Evolutionary Educational Psychology
The distinction between biologically primary and biologically secondary knowledge
2 described by Geary is an instructionally important categorisation scheme (Geary
2008,
2012; Geary and Berch
2016).
Biologically primary knowledge is knowledge we, as a species, have specifically evolved to acquire over many generations. Primary skills, such as learning general problem solving strategies, imitation, recognising faces, communication through listening and speaking a native language, and social relations including our ability to communicate with each other, are modular with each skill likely to have evolved during different evolutionary epochs. We can acquire primary knowledge, easily, unconsciously, and without explicit instruction merely by membership in a group. Generally, because primary skills are acquired effortlessly, they do not need to be formally taught. Most generic-cognitive skills such as problem-solving, planning, or generalising are biologically primary (Sweller
2015; Tricot and Sweller
2014). Communicating by speaking and joint attention is a generic-cognitive skill (Callaghan et al.
2011; Tomasello & Rakoczy,
2003).
The ability to acquire vast aspects of the culture we grow up in is biologically primary. Nevertheless, in most cultures, there are many concepts and procedures that we have not specifically evolved to acquire such as reading, doing mathematics, working with a computer, or searching the internet. Those
biologically secondary skills are acquired consciously, often requiring considerable effort. Unlike primary knowledge and skills, explicit instruction is important when dealing with secondary knowledge and skills (Kirschner et al.
2006; Sweller et al.
2007). Without explicit instruction, this knowledge acquisition is likely to be severely compromised.
Unlike the generic-cognitive skills that tend to be biologically primary, biologically secondary skills tend to be
domain-specific (Sweller
2015; Tricot and Sweller
2014). Examples of biologically secondary skills include almost everything that is taught in education and training institutions. The distinction between primary, generic-cognitive knowledge and secondary, domain-specific knowledge explains why information tends to be acquired differently outside as opposed to inside educational contexts. We use primary knowledge to leverage acquiring secondary knowledge (Paas and Sweller
2012). For example, to learn geometry in a conventional class or using computer-supported material requires primary skills such as visual recognition, join attention, and schemas about space, time and sequence (Casasanto et al.
2010; Núñez and Cooperrider
2013; Siegel and White
1975), to name a few.
In this way, the theoretical machinery of evolutionary educational psychology can be used to suggest that the primary, generic-cognitive knowledge associated with collaborative learning may, under some circumstances, improve the acquisition of the biologically secondary, domain-specific knowledge that is taught.
Human Cognitive Architecture
The manner in which biologically secondary knowledge is processed by the human cognitive system is analogous to the way in which evolution by natural selection processes information. Both are examples of natural information processing systems (Sweller and Sweller
2006) which can be described using five principles summarised in Table
1.
Table 1
Natural information processing system principles
Information store | Store information in long-term memory for indefinite periods |
Borrowing and reorganising | Permit the rapid building of a long-term memory store by borrowing information from another person’s long-term memory |
Randomness as genesis | Create novel ideas |
Narrow limits of change | Use limited working memory to process novel information |
Environmental organising and linking | Use environmental signals to transfer organised information from long-term memory to working memory in order to effect appropriate action |
The
information store principle indicates that in order to function, natural information processing systems require an enormous store of information. Long-term memory provides that store for primary and secondary knowledge in the case of human cognition. The finding that skilled performance in any complex area requires the memorisation of tens of thousands of problem states and the best moves for each state (De Groot and Gobet
1996; Egan and Schwartz
1979; Jeffries et al.
1981; Sweller and Cooper
1985) provided evidence for the importance of long-term memory to general cognition. The ability to store information in long-term memory is a biologically primary skill that does not need to be taught.
The second principle, the
borrowing and reorganising principle, suggests that most of the information acquired by and stored in long-term memory is borrowed from the long-term memories of other people. We imitate others, listen to what they say and read what they write. Once information is acquired from others, it is reorganised by us using information previously stored in our long-term memory (Bartlett
1932).
For the purpose of this article, there are two aspects of this principle that need to be noted. First, borrowing and reorganising knowledge from others does not need to be taught because it is biologically primary. We are one of the few species that has evolved to obtain information from others (Brownell et al.
2006). Second, collaborative learning makes use of the borrowing and reorganising principle and is one of the justifications for hypothesising that collaboration can be effective for learning. During collaboration, we can obtain important information from others that may be difficult to obtain by other means. Of course, while most explicit instruction, both oral and written, also makes use of this principle, collaboration differs from non-collaborative instructional methods because there may be a greater emphasis on the reorganising aspect of this principle.
The randomness as genesis principle explains how information is first generated. If we are unable to obtain needed information from others, we need to use our primary skills to generate information ourselves during problem solving. In the absence of sources that allow us to borrow required information, we must randomly generate problem-solving moves and test them for their effectiveness. Again, this procedure is biologically primary and does not need to be formally taught. We have evolved to use general problem solving strategies and to generate moves randomly and test them for effectiveness.
The fact that the randomness as genesis principle is used in important activities such as research does not justify its use when information can readily be borrowed from others. Problem solving is only useful when we do not have alternative access to problem solutions. Under appropriate circumstances, collaborative learning can provide that access by increasing the range of information available to us.
The randomness as genesis principle has functional implications for the cognitive system, leading to the the fourth principle, namely the
narrow limits of change principle. In order to avoid combinatorial overload and explosions, we need a structure that limits the number of elements of information that we can consider at one time. Those limits are imposed by our working memory that is severely limited in both capacity (Miller
1956) and duration (Peterson and Peterson
1959). It needs to be noted that those limits only apply to novel information and not to familiar information retrieved from long-term memory, as will be discussed under the next principle. It also needs to be noted that collaborative learning may ameliorate some of the limitations of working memory (F. Kirschner et al.
2011) and especially that of asynchronous CSCL where written text is often used which may lead to cognitive offloading (Hmelo-Silver,
2002; Suthers
2006). By having multiple working memories working together on the same task, the effective capacity of the multiple working memories may be increased due to a
collective working memory effect that is discussed in more detail below.
The environmental organising and linking principle is the fifth principle and provides a justification of the preceding principles. Signals from the environment trigger the transfer of appropriate information from long-term memory to working memory. That information can then be used to generate action appropriate to the environment. While working memory is limited when dealing with novel information, it has no known limits when dealing with organised information from the information store of long-term memory. Based on this principle, we are transformed by our ability to marshal large amounts of information transferred from long-term memory to working memory. These large amounts of information from long-term memory can be held in working memory indefinitely giving us an ability to carry out actions that otherwise we could not consider. Accordingly, one of the primary aims of instruction is to help learners to accumulate the large stores of secondary knowledge and skills in long-term memory for later use. Collaborative learning aims to facilitate that procedure by increasing our ability to collectively process novel information.
In considering the advances of the evolutionary perspective and the application of the principles of human architecture to collaborative learning leads to a sub-principle, the
mutual cognitive interdependence principle (Tomasello and Gonzalez-Cabrera
2017; Tomasello, Melis, Tennie, Wyman, & Herrmann,
2012). This sub-principle acts as a subsidiary of the borrowing and re-organising principle by detailing how cognitive systems (i.e., inter-cognitive processes) acquire information between them. Systems develop, process, create, acquire, and share knowledge in mutual openness and collaboration with other systems. The knowledge in long-term memory that has been acquired by students consists of elaborations and structures intrinsically related to the type of relationship with others (i.e., instructors, other learners) and the means by which they carry out their cognitive transaction activities (i.e., face-to-face, mediated by computers). Individuals depend on an instructor’s explicit guidance and appropriate interactions with others as part of a group, but also on appropriate instructional environments of collaboration with other learners. This principle presupposes a relative openness between cognitive systems (Scheler
1994) and pays attention to the intrinsic transactive processes that allow cognitive exchange between them (Zambrano et al.
2017b). In addition, it takes into account the relationship between the system(s) and the environment without reducing them to the cognitive components of an individual system. Consequently, the evolution of human cognitive architecture depends on the mutual and simultaneous relationship between the components of an information-processing system, between systems, and between the systems and their environment.
Instructional Design
Cognitive load theory has used this cognitive architecture to devise cognitively effective and efficient instructional procedures. Cognitive load refers to the total working memory resources required to carry out a learning task. It assumes that human memory can be divided into two basic forms, working memory and long-term memory, that the information that is stored in long-term memory takes the form of schemas, and that the processing of new information requires mental effort resulting in cognitive load on working memory which affects learning outcomes (Sweller,
1998).
When presented with novel information, there are two additive sources of cognitive load imposed on working memory (Sweller
2010): intrinsic and extraneous load. In addition, germane cognitive load, defined as the working memory resources devoted to dealing with intrinsic cognitive load, is frequently discussed but it is closely related to intrinsic cognitive load. Kalyuga (
2011, p. 1), for example posited that:
[I]n its traditional treatment, germane load is essentially indistinguishable from intrinsic load, and therefore this concept may be redundant … the dual intrinsic/extraneous framework is sufficient and non-redundant and makes boundaries of the theory transparent. The idea of germane load might have an independent role within this framework if (as recently suggested by John Sweller) it is redefined as referring to the actual working memory resources devoted to dealing with intrinsic rather than extraneous load.
As such, germane load is not treated as an additive source of load here.
Intrinsic cognitive load deals with the inherent complexity of the information that needs to be processed. Complexity, in turn, is defined in terms of element interactivity. Consider a learning task given by a teacher such as learning the translation of a list of 50 words from one language to another (i.e., word-pairs) within a certain period of time (e.g., 60 min). Despite the difficulty of learning the words, it is not a complex task because each word-pair can be learned independently of every other word-pair. Learning that chat is the French word for cat can be learned without reference to the fact that chien is the French word for dog. The two word-pairs do not interact. For this task, element interactivity and intrinsic cognitive load are low because working memory does not have to process more than one or two word-pairs simultaneously. Of course this intrinsic load will be influenced to a certain extent by the learner’s prior knowledge, for example if the learner knows a different Indo-European or more importantly Romance language than French (e.g., Spanish, Italian, Portugese, etc.) the task could be, intrinsically, less complex and also less difficult while for a learner without knowledge of either English or French (e.g. someone who speaks only Slavic or Afroasiatic languages) the task is intrinsically more complex and more difficult. In contrast, using those same words to write a few simple sentences requires far fewer elements but the elements interact with each other. All of the words in the sentence have relations with other words (e.g., gender, gender-related articles, plurals, verb conjugation, etc.) and thus must be considered as a whole unit in working memory when learning to carry out this task. We are often unable to make any change to any of the parts of the sentence without affecting other elements and so element interactivity and intrinsic cognitive load are high.
Element interactivity is affected by both the nature of the task (as indicated) and by levels of learner expertise. In the above example, learners who are competent in a language have stored the grammatical relations between words in long-term memory. According to the
environmental organising and linking principle, that stored knowledge can be transferred to working memory as a single entity. In other words, the interacting elements are incorporated in the stored knowledge and so an entire problem such as writing The translation of the sentence
“cats and dogs are both pets” into French “
les chats et les chiens sont tous deux des animaux domestiques” constitute a single larger and more complex chunk (Egan and Schwartz
1979) or encapsulated element (Boshuizen & Schmidt,
1992). The individual elements are incorporated in that chunk and so for an expert in the language, element interactivity is low. For a novice, element interactivity for this collection of words may be very high. A change in either the nature of the task or the expertise of the learner results in a change in element interactivity which otherwise, remains constant.
Element interactivity also determines the level of
extraneous cognitive load. This form of working memory load refers to the load imposed by information elements unrelated to the learning task such as the way the information or the task is presented (Chen et al.
2016). These elements can be produced by instructional procedures and so it is under the control of instructors and can be varied by using different instructional procedures. Cognitive load theory has developed a wide range of instructional procedures designed to reduce extraneous cognitive load (Sweller
2011). Another example is the worked example effect which occurs when problem-solving skill is enhanced more by studying worked examples rather than solving the equivalent problems. The effect occurs because element interactivity is reduced by studying worked examples in comparison to problem solving. During problem solving, learners must search for appropriate moves using the
randomness as genesis principle. In the case of an algebra problem such as
(a + b)/c = d, solve for a, all of the elements of the problem statement must be considered in an interconnected way along with the consequences of the series of possible moves at each choice point. Element interactivity may result in a working memory load far above working memory limits. In contrast, studying a worked example demonstrates a use of the
borrowing and reorganising principle. When studying a worked example, each step can be considered without concurrently considering alternative moves because an appropriate move has been provided. Element interactivity and the extraneous load on working memory are reduced by the use of worked examples.
Relation to Collaborative Learning
Collaborative learning occurs when two or more students actively contribute to the attainment of a mutual learning goal and try to share the effort required to reach this goal, either face-to-face or supported by a computer (Teasley and Roschelle
1993). This activity is most often initiated by the posing of a learning task or problem by the instructor. The task may be well-defined (i.e., a task with specific goals, clearly defined solution paths, and clear expected solutions), ill-defined (i.e., a task with no clear goals, solution paths, or expected solutions) or even wicked (i.e., a task with incomplete, contradictory, and/or changing requirements that are often difficult to recognize; Rittel and Webber
1984). Many researchers of CSCL have emphasised the use of learning in groups for all three types of tasks/problems (Baghaei et al.
2007; Le et al.
2013; Scheuer et al.
2010; Strijbos, Kirschner, & Martens, 2003; Suthers
2003). The use of cognitive load theory for these different types of tasks has also been well recorded, for example Van Merriënboer and Sweller (
2010), Rourke and Sweller (
2009), and Sweller et al. (
2011a,
2011b).
Although in the short run, collaborative learning results in group members trying to successfully perform a certain learning task or solve a specific problem together, in the long run, as an instructional method, it is very important that all members of the group develop effective experience working together (i.e., domain-generalised group knowledge, (Kalyuga
2013)) that facilitates every member in acquiring domain-specific knowledge from this combined effort.
The use of collaborative learning has implications for extraneous cognitive load. Let us assume students are learning to solve a particular class of geometry problems. Depending on the extent to which the elements interact, there is an intrinsic cognitive load associated with that task irrespective of how it is taught. In addition, we need to choose whether to have the students learn this material individually or collaboratively. Both instructional procedures have levels of element interactivity associated with them that are independent of the intrinsic cognitive load. Indeed, since ‘individual’ learning and ‘collaborative learning’ are extremely broad umbrella terms, the levels of element interactivity associated with both depend on the particular version of individual or collaborative learning we use. Our aim is to reduce the element interactivity associated with extraneous cognitive load and optimise the elements associated with instrinsic cognitive load by changing the instructional technique we use. If the element interactivity associated with collaborative learning (i.e., for the individual group member) is less than the element interactivity associated with individual learning, then extraneous cognitive load is reduced by using collaboration.
There are theoretical grounds for hypothesising that the use of collaborative learning can reduce element interactivity and its concomitant cognitive load. According to the
mutual cognitive interdependence principle, appropriate collaborative learning introduces a
collective working memory (F. Kirschner et al.
2011) that otherwise does not exist. This collective working memory is part of a collective working space that is created by communicating and coordinating (relevant) knowledge held by each individual group member. Through communication and the resulting socio-cognitive processes within the group, a collective knowledge structure, or mutually shared cognition, consisting of shared mental models is formed. Research shows that these collective knowledge structures are conditional for the effectiveness of collaboration (Van den Bossche et al.
2006). The concept of a collective working memory is strongly linked to the theory of group cognition (Stahl
2014) which considers a larger unit of analysis than the individual mind as a producer of cognitive activities such as complex problem solving. However, the collective working memory concept has an important focus on the learning of individuals in the group. Under individual learning, all interacting elements must be processed in a single working memory of that individual. Under collaborative learning, various interacting elements can be distributed among multiple working memories (i.e., the working memories of the different group members) thus reducing the cognitive load on a single working memory. Those multiple working memories constitute a collective working memory that is larger than a single memory. One could state that for complex tasks/problems, collaboration becomes a scaffold (just like worked examples) for individuals’ knowledge acquisition processes. Collaboration, then, will be effective if it becomes a scaffold in this sense. If it does not, or if it in itself adds too much extraneous load, it will be harmful. The process of creating a collective working memory can be supported by helping the members of a group to exchange knowledge and information. Making learners dependent on each other, either for successfully carrying out and completing a task (i.e., task/goal interdependence) or for exchanging resources (i.e., positive resource interdependence), has been shown to be way of doing this (Johnson et al.
2001; Langfred
2000).
In summary, extraneous – and thus also total - cognitive load is changed because having learners collaborate, in effect, changes the instructional procedure (P. Kirschner et al.
2014). A collective working memory function is also seen in CSCL when learners socially share learning, their resources and regulation as is the case in co-regulated and socially shared regulation of CSCL (Järvelä et al.,
2015). During collaborative learning, some information comes from collaborators rather than other sources and that information is likely to become available exactly when it is needed resulting in a decreased load and increased learning.
Collaborative learning and evolutionary categories of knowledge
We can assume humans have evolved to work together, with the existence of language providing strong evidence. Collaboration provides a major purpose for the evolutionary development of language (Tomasello
2008; Tomasello & Rakoczy,
2003). If we have evolved to collaborate, then the act of collaboration is biologically primary. Nevertheless, while we may have evolved to collaborate, it does not necessarily follow that we collaborate effectively and efficiently while acquiring biologically secondary information under all circumstances. A failure to collaborate appropriately may be even more prevalent in CSCL where some affordances/conventions of contiguous collaboration do not apply (e.g., Kirschner
2002a; Jeong and Hmelo-Silver
2016) and where others (e.g., deixis, body language, facial expressions) are often not available (Dwyer & Suthers
2006). Acquiring biologically secondary information during collaboration requires learners to collaborate on a specific secondary task including obtaining the necessary support and guidance to collaborate appropriately. While collaborating is biologically primary, the manner in which we collaborate may differ when, for example, we collaborate to solve a mathematical problem as opposed to write prose or design an artefact, or when we collaborate face-to-face in a project room setting or do the same in a text-based CSCL setting. We may need to learn the differing collaborative techniques for each activity and each setting. It is possible that under some circumstances, collaboration facilitates the learning of biologically secondary information while under other circumstances it interferes with that learning.
Consider two conditions under which collaboration may occur. First, individuals may collaborate because the learning task is highly complex. However, the knowledge held by different people is asymmetric (i.e., each learner may possess some of the necessary information, but not other information that is possessed by other people). In this situation, the task requires collaboration considering the different levels of knowledge and expertise. The goal is learning while carrying out a complex task. However, if the prior knowledge differences have not been recognized before carrying out the task and the members have not had previous experience working together, their learning will be negatively affected (Zambrano et al.
2017b; Zhang et al.
2016). Collaborators will experience extraneous cognitive load due to task-unrelated transactive activities. Some of them may learn incidentally due to primary knowledge, but may not learn as a group.
A second circumstance in which collaboration may occur is when the learning task is highly complex but group members have worked together as a team or they are provided with external collaboration scripts (Fischer et al.
2013). As in the first situation, group members are going to carry out the task. The difference is that they have had experience of how to work together (i.e., how to organize the information, how to distribute the activities among them, how and when to exchange roles according to the type of activity, and so forth), or are explicitly guided by the learning environment as to how to effectively collaborate (e.g., via external scripts, just-in-time support). In other words, collaborators are using their own experience of how to work together or other people’s experience of how to work together so that hey are able to focus their cognitive resources on acquiring relevant knowledge in long-term memory. These collaborators will experience less cognitive load and better knowledge structures due to
task related transactive activities. A recent meta-analysis provides evidence that CSCL scripts substantially improve learning outcomes for domain-specific knowledge and collaborative skills compared to unstructured CSCL (Vogel et al.
2016).
The above examples show the importance of making well-thought-out choices when it comes to the learning goals of a collaborative task. While in education the goal of learning domain-specific knowledge is often accompanied with the goal of learning how to collaborate, it is important to realise that both require different guidance and support and that what may cause intrinsic load with respect to one goal may produce extraneous load with respect to the other and vice versa. For example, a collaboration script may provide intrinsic load with respect to the learning of a collaboration skill, but attract students’ attention away from a deep processing of the content material being discussed.
From a cognitive load theory perspective, there are conditions under which collaboration may or may not facilitate learning depending on element interactivity and interactions between the
information store principle, the
borrowing and reorganising principle and the
narrow limits of change principle. Collaborative learning is beneficial when the task exceeds individual working memory capacity (under time restrictions) assuming members have not stored relevant prior knowledge structures. Under those circumstances and where individuals have prior experience working together on similar tasks, they can appropriately distribute the elements and cognitive activities of the task at hand and take advantage of their greater capacity and inter-individual communication to acquire better knowledge structures. However, if most or all members already have relevant knowledge structures about the task in their long-term memory, then previous group experience, greater group cognitive capacity, and inter-individual communication are unnecessary. Finally, if groups are composed of advanced students (i.e., more-knowledgeable learners) and are instructed with information already learned, collaboration can even be detrimental as the group members can experience an
expertise reversal effect that occurs when instructional procedures that are beneficial for novices have negative consequences for more expert learners (Sweller et al.
2011a,
2011b). In sum, what students already know may determine whether collaboration is effective (Retnowati et al.
2017; Zambrano et al.
2017b; Zhang et al.
2016).
Incorporating the
mutual cognitive interdependence principle into human cognitive architecture used by cognitive load theory provides the basis for the
collective working memory effect (F. Kirschner et al.
2011; P. Kirschner et al.
2014). This effect suggests that learning in a team is more effective than individual learning if the complexity of the to-be-learned material is so high that it exceeds the limits of each individual learner’s working memory. In this situation, the cognitive load of processing this complex material is shared among the members of a collaborative learning team enabling more effective processing and easier comprehension of the material. In other words, when the complexity of the material which is to be learned and/or the learning task that needs to be carried out is so complex that it exceeds the working memory capacity of the individual learner, the collective working-memory effect will make group learning more effective than individual learning. F. Kirschner et al. (
2011) have experimentally confirmed this hypothesis, suggesting that
…for high-complexity tasks, group members would learn in a more efficient way - both in terms of the learning process and outcomes - than individual learners, while for low-complexity tasks, individual learning would be more efficient… This efficiency is affected by the trade-off between the possibility to divide information processing amongst the WMs of the group members (i.e. collective working memory effect) and the associated costs of information communication and action coordination. (p. 621)
Communication and coordination, depending on their content, can be divided into two categories: firstly, general communication and coordination which can be biologically primary, and secondly, school task-specific communication and coordination which is biologically secondary and is based on knowledge of general communication. Biologically primary knowledge will impose little load on working memory (e.g., reading nonverbal communication of team members or making facial expressions in quotidian situations), while biologically secondary knowledge will probably impose a greater load on working memory. Concerning the load on working memory when dealing with collaboration and the channel through which this communication takes place should be taken into account. The more the channel of collaboration mimics a face-to-face interaction, the less of a load collaboration will place on working memory because it relies on biologically primary knowledge we have on how to collaborate with each other. Whether the costs are low or high, both should be taken into account when deciding the effectiveness of collaborative learning as an instructional method. Within the collective working memory effect these costs are refered to as transactive activities, which were introduced above but will be discussed in more detail in the next section.