Children acquire a large proportion of their knowledge through the internet and social media (Livingstone, 2009). This has been particularly true during the COVID-19 pandemic where many school students are learning exclusively through remote means (e.g., Fiialka, 2020; Garbe et al., 2020; Mabeya, 2020; UNESCO, 2020). Children are actively encouraged by educators to research school projects, and children also rely on the internet for personal matters. Although the internet contains many exceptional websites whose content has been accurately curated, many websites do not contain credible information and/or may have been authored by people who are not experts in a given area. How susceptible are children to accepting such misinformation? Are children able to detect ‘fake news’ and ‘fake content’ on the internet? The question is not trivial—computers are used on all continents of the world in education and the amount of available and accessible information is not likely to reduce in the future. It is essential that the current generation of children are helped to be intelligent consumers of information, and to be able to detect misinformation in websites. In two studies, children aged 7- to 12-years used credible and not credible websites to learn about environmental sustainability. There was little evidence of children spontaneously evaluating the credibility of the websites, but 10- to 12-year-olds could ‘edit out’ fake information from their knowledge when explicitly asked to. The results suggest that we cannot let this opportunity to help children resist ‘fake news’ and ‘fake information’ pass by. Researchers should aim to discover more about what processes children need to master in order to be intelligent consumers of digital information.
Children learn from a multitude of sources, for example, teachers, parents, peers, books, television programming, and the internet. Increasingly, school children are encouraged to research a topic by searching the internet for sources of information, that is, websites. In the preschool and early elementary grades when children are aged 4- to 8-years-old, many children (particularly within North America) are shown how to use the internet for research, but their searches are largely guided by the websites chosen by their educators (e.g., Faulk & Evanshen, 2013; the Kindergarten Program, 2016; Long Beach Unified School District, 2013). In contrast, in later grades, when children are aged 9- to 10-years and older, they are expected to take the initiative of finding websites themselves to research a topic.
The sheer amount of information available on the internet is astounding, however, the quality of information varies widely. To learn about a medical condition, for example, sources range from highly credible sites like those of the National Institutes of Health or the Mayo Clinic, to websites listing personal experiences as if they were generalized facts, as well as patently false information (e.g., vaccine risks). To maintain the integrity of our knowledge, we must actively evaluate the credibility of each website and the information it contains. Otherwise, we risk having a knowledge base that is inaccurate. In the case of children, without proper constraint, we risk an entire generation that is misinformed or, at the very least, has compromised knowledge. Rather than analyzing children’s internet search strategies, the purpose of the current study was to assess whether children use information about a website’s credibility when they are asked about information from internet sources.
There are two issues that impact the quality of information gleaned from the internet. The first is the accuracy of source monitoring (Johnson et al., 1993), which refers to knowing where (i.e., from what source) information originated. The second issue refers to the role of source monitoring during active learning. We discuss each in turn.
The development of source monitoring
There are many instances when we confuse where we have learned information (see Roberts, 2002, for a review). For example, people can blend or confuse memories of live events and stories (Thierry, 2009), film and narrative about film (e.g., Ackil & Zaragoza, 1995, Thierry & Pipe, 2009), different instances of a similar event (Brubacher et al., 2011; Connolly & Lindsay, 2001; Powell et al., 1999; Zhang et al., 2019), or real-life and suggestions about real-life events (Welch-Ross, 1999). The theoretical role of source monitoring has been used to develop our understanding of eyewitness suggestibility (Gudjonsson et al., 2016), theory of mind development (Bright-Paul et al., 2008), and language devices such as evidentiality in Korean (Papafragou et al., 2007) and Turkish (Lucas et al., 2013) children. Finally, source-monitoring processes have been identified as significant cognitive factors for children with autism (Spitzer et al., 2017), fetal alcohol spectrum disorders (Kully-Martens et al., 2012), and phobias (Klein et al., 2014).
Many factors are responsible for source errors. We are more likely to be confused, for example, between two similar versus dissimilar sources (Lindsay et al., 1991; Roberts & Blades, 1999), when we experience an event repeatedly (Brubacher et al., 2011), and when substantial time has passed and source information is forgotten or cannot be retrieved (Roberts & Powell, 2007). Importantly, source monitoring requires both memorial and metamemorial capacity as it involves reasoning about sources based on the quality of our memories or other knowledge. Non-memorial factors such as executive function (working memory, inhibitory control, set switching, and self regulation) also play an important part. In one study, for example, working memory was related to source monitoring when children aged 4- to 8-years-old were asked to distinguish between a science demonstration and a slide show (Earhart & Roberts, 2014). Even children as young as 2.5- to 3-years-old showed relations between conflict inhibition and source monitoring when children had to decide whether they themselves or a confederate placed animal pieces on a farm (Hala et al., 2016).
Not surprisingly, then, source monitoring is a skill that has a protracted development. Evidence of early source monitoring is concurrent with the accelerated development of the frontal lobes (i.e., the early preschool period; Drummey & Newcombe, 2002). Improvements are seen in both accuracy and breadth in the different types of source decisions until, at least, the early teenage years (ages 10–14; Menon et al., 2005; Raj & Bell, 2010; Ruffman et al., 2001) although most research has been focused on young children aged 10-years and younger (Roberts, 2002).
Understanding the development of source-monitoring skills from basic to adult competencies is informed by Johnson’s Multiple-Entry Modular Memory model (MEM; Johnson et al., 1993; Mitchell & Johnson, 2009). The model outlines two sets of processes: a perceptual set and a reflective set. According to the model, a source is attributed at retrieval as the result of these processes. Memories containing high levels of perceptual detail can lead us to assume that an event was observed rather than imagined (because we assume that such detail would not be present in a memory of an imagined event). Sometimes these decisions are made effortlessly and without awareness. At other times, we need to be more intentionally reflective to attribute a source correctly (e.g., I couldn’t have seen it because I wasn’t there). Using these processes enables us to distinguish between ‘internal’ sources (those originating from the self; e.g., thinking, dreaming, self-performed actions) and ‘external’ sources (originating from outside of the self; e.g., observing, hearing).
In the case of learning from websites, however, the source is always external to the self, as websites are presented on a screen (e.g., computer display, smartphone, tablet). Further, although the content and look of websites may differ, the actual medium through which the information is delivered (i.e., the screen) is nearly identical (and not distinctive) when comparing multiple websites. Imagine, for example, a person searching for a topic and then clicking on links to different websites. The websites are listed in the same search, they are both accessed in the same study session, presented on the same device, and so on. Given the similarity of the sources of websites, then, we would expect that a significant amount of development in source monitoring would have to be achieved before it is possible to successfully distinguish websites and information gleaned from them. As children progress through the school grades, there is increasing responsibility placed on them to independently use the internet to find sources of information (children aged 9-years and older).
It has not yet been demonstrated, however, whether children of this age are capable of judging the credibility of websites to maintain integrity of their knowledge. In contrast to much prior research where children choose directly between two sources at the test (e.g., Was that in the video or the story? Earhart & Roberts, 2019), using website credibility information involves more indirect use of sources. In other words, children must use source information in the process of choosing and reporting information. Given the difficulty of distinguishing between two different websites, it is important to know whether children can use source information in this way, and subsequently build a reliable and credible knowledge base. Can children later edit out their knowledge, for example, information that they read on a website that contained errors? Can children identify credibility markers spontaneously? Can children use credibility information with adult guidance? And what is the developmental pattern and timing of these processes? We sought to answer these questions in the current set of studies.
The role of binding processes in source monitoring
It has been demonstrated many times that deficits in source monitoring do not simply reflect a loss of memory (see Raj & Bell, 2010, for a review; Roberts et al., 2016). In studies of repeated-event memory, for example, children can retrieve details from 3 out of 4 different instances of an event (Brubacher et al., 2018). Despite recalling the details, however, children often attribute particular details to the wrong instances (e.g., claiming they did a puzzle during the second instance, when it was actually the third; Brubacher et al., 2018).
Source-monitoring errors in school-aged children are also not caused by a lack of awareness that information is gleaned from different sources. Although preschoolers are still developing the understanding that all knowledge originates from a source (Wimmer et al., 1988), children aged 7-years and above clearly show that they understand the separation of one event from another (Brubacher et al., 2011) and can sometimes identify the differences between them (Brubacher et al., 2011; Danby et al., 2017; Roberts et al., 2015).
An increasingly popular explanation involves the role of binding processes in source monitoring (Bemis & Leichtman, 2019; Burns et al., 2016; Kovacs & Newcombe, 2006; Lloyd et al., 2009; Roberts et al., 2017; Sluzenski et al., 2006). In the work by Newcombe and colleagues, for example, there were few age differences when children identified objects or contextual information (location, scenes) they had seen before, but children aged 3–4 showed impairments relative to older children when recognizing item and context information concurrently (e.g., Did you see the pig in this square?). Newcombe and colleagues argue that these results reflect the difficulty that young children have when binding different aspects of experiences together, in this case, the pig and its location. A binding effect was also observed in a study on children’s episodic memory. Specifically, Roberts et al. (2016) found that increased memory for the details of two different events actually corresponded with increased source confusion. When the source information was used as a cue to promote reflective attribution, however, source monitoring was improved and the children were less confused (e.g., “Remember that when you wore the cape it was the time that I was the doctor”). According to the MEM model, binding content with source is important in source monitoring because the perceptual (e.g., Gestalt processes) and reflective (e.g., noting relations between stimuli) processes act on information to bind content and its context together (Johnson et al., 1993).
Impairments in binding in young children also make sense in the context of other cognitive developments. Between 3 and 6 years of age, substantial progress is made in tasks involving working memory and executive processes (Kanakogi et al., 2012; Zelazo & Muller, 2002). Both the episodic buffer (Baddeley, 2000), which binds features from short- and long-term memory, and the MEM framework suppose that executive processes direct attention and resources to particular features—those most characteristic of the to-be-remembered stimuli (Johnson et al., 1993; Ruffman et al., 2001). Some features are processed at the expense of others, so children who are still strengthening their working memory and executive systems may not be able to adequately direct their attention to and encode those features most relevant for later source attribution.
Interestingly, source monitoring, executive functions (e.g., working memory), and binding processes have been localized in similar brain regions, and development of these neural structures coincides with known developmental patterns of cognitive functioning. Much of the neurological evidence regarding the development of source monitoring comes from studies of aging because older adults tend to show lower accuracy when monitoring sources than do younger adults (i.e., paralleling comparisons between young children and adults). For example, Glisky et al. (2001) showed that older adults were less accurate when monitoring sources than younger adults only when the seniors had below average frontal function. In investigations of episodic memory with adults, activation of the medial temporal lobe (including the hippocampal regions and the amygdala) is correlated with item and source memory, and the hippocampus is particularly activated when correct source judgments are made (Davachi et al., 2003). Lesion (e.g., Cabeza et al., 2008), ERP (e.g., Wilding & Rugg, 1996), and fMRI studies (e.g., Nolde et al., 1998) provide converging evidence. From fMRI data with older adults in a correct rejection task, brain activations and interactions decreased in the inferior frontal gyrus, supramarginal gyrus, and hippocampus (Tsukiura et al., 2014). Tsukiura and colleagues suggested that the ventral prefrontal region, which is involved in source monitoring, and the inferior parietal region, which is associated with recollection by cooperating with the hippocampus, led to confusion between old and novel stimuli. Parallel neural investigations with children are sparse although the evidence to date reveals a similar neurological profile. For example, Ghetti and colleagues, in their study of middle childhood (aged 6- to 10-years), found that hippocampal and parahippocampal regions of the medial temporal lobe were associated with the retrieval of source-specifying information (Ghetti et al., 2010). Finally, children with autism present with abnormalities of the frontal lobes and related structures and behaviorally show impaired performance on source monitoring and executive function tasks (Spitzer et al., 2017).
The role of source monitoring and learning
Although source confusions can sometimes be troublesome (e.g., when they affect personal relationships, or a court case hinging on eyewitness testimony), confusing sources may be beneficial in other cases. Specifically, there are demonstrations of an inverse relationship between item and source memory (Ratner & Foley, 2020; Roberts et al., 2016; Sommerville & Hammond, 2007). Ratner and Foley found that when a child and an adult jointly perform a task (making a collage), the children who showed the greatest gains in learning (i.e., item memory) were also the most confused about who placed the pieces on the collage (claiming responsibility for placing pieces on the collage that the adult actually placed; also see Ratner et al., 2002). Naturally, teachers often run their classrooms in terms of one topic at a time using various sources to accomplish learning. As described earlier, children’s memory for the details of two similar events was improved when they were explicitly instructed to do so, but this was at the expense of their source monitoring (remembering what happened, but confusing in which of the events it happened; Roberts et al., 2016). One possibility is that presenting similar sources close together leads children to blend the information without reference to source. Indeed, this is one way we can build up knowledge bases in different domains using multiple sources.
The mechanism of blending information from different sources can, therefore, be helpful for learning, but there is a caveat when this blending of sources is applied to learning from the internet. While children may trust their teachers to provide accurate information and teachers can carefully choose which sources to use, information that appears in internet searches is not formally vetted and information consumers must actively judge the credibility of websites. Trusting the information on every website would be naïve and likely lead to inaccuracies or omissions in knowledge. Hence, confusing sources may ordinarily help children to build up a knowledge base; but living in the digital world necessitates that children are also able to evaluate the credibility of some sources, such as websites. When an inaccurate website has been identified, for example, a child must identify the origin or source of the compromised information and filter out those details from their knowledge base.
The current study: learning environmental sustainability information from websites
To investigate whether children can benefit from credibility information about websites, we ran two experiments with 7- to 12-year-olds (an age where most children would be doing at least some independent research using the internet; Holloway et al., 2013). We varied whether children could use credibility information through guidance or spontaneously (without guidance). We also tested whether children can ‘filter’ or ‘edit out’ information that had been encoded, but later discover that the source of this information was not credible. The opposite of ‘credible’ is ‘noncredible’ according to the Merriam-Webster dictionary. For the two website sources, we developed unique (but plausible) websites associated with two different “authors” (confederates; one male, one female) on environmental sustainability. We chose this topic because it is highly relevant to the younger generation who will suffer from the environmental mistakes of previous generations, but also because there is disagreement over some environmental claims. For example, many people display a ‘Christmas tree’ in December but the jury is still out on whether it is better for our environment to buy a natural pine tree or an artificial one. On one hand, natural trees can safely decompose and will re-grow thus sustaining the forest. On the other hand, artificial trees reduce the need to kill a tree each year and can be re-used in subsequent years. Therefore, we could present different ‘facts’ without actually corrupting children’s knowledge of the environment. The important point here is for the reader to have an idea of why environmentalism was a good topic to use to investigate source monitoring of websites.