Introduction
Today, with regard to the pervasive availability of technological aids such as smartphones or tablets, individuals can constantly decide between either externalizing cognitive processes into these aids by, for example, offloading a shopping list onto one’s smartphone or relying on their own internal cognitive processing by memorizing the shopping list instead. Technological aids serve as a digital expansion of the individual mind (Clark & Chalmers,
1998) and individuals perceive their external memories as part of themselves (Finley, Naaz, & Goh,
2018). The determinants of utilizing either internal cognitive processes or external cognitive resources have been the focus of recent research (e.g., Gilbert,
2015a; Gray, Sims, Fu, & Schoelles,
2006; Grinschgl, Meyerhoff, & Papenmeier,
2020; Risko & Dunn,
2015; Weis & Wiese,
2018). With the present experiment, we probed whether there is a causal relationship between metacognitive beliefs and offloading behavior by manipulating participants’ metacognitive beliefs about their own working memory performance with fake performance feedback.
The externalization of cognitive processes into technological aids is known as cognitive offloading (Risko & Gilbert,
2016). Cognitive offloading reduces demands on internal cognitive processing and thus minimizes cognitive effort when performing a task. Furthermore, due to cognitive offloading, individuals can store and handle more information simultaneously than within the restrictions of their internal memory capacity. In other words, cognitive offloading allows for overcoming capacity limitations of internal cognitive processing such as in working memory (Risko & Gilbert,
2016). With regard to working memory, cognitive offloading avoids the internal encoding or actively holding of information that is present in the immediate environment (Wilson,
2002). Instead, individuals can rely on the environment, for example, using a technological aid to externally store and/or manipulate information and only access the information when needed (Wilson,
2002).
Over the last years, research has identified multiple determinants for offloading behavior (see Risko & Gilbert,
2016, for a review), such as the characteristics of the technological aid and the task at hand. For example, the likelihood of offloading cognitive processes onto a tablet device depends on the responsivity of the device and the smoothness of the control type (Grinschgl et al.,
2020). Current research suggests that cognitive offloading is based on cost–benefit considerations (e.g., Gray et al.,
2006). When cognitive offloading is associated with low temporal and/or physical costs while interacting with tools, offloading behavior is more pronounced than with high associated costs (e.g., Cary & Carlson,
2001; Gray et al.,
2006; Grinschgl et al.,
2020). Regarding the task at hand, the information that needs to be processed also influences offloading behavior. For instance, increases in complexity (Schönpflug,
1986), difficulty (Hu, Luo, & Fleming,
2019), or amount of information (Gilbert,
2015a; Morrison & Richmond,
2020; Risko & Dunn,
2015) results in an increased offloading behavior.
Recently, researchers interested in cognitive offloading started considering determinants of cognitive offloading related to the user of technological aids, such as users’ memory capacity or metacognitive beliefs about their own internal abilities. Individuals offloading behavior is more pronounced, the lower their own internal performance is (Gilbert,
2015b; Risko & Dunn,
2015; but see Morrison & Richmond,
2020, for conflicting results). Importantly, however, prior research suggests that not only objective memory abilities but also metacognitive beliefs about one’s internal memory abilities and one’s environment might affect offloading behavior (Arango-Muñoz,
2013). In their review article, Risko and Gilbert (
2016) proposed a metacognitive model of cognitive offloading. This model states that the decision between internal and external strategies is guided by metacognitive beliefs about one’s environment—such as the properties of technological aids—and one’s internal memory abilities. Regarding the former, that is, the metacognitive beliefs about one’s environment, studies have shown that individuals adapt their offloading behavior according to their beliefs about the benefits of an offloading strategy (Dunn & Risko,
2015) or the reliability of a technological aid (Weis & Wiese,
2018). If individuals expected an offloading strategy to be inefficient for reaching their goal (Dunn & Risko,
2015) or a technological aid to be unreliable (Weis & Wiese,
2018), they offloaded less and relied more on their own internal resources. Regarding the latter—metacognitive beliefs about one’s internal abilities—Gilbert (
2015b) observed in a prospective memory task that the subjective confidence in one’s memory performance predicted offloading behavior, regardless of objective accuracy. Lower confidence in one’s memory performance (i.e. less positive metacognitive evaluations about one’s memory) was associated with a more extensive use of external reminders, thus more cognitive offloading (Boldt & Gilbert,
2019; Gilbert,
2015b; similar results were obtained by Hu, et al.,
2019; Risko & Dunn,
2015). Therefore, individuals might use cognitive offloading as a compensatory strategy if they believe that their internal memory abilities are poor. In a recent experimental study using the same prospective memory task, Gilbert et al. (
2020) manipulated the difficulty of practice trials as well as the valence of provided feedback on each trial (positive vs. negative). After performing the practice trials, the participants provided metacognitive performance estimations and then performed the task with the possibility of offloading memory demands. The participants rated their own memory performance to be more accurate when they received positively framed feedback or easier practice trials than when they received negatively framed feedback or more difficult practice trials. This shift in metacognitive evaluations was accompanied by a matching shift in offloading behavior. When a manipulation resulted in less confidence in one’s memory abilities, this led to more cognitive offloading. However, all participants showed a bias towards using cognitive offloading extensively, thus metacognitions cannot fully explain offloading behavior in this study (Gilbert et al.,
2020). While these findings are a first indication of the connection between metacognitions and cognitive offloading beyond correlational approaches, further investigations are needed to explain their causal relationship as well as the involved processes.
In the present study, we set out to investigate the causal relationship between metacognitive beliefs and offloading behavior by manipulating metacognitive beliefs with fake performance feedback. Performance feedback can influence motivation (Venables & Fairclough,
2009), effort spent on a task (Raaijmakers, Baar, Schaap, Paas, & Van Gog,
2017) as well as goals (Fishbach, Eyal, & Finkenstein,
2010; Ilies & Judge,
2005), even if the feedback is manipulated and therefore false (Ilies & Judge,
2005). With regard to perceptual learning, fake performance feedback has an even higher impact than genuine feedback (Shibata, Yamagishi, Ishii, & Kawato,
2009). Additionally, positive and negative performance feedback can influence beliefs about one’s self-efficacy (Nease, Mudgett, & Quiñones,
1999). Individuals often evaluate their own performance in comparison to other individuals, such as those in their peer group (Ilies & Judge,
2005; MacFarland & Miller,
1994). Thus, performance feedback including a social comparison (e.g., “you performed worse/better than your peers”) might have a particularly strong effect on metacognitive beliefs. In addition, participants might be less able to judge their own performance in relation to their peers compared to directly estimating their own abilities (without any social comparison). Thus, they might be more vulnerable to fake performance feedback with rather than without social comparisons. For these reasons, we provided the participants of our study with fake performance feedback indicating a below-average or above-average performance compared to their peers (i.e. other students), and we measured participants’ metacognitive evaluations with subjective performance ratings similar to the feedback.
We predicted that fake performance feedback should influence participants’ metacognitive beliefs about their own working memory performance. We further hypothesized that the manipulated metacognitive beliefs should transfer into the control of offloading behavior in a working memory task. Therefore, we expected that the participants receiving below-average performance feedback rely more on cognitive offloading while performing a working memory task than those participants receiving above-average performance feedback, with the control group (i.e. no feedback) in between the two. We expected this effect to be due to metacognitive beliefs about the reliability of the internal working memory resources. Whereas the participants receiving below-average feedback should expect their memory to be poor, thus relying more on offloading, those participants receiving above-average feedback should expect their memory to be good, thus relying more on internal processing.
Discussion
In the present study, we investigated the causal impact of metacognitive beliefs about one’s working memory on cognitive offloading in a working memory task. Metacognitive beliefs are supposed to influence one’s decision for using specific strategies when performing a task based on metacognitive monitoring and control. Thus, metacognitive beliefs should affect the use of technological aids (and likewise cognitive offloading) or one’s internal working memory resources. To experimentally test the determining role of metacognitive beliefs when offloading working memory processes, we used fake performance feedback. Our fake performance feedback successfully manipulated metacognitive beliefs about one’s working memory. Before receiving any feedback, the three feedback groups did not differ in their pre-rating about their upcoming working memory performance, but after receiving fake performance feedback, they differed accordingly. The participants receiving below-average performance feedback rated their working memory performance the lowest, and those participants receiving above-average performance feedback rated their working memory performance the highest, with the control group that did not receive any feedback in between. Remarkably, the effect of fake performance feedback was so strong that it even spilled over to general beliefs about one’s memory abilities (measured by the MMQ) at the end of the experiment. The participants in the below-average group estimated their general memory abilities lower than the other two groups. Thus, especially below-average performance feedback affected metacognitive beliefs broadly and persistently. Although our manipulation of metacognitive beliefs altered the participants’ subjective working memory ratings, it had clearly no impact on offloading behavior within the Pattern Copy Task. Within this task, participants could either rely more on a technological aid by looking up information more often (i.e. more cognitive offloading) or rely more on their own internal memory by looking up the information less often (i.e. less cognitive offloading). We observed that spontaneous offloading behavior within the Pattern Copy Task was nearly identical across all feedback groups.
Previous research suggests that metacognitions are negatively correlated with offloading behavior (Boldt & Gilbert,
2019; Gilbert,
2015b; Hu, et al.,
2019; Risko & Dunn,
2015). For instance, in studies applying a prospective memory paradigm (Gilbert,
2015b; see also Boldt & Gilbert,
2019; Gilbert et al.,
2020), the participants had to drag circles with ascending numbers one after another to the bottom of the screen. At the beginning of a trial, the participants were instructed that some special circles (e.g., the circle with the number 3) had to be dragged to another side of the screen (e.g., the left side) when it was their turn. These special circles induced intentions that the participants needed to fulfill later on. After performing practice trials, the participants were asked to rate their upcoming performance (0–100% of special circles dragged to the correct location). The participants then performed several trials of the task without the option to offload, followed by several trials that allowed cognitive offloading. In these latter trials the participants could offload the intentions by placing the special circles close to the correct side of the screen already at the beginning of a trial. More positive evaluations about one’s unaided memory performance were associated with less cognitive offloading (Boldt & Gilbert,
2019; Gilbert,
2015b; see also Hu et al.,
2019; Risko & Dunn,
2015, for similar results). While these correlational findings suggest a relationship between metacognitions and offloading behavior, we did not observe a matching impact of metacognitions on cognitive offloading in the present experimental study. To resolve these seemingly conflicting results, we suggest that the differentiation of metacognitions into metacognitive beliefs and metacognitive experiences might explain the diverging results across studies.
In the present study, we manipulated metacognitive beliefs by providing the participants with fake performance feedback on three different working memory tasks but—importantly—before they gained any actual experience in performing the Pattern Copy Task. In contrast, previous research reporting significant effects of metacognitions on offloading behavior collected metacognitive performance estimations after participants performed practice trials of the offloading task (Boldt & Gilbert,
2019; Gilbert,
2015b) or the presentation of the relevant stimuli (Hu, et al.,
2019; Risko & Dunn,
2015). Theoretical accounts of metacognitions often differentiate between metacognitive beliefs—referring to general beliefs about one’s person stored in long-term memory (e.g., beliefs about one’s memory abilities)—and metacognitive experiences—referring to task-specific knowledge that is present before, during, or while performing a cognitive task (Efklides,
2008; Flavell,
1979). Metacognitions manipulated in our study were supposed to reflect the former—general beliefs about one’s working memory based on fake performance feedback in other tasks. On the other hand, metacognitions measured in the previous studies (e.g., Gilbert,
2015b) rather reflect the latter—metacognitive experiences—due to the measurement after performing practice trials, for instance. This latter design was also used in a recent study showing that the manipulated valence of feedback on task trials influenced metacognitions and in return offloading behavior (Gilbert et al.,
2020). Therefore, we suggest that it might actually be metacognitive experiences (as measured by the previous studies) rather than metacognitive beliefs (as manipulated in our study) that drive offloading behavior.
The suggestion that metacognitive experiences rather than metacognitive beliefs alter offloading behavior fits in well with research showing that actual offloading behavior is determined by the properties of the task at hand and thus probably metacognitive experiences. In like manner, cognitive offloading is known to be driven by external factors such as tool design (Grinschgl et al.,
2020) costs when interacting with external tools (e.g., Cary & Carlson,
2001; Gray et al.,
2006; Grinschgl et al.,
2020), or characteristics of processed information (Gilbert,
2015a; Hu et al.,
2019; Morrison & Richmond,
2020; Risko & Dunn,
2015; Schönpflug,
1986). Such external factors are likely to influence metacognitive experiences while performing a task and in turn influence offloading behavior. Within this context, the new finding of our study is that metacognitive beliefs in contrast to metacognitive experiences had no influence on offloading behavior—at least not within the Pattern Copy Task.
Interestingly, we observed an influence of fake performance feedback on subjective judgements regarding the offloading strategy in the Pattern Copy Task. The participants in the below-average group were more likely to report an offloading strategy over an internal strategy than the participants from the above-average group, although their actual offloading behavior was nearly identical. The distinction between metacognitive beliefs and metacognitive experiences also provides a way to resolve the apparent contradiction between perceived and actual strategy use. Whereas metacognitive experiences could be the main determinant of actual offloading behavior in the Pattern Copy Task, participants might rather consider their metacognitive beliefs when giving subjective judgements on their behavior. For instance, negative beliefs about one’s performance might lead participants to judge their behavior as offloading more (although they actually did not offload more) than positive beliefs about one’s performance. Thus, based on metacognitive beliefs, the same actual behavior might be interpreted differently by the participants. This assumption was further supported by exploratory correlations showing that the reported strategy selection did not correlate with the actual offloading behavior in the below-average group as well as in the above-average group across most offloading-variables. Interestingly, however, in the control group we did indeed observe such a correlation; that is, participants that reported to have used an offloading strategy also offloaded more within the Pattern Copy Task. Thus, without a manipulation of metacognitive beliefs with fake performance feedback, participants could correctly judge their own performance.
When relating the findings of our present study to previous research, it is also important to consider the differences between the offloading tasks applied. For instance, in a prospective memory task that has established a correlation between metacognitions and cognitive offloading (Boldt & Gilbert,
2019; Gilbert,
2015b), the participants offloaded future intentions, whereas in the Pattern Copy Task the participants offloaded by looking up relevant information. These two kinds of offloading behavior might be different per se, thus also be guided by different determinants. It is possible that fake performance feedback and in return metacognitive beliefs would indeed drive the offloading of intentions in a prospective memory task, but not offloading behavior in the Pattern Copy Task. We can only speculate about the different processes involved in these offloading paradigms as no study has directly compared them. However, one important difference might be the involved timing when offloading memory processes. Whereas in studies using the prospective memory task (Boldt & Gilbert,
2019; Gilbert,
2015b; Gilbert et al.,
2020), the participants offloaded future intentions (i.e. the information is offloaded for remembering it later on), in the Pattern Copy Task the participants offloaded information for instantaneous use (i.e. looking up information more often for the ongoing copy task). Thus, offloading of future intentions might be related to planning before actual task performance, while offloading in the Pattern Copy Task might be related to ongoing processes throughout the task.
4 Metacognitive beliefs could possibly play a greater role for planning before action (thus affecting the offloading of intentions) rather than for offloading during ongoing task processing. Further research is needed to investigate the different as well as shared processes involved in cognitive offloading across various paradigms.
Another difference between previous studies investigating metacognitions as determinant of cognitive offloading (e.g., Gilbert,
2015b) and the present study is the specific framing of participants’ performance estimations. While in previous studies the participants estimated their own performance based on how accurate they think their own performance is (0–100% accuracy; Boldt & Gilbert,
2019; Gilbert,
2015b; Gilbert et al.,
2020; Risko & Dunn,
2015), in our experiment they estimated their performance in comparison to other students via a percentile rank. This latter estimation in our study was in line with the provided fake performance feedback that was designed to have a strong impact due to social comparisons (MacFarland & Miller,
1994). It might be argued that cognitive offloading was guided rather by metacognitive beliefs without any comparison (i.e. the participants might adopt their offloading behavior based in their confidence in their own memory, independent of its relation to other individuals). However, in our study we also measured metacognitive beliefs with the MMQ that did not include any estimations compared to other individuals and—importantly—our manipulation also affected the metacognitive beliefs as measured in this questionnaire following below-average performance feedback. We thus consider it unlikely that the specific framing of the fake performance feedback as well as performance estimations was a key factor in explaining differences in results between our present study and previous research on cognitive offloading.
The participants in the below-average group estimated their general memory abilities lower than the other two feedback groups. Thus, it seems that below-average performance feedback has a particularly strong influence on metacognitive beliefs and self-perception. In a similar vein, Davis and Brock (
1975) showed that below-average performance feedback influences the participants’ self-awareness compared to no feedback or above-average feedback, while the latter two conditions did not differ from each other. However, it might not only be below-average feedback per se that strongly influences metacognitive beliefs. Another possibility could be that below-average feedback induces a large deviation from one’s primary beliefs before receiving feedback. For instance, one might think that his or her performance is slightly above average. In this case, receiving above-average feedback suggesting a percentile rank of 79% might be less unexpected and thus have less impact than below-average feedback suggesting a percentile rank of 21%, which might largely deviate from one’s primary beliefs. Nonetheless, our findings suggest that below-average performance feedback is particularly suited to experimentally manipulate the participants’ self-perception – an important insight for future experiments.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.