believe that the feedback (anchor) administered by the experimenter is informed and close to the actual result (Jacowitz & Kahneman, 1995). This theory has been criticised because anchoring effect also occurs in situations where feedback is extremely unlikely (e.g., when using number 212 as an anchor for the age of Mahatma Gandhi; see Bahník, Englich, & Strack, 2017). However, this is not the case with performance feedback in our designs because participants could trust that the experimenter’s feedback was correct and could be relied upon when making subsequent judgments. Raaijmakers, Baars, Schaap, Paas, and van Gog (2017) in a similar setting manipulated performance feedback in order to test the effect of dominantly positive or dominantly negative feedback and none of the participants noticed them doing so. Future research should address whether the perceived validity of judgments relies on social rules by explicitly questioning the validity of the experimenter or feedback. This could examine whether participants rely on feedback even when compromised, administered by an “uninformed” experimenter lacking credibility or that is inconsistent with the participants’ internal cues (see also Baadte & Kurenbach, 2017; Brunot, Huguet, & Monteil, 1999).

Furthermore, the selective accessibility theory of anchoring (Mussweiler & Strack, 1999) states that anchor improves accessibility of topic related knowledge in the memory. Individuals then test whether knowledge in their memory matches the value of the anchor. This is resulting in stronger anchoring effect in unfamiliar situations, where the individual lacks expertise or feels uncertain (Mussweiler & Strack, 2000). Research by Schneider (1998) shows that children’s judgments have higher metacognitive accuracy when they are doing familiar tasks, because they can relate to their previous experience. However, in the research on feedback effect there are often used unfamiliar tasks to ensure the results are unbiased (Geurten & Meulemans, 2017; Urban & Urban, 2019; van Loon et al., 2017). As Winne (2011) pointed out students perform better on unfamiliar tasks when the overall evaluation standards are set externally by the teachers (Hamilton, 1985; Kiewra, Benton, Kim, Risch, & Christensen, 1995), but once students gain experience of the subject they can begin to use task familiarity as an internal cue on their own (Begg, Anas, & Farinacci, 1992). However, the high degree of familiarity with the ideas prevents the feedback from having an effect, which is especially alarming when related to prejudice or stereotype reduction (Swire, Ecker, & Lewandowsky, 2017). Applying this theory to preschoolers, future research should address differences in the effect size of performance feedback in familiar and unfamiliar tasks to see whether as preschoolers gain more experience of the task, the performance feedback loses its’ potential to improve accuracy in metacognitive monitoring.

Both conversational inferences theory and selective accessibility theory of anchoring have therefore practical implications in administration of feedback in educational process. Stipek and Tannatt (1984) found that children from preschool age, perceiving their teachers as authorities, introject teachers’ feedback and use it as the general assessment criteria for their own success. The repetition of information without any feedback can, on the other side, results in the reverse situation when children introject the information itself. Plichtová (2013) described the situation in which students were unable to accept feedback on the familiar concepts and an attempt to administer feedback was perceived as a personal threat. Moreover, Urban (2017) in research with art students described that individuals become reluctant to pursue personal goals when receiving negative feedback from peers or teachers. These results are in line with research on performance feedback in which Venables and Fairclough (2009) concluded that feedback following failure led to adverse changes in mood and motivation and participants were on the verge of abandoning the task. Winne (1996) therefore proposed that it is necessary to create learning environment allowing individuals to decide when to adjust knowledge in line with external feedback and when to rely on their own internal cues, such as personal experience or problem familiarity. In this context, the conceptual connection between anchoring effect theory and feedback administration may bring us closer to understanding the cue selection process that begins at preschool age.