Balancing Collaboration and Cognitive Load to Optimize Individual and Group Desirable Difficulties

Abstract

We sought to identify factors that optimize individual learning in complex, technology-enhanced learning environments. Undergraduates viewed tutorials and played a simulation-based game either alone or in groups and in either high or low cognitive load sequences and later took tests measuring comprehension of tutorials and transfer of computer networking skills. A cognitive load by collaboration interaction was found for both immediate and delayed transfer measures, but not comprehension measures. Students working in groups performed best under high cognitive load whereas students working individually performed best under low cognitive load. These findings support the notions of optimal individual and group cognitive load and have implications for leveraging technology to design learning environments that allow students to collaborate and maximize individual learning.

Keywords

cognitive load collaboration desirable difficulty simulations games

Educational institutions face numerous challenges in 21st century skill development, specifically collaboration or teamwork, and in leveraging technology to aid students in attaining 21st century skills and knowledge. Although collaboration is widespread within educational contexts, individual-level performance often fails to reflect learning gains (Johnson & Johnson, 1999) suggesting a mismatch between group activities and the quality of learning outcomes for the individual learner. Also, little is known with regards to computer-mediated collaborative learning. Given increased technology-enhanced learning environments, are there specific factors that facilitate the effectiveness of collaboration with a digital game? The present study examined when collaborative and individual learning may be most effective as a function of the cognitive load imposed by the instructional sequence.

The current shift in valued skills results from an economy transformed into one that is globalized and characterized by decentralized decision-making, widely shared information, and project teams (Binkley et al., 2009). Progress is increasingly contingent on teams that, opposed to individuals, often collaborate, and communicate through digital technologies. A key criticism of education today is that students are unable to demonstrate collaborative skills in real-world contexts (Scardamalia et al., 2010). Technology-enhanced learning environments increasingly address this challenge especially within domains where skills training is difficult or impractical (e.g., critical thinking, media literacy; Gibson et al., 2007).

Simulation-Based Games

Digital games have been described as “the most engaging intellectual [activity] that we have invented” (Foreman et al., 2004). They can provide new opportunities to engage learners actively with authentic experiences. When merged with simulations to visually represent complex systems, games provide powerful, situated experiences (Linn et al., 2004). For example, in 2002, America’s Army became the first 3 D role-playing game that developed Army-related values to educate the general public as well as to increase recruiting. In 2003, the Defense Advanced Research Projects Agency (DARPA) began using a simulation-based gaming system to develop team-level defense tactics for situations such as when convoy units are ambushed, natural disaster relief, and search and rescue skills (Chatham, 2007). With similar goals for enhancing cognitive skills, the Cisco Networking Academy developed a simulation-based game to train students for careers in information technology (IT). Cisco Aspire develops and assesses entrepreneurial and decision-making skills within the IT industry. Educators continue to strive to capture the motivational power of games for student learning (Plass et al., 2015). Although digital games were not a main focus of the present study, we chose to situate our study within a digital game environment to test the effects of two important learning elements: collaboration and cognitive load.

Individual Versus Collaborative Learning

Is learning in groups always better than learning alone? Not necessarily. Sometimes having an individual brainstorm or recall in groups actually produces fewer unique instances than comparison groups who meet separately and pool their contributions (Peker & Tekcan, 2009; Weldon & Bellinger, 1997), often referred to as the collaborative inhibition effect. Further, when individuals learn in groups they can sometimes exhibit greater confidence in their answer choices, including when they are wrong (Puncochar & Fox, 2004).

Other times, there are clear advantages to working in groups. For example, cross-cuing, a process in which individuals may spark each other’s memory of one thing by mentioning something else, may happen when groups are attempting to recall the answer to a test question (Cranney et al., 2009; Meudell et al., 1995). Moreover, collaborative groups may outperform individuals in the long run even if their performance is inferior at the first place, a delayed benefit which appears only when group members are tested individually after a delay (Basden et al., 2000).

Collaborative learning models are based on the premise that active knowledge construction is best achieved through shared social interactions as opposed to traditional approaches in which learning takes place as strictly an individual activity (Murphy & Alexander, 2005). However, existing applications that seemingly demonstrate positive collaboration effects (i.e., higher-order skills) often implement “extra” steps as part of a collaborative script that often constrains students’ discourse and inhibits the fun and richness of the activity (Dillenbourg, 2002). Such approaches do not address under which conditions collaboration is most effective in learning environments. Given recent research, collaboration may not be effective if a mismatch exists between group-level learning and the cognitive resources required from each group member during that time (Moreno, 2009; Nihalani et al., 2011). It may be the case that some instruction is more appropriately designed for the individual learner rather than groups. How can such a mismatch be avoided?

Cognitive Load

Cognitive load theory (CLT) was developed to explain how cognitive resources might be best distributed during learning and has traditionally been concerned with individual learning from complex cognitive activities (Sweller et al., 1998; but see Janssen & Kirschner, 2020, for an exception). It posits that learners have limited cognitive and attentional resources. Types of cognitive load that are involved in learning include intrinsic and extraneous (Sweller, 2020; van Merrienboer & Sweller, 2005).

Overall cognitive load is determined by the complexity of the learning materials. The more complex, the more difficult for the learner due to higher demand for resources. Intrinsic cognitive load is determined by the difficulty of integrating information between long-term and working memory. For example, making sense of a concept or practicing skills require intrinsic load. In contrast, extraneous cognitive load is determined by irrelevant stimuli that take the learner’s attention away from schema construction.

CLT can perhaps be best understood by drawing an analogy with cholesterol. Total cognitive load is similar to a person’s overall cholesterol score. To be healthy, the score must not be too high. The same goes for intrinsic load – if it is too high, then the material is simply too difficult for the learner. Similar to the ratio of HDL (good) and LDL (bad) cholesterol, for learning to be successful, it is optimal to have a high ratio of intrinsic to extraneous cognitive load. Both intrinsic and extraneous load can be either increased or decreased by instructional design manipulations (van Merrienboer & Sweller, 2005). For example, when element interactivity (intrinsic load) is increased, students are better able to write Chinese characters (Lu et al., 2020). In contrast, students who take tests rather than study do worse on immediate tests due to increased working memory depletion caused by extraneous load (Leahy & Sweller, 2019).

Thus, CLT is particularly informative for designing technology-enhanced environments as it provides a cognitive perspective of learning, as opposed to one focused on information delivery. Recent applications of CLT to groups aim to predict when collaboration facilitates performance compared to working alone (e.g., F. Kirschner et al., 2009a). For example, when tasks are complex, groups may benefit from collective working memory that is not available for individuals (P. A. Kirschner et al., 2018). Thus, collaborative groups may experience an advantage where information is distributed across members making the group better able to manage highly complex tasks than individuals. Therefore, the more complex the task, the more effective it will be for individuals to collaborate than work alone. Conversely, if highly complex instruction or assessment tasks can be managed through groups’ expanded processing capacity, can groups also manage less complex instruction or tasks just as effectively?

Game-Based Instruction and Cognitive Processing

Integrating digital games into the classroom can be difficult and previous instructional strategies used with traditional learning environments may not generalize to computer-based tools (Mayer & Moreno, 2003; Nihalani et al., 2011). Low prior knowledge students, for instance, may be at a disadvantage, in contrast to high prior knowledge students, because of their lower cognitive capacity. Attempts to teach common strategies to such students (i.e., collaboration, feedback) may hamper learning outcomes if not matched properly. Intuitively, and based on CLT, it seems computer-delivered instruction should present low prior knowledge students with smaller chunks, or bites, of instruction while still allowing for experimentation and exploration. If educators are able to control the amount of instruction that is delivered to students, then perhaps they can also fashion the instructional sequence, composed of instruction and practice tasks, to fit their students’ needs. Unfortunately, there is little research on how to design instructional sequences that match the cognitive processing capacity of both individuals and collaborative groups.

The present study was designed to provide practical methods for educators to implement individual and collaborative learning effectively when using a simulation-based game. Designers typically do not produce multiple versions of the same game to be used across different instructional strategies. However, based on CLT, restructuring the instructional sequence to match students’ and groups’ cognitive capacity may improve individual learning. Further, giving educators the option to restructure instructional sequences themselves may increase the use and effectiveness of simulation-based games in the classroom.

The present study examined the effects of collaboration (individual vs. collaborative) and cognitive load (small bites vs. large bites) on individual comprehension and transfer in a simulation-based game. We were particularly interested in the potential of a collaborative group’s expanded cognitive capacity within the framework of the cognitive-affective theory of multimedia learning (CATML) (Moreno, 2005; Moreno & Mayer, 2007) and CLT (Sweller et al., 1998).

Advocates for collaborative learning have argued that group discourse stimulates conditions necessary for learning such as allowing learners to monitor their conceptual understanding and clarifying inconsistencies in knowledge (Brown & Palincsar, 1989; Murphy & Alexander, 2005). From a CLT perspective, if learners are expending cognitive resources that are beneficial to learning, then they are more likely to engage in generative cognitive processing (Moreno & Mayer, 2007; Moreno, 2009). With regard to collaborative learning from instructional sequences that impose high cognitive load, the opportunity to divide the processing of information among group members was predicted to result in positive learning outcomes (F. Kirschner et al., 2009b; Moreno, 2009; Nihalani et al., 2011).

In contrast, if learners are ineffectively expending cognitive resources elicited by mismatched instruction, they may experience an increase in extraneous load. Therefore, it was predicted that there would be an interaction on measures of cognitive processing between collaboration (individual vs. collaborative) and the cognitive load required by the instructional sequence (low vs. high). When presented with the high cognitive load instructional sequence, students working in collaborative groups were expected to outperform those working individually by pooling individual processing capacities to share the cognitive load imposed by the instructional sequence. Individual learners, in contrast, were expected to experience extraneous processing in the high load sequence because they had only the option to rely on themselves.

Method

Sample size was estimated using G*Power (Faul et al., 2007). Using a two by two factorial design, an alpha of 0.05, and expecting a medium effect size (d = .50), a minimum sample size of 136 (34 per cell) would be needed to obtain statistical power of 0.80.

Participants and Design

One hundred forty-nine undergraduates (85 males, 15 freshmen, 23 sophomores, 28 juniors, and 83 seniors) were recruited from a subject pool at a large state university in the Southwest and received course credit. There were 35 liberal arts majors, 27 natural sciences, 25 business, 21 communications, 11 education, four engineering, two geosciences, eight other, and 16 undeclared. They were randomly assigned to one of four conditions in a completely crossed, 2 collaboration (individual vs. collaborative) by 2 cognitive load (low vs. high) factorial design. Collaborative groups were formed by random assignment.

Instructional Sequences and Collaborations

In the low cognitive load instructional sequence, both tutorials were presented with tests directly following each tutorial. This technique – breaking the multimedia presentation into smaller bites – is an application of the segmenting principle. In the high cognitive load instructional sequence, both tutorials were presented in a single continuous sequence followed by the tests over the first and second tutorial. Collaborative learning conditions required group work on the Aspire contract tasks. Students in the individual learning conditions simply completed each of the tasks and measures on their own without the help of others.

Assessments

All assessments were designed to be consistent with those used in previous studies examining CLT (Sweller et al., 1998; Sweller & Chandler, 1994) and CATML (Moreno, 2005, 2009) frameworks.

The immediate cognitive load manipulation test over the first tutorial was comprised of ten multiple choice items. An example item is below:

What does WAN stand for?

Wireless Area Network

Wide Area Network

Wireless Access Network

Wide Access Network

Both transfer tests had multiple choice activities/questions that were part of the Quest contracts. The immediate transfer test included six multiple choice items measuring transfer from the two tutorials. The delayed transfer test included five multiple choice items that each presented a hypothetical problem that was an extension of the first and second contracts. The student had to determine the “first step” in troubleshooting the problem depicted. An example item/activity involved the student having to purchase the correct equipment (e.g., three personal computers, certain cables, and a specific router). If students purchased only two computers, for example, they could still complete the activity, but the final set up would not work. The cognitive load scale included eight items that measured students’ mental effort and perceived difficulty after completing the second and third contracts. The items were adapted from Moreno (2006, 2009) and Moreno and Mayer (2002). Students were asked to rate the mental effort they had expended while completing each contract on a five-point Likert scale ranging from (1) Extremely low mental effort to (5) Extremely high mental effort. See example below:

1	2	3	4	5
Extremely low mental effort				Extremely high mental effort

Please rate your level of mental effort during Aspire tutorial (one; two).

The comprehensive retention test contained 15 multiple choice items that covered information in both tutorials. An example is below:

Which of the following cables would be used to connect a PC to a server?

Copper non-terminating Ethernet cable

Copper straight-through Ethernet cable

Copper mixed-use Ethernet cable

None of the above

Instructional Materials

Tutorials

Cisco Aspire contains three sets of instructional content that were built into the game. The first set of instructional content is a series of tutorials that demonstrate the basic functions, features, and aspects of Aspire through animated Captivate screencasts. Each tutorial delivers narration through pop-up text-boxes that accompany the animation as well as voice narration. Users can turn off the audio narration if they choose. The second set of instructional content is a series of tutorials for Packet Tracer, the simulation under Aspire’s virtual world, called My First PT labs. The tutorials were, on average, four-minute animated Captivate screencasts that orient the user to Packet Tracer’s interface by displaying how to drag network icons and symbols to the virtual workspace. Unlike the Aspire tutorials, the narration in each Packet Tracer tutorial is only delivered through pop-up text-boxes that accompany the animation; no voice narration accompanies the text-boxes. The third set of instructional content is a Help menu that can be used as a reference guide and contains text files with annotated screenshots covering a broad range of topics.

For purposes of this study, content was pulled from the three existing sets to design two tutorials that cover navigating the virtual world and fundamental computer networking and business knowledge needed to complete quest contracts. Because the quest contracts used in this study were pre-existing, task analyses for each contract were reverse-engineered to identify knowledge needed to complete the quests and evidence for the skills being practiced. The tutorials were designed according to principles grounded in CTML (Mayer & Moreno, 2003) and CATML (Moreno, 2005, 2006, 2009). Instead of including popup boxes, narration was delivered aurally. To maintain experimental control of pace, the tutorials were set to be system-paced so that students would not be able to stop, pause, rewind, and fast-forward the presentation.

The first Aspire tutorial was a nine-minute screencast created in Camtasia Studio and used a worked-out example to familiarize first-time users with how to navigate within Aspire’s virtual world and how to play the game. By virtue of the worked-example, the user is shown how to: (1) accept a contract from a client, (2) navigate the city to enter specific places of interest (see Figure 1) (3) search for and purchase equipment from the Networking Equipment Store, (4) view their profile and game progress, (5) access and use tools (e.g., task lists in the Contract Binder, see Figure 2), and (6) work through the contract. Additionally, a three-minute introduction covering the rudiments of Cisco’s Networking Academy and Aspire’s intended user audience for developing entrepreneurial and technical skills was included.

Figure 1.

Screenshot of Aspire Tutorial One.

Figure 2.

Screenshot of Contract Binder Tool.

The second Aspire tutorial was a seven-minute screencast that demonstrated the: (1) fundamental networking connection, a Local Area Network (LAN), as a simple connection between a PC and a router, (2) defining characteristics of a LAN, (3) three steps involved in creating a network (physically building it, setting it up, testing it), and (4) configuring a wireless connection. Next, a worked-example was presented that showed how to: (5) search for and purchase specific equipment (modules, accessories, and software), (6) access and use tools (e.g., available equipment left over from previous contracts), (7) work through building a network, and (8) configure network connections. Business knowledge was also included regarding: (9) opportunities to donate materials and services, (10) maintaining relationships with previous clients, and (11) choosing vendor relationships that balance budget and quality of services. In contrast to the first tutorial, the second tutorial only presented instructional content for transfer of knowledge to complete Contract 2.

Quest Contracts

Students’ progress in Aspire was tracked through score calculations, proficiency points, and time. Score calculations were based on various factors and awarded at different points during gameplay, but the majority were calculated when users completed a contract. Aspire’s Help menu lists the following examples as factors that affect scoring: completing contracts (with various aspects scored independently), credits given, achieving proficiencies and badges, and time.

Aspire also measured students’ Business and Computer Networking proficiencies determined by algorithms on student’s gameplay. Each proficiency further measures three interrelated skill sets. For Business, these are money management, business sense, and reputation. For Computer Networking, these are configuration, troubleshooting, and physical labor.

Contract One (Internet Café)

Upon entering the city, the client (Maria) called the student’s avatar requesting assistance with setting up computers in a café. After the call was answered and contract accepted, as presented in the first tutorial, Maria instructed the player to visit the contract site, the Internet Café, shown below in Figure 3. When the student’s avatar entered the contract site, a detailed task list that was stored in the Contract Binder appeared. This contract required the user to purchase and set up four PCs within an allotted budget.

Figure 3.

Screenshot of Contract One Venue (Internet Café).

The first contract required that students purchase and setup four PCs. The allotted budget for this contract was not enough for four new PCs unless a loan was taken out from the bank. Thus, they had to purchase four refurbished PCs as well as maintenance contracts in case computer problems arose.

Contract Two (State Office Building)

During the first contract, the client (Maria) gave the avatar the contact information for a friend of hers (Mrs. Judy Jones). Once players followed through by making contact, they were recommended for a job at a state office building. The client, Michael, introduced the contract with the following message:

Hi [Aspire Player]. The Employment Agency is expanding the computer training lab and adding PCs to use for Internet calling and job interviews. They need someone to help purchase and install the new PCs and accessories.

After the contract was accepted, the player was directed inside the contract site, the State Office Building, shown below in Figure 4.

Figure 4.

Screenshot of Contract Two Venue (State Office Building).

The second contract required students to: (1) correctly purchase equipment (PCs, headphone, microphone, camera, wireless card, cables), (2) install them in a lab, (3) physically connect three PCs to a router, (4) install a wireless card in a fourth PC, and (5) enable the computers to get their IP address configuration from the router (set to DHCP). There were numerous ways in which this contract could be completed and all students were successful within the allotted virtual time.

Procedure

All experimental sessions, including the pilot, were conducted in a computer lab with 25 Mac computers. During the tutorials, students used USB headphones.

Day One

Individual + Low cognitive load students (1) viewed the first tutorial individually, (2) completed the immediate retention test over the first tutorial, (3) completed the first contract, (4) viewed the second tutorial, (5) completed the second contract, (6) immediate transfer test, and (7) cognitive load scale.

Individual + High cognitive load students (1) viewed the first tutorial individually, (2) viewed the second tutorial, (3) completed the immediate retention test over the first tutorial, completed the (4) first and (5) second quest contracts, (6) immediate transfer test, and (7) cognitive load scale.

All collaborative students were asked to rearrange seating so that members of their group could sit together. Collaborative + Low cognitive load students (1) viewed the first tutorial, and (2) completed the immediate retention test on their own. Next, as a group, students (3) completed the first contract. Although members worked together, they each needed to complete the contract on their respective computers individually. Next, the groups (4) viewed the second tutorial and (5) completed the second contract, followed by individually completing (6) the immediate transfer test, and (7) cognitive load scale.

Collaborative + High cognitive load students (1) viewed the first tutorial followed by the (2) second tutorial in groups. Next, independently, students (3) completed the immediate retention test. Groups then worked together on the (4) first and (5) second contracts. Similar to the other collaboration, members needed to complete the measures on their respective computers. Students then worked independently on the (6) immediate transfer test and (7) cognitive load scale.

At the end of session one, all students received instructions not to discuss tutorials, contracts, or measures.

Day Two

The experimental procedure for day two, 48 hours later, was the same for all students. Students completed the following measures individually: the delayed transfer test and the delayed comprehensive test.

Results

The dependent variables under investigation were the immediate retention and comprehension tests, and the delayed transfer and comprehensive tests. All statistical analyses were conducted using SPSS version 24 and an alpha level of 0.05. Percent correct scores and split-half reliability estimates with Spearman-Brown adjustment were computed for all measures. Estimates of effect size are reported as Cohen’s (1988) f and interpretations correspond to .4 = large, around .25 = medium, and .1 = small. Table 1 displays means and standard deviations of scores on all measures and reliability scores. A Levene’s test was conducted to test for homogeneity of error variance for each measure and found no significant effects; thus, supporting the assumption of equal variances among conditions.

Table 1.

Means (and Standard Deviations) of Percent Correct Scores and Reliability Estimates.

Dependent	Individual		Collaborative
Variables	Low load	High load	Low load	High load
	n = 36	n = 35	n = 39	n = 39	r
		Day 1
Retention	66 (18)	51 (16)	63 (16)	44 (19)	.75
Comprehension	74 (15)	46 (22)	57 (23)	60 (23)	.61
		Day 2
Transfer	62 (15)	54 (15)	60 (17)	62 (19)	.69
Comprehensive	76 (10)	76 (13)	74 (14)	71 (15)	.60

Immediate Retention Test

Retention was assessed for the low and high cognitive load conditions after viewing the first and second tutorials respectively. To verify experimental manipulation of cognitive load, an independent-samples t-test was conducted on retention test scores. Scores in the low cognitive load condition (M = .64; SD = .17) were greater than scores in the high cognitive load condition (M = .47; SD = .18), t(147) = 6.05, p < .01, f = .50, confirming the attempt to manipulate cognitive load through the instructional sequence was successful.

Three Learning Measures

Initially, a MANOVA with collaboration and cognitive load as between-subjects factors was conducted on the three learning measures: immediate comprehension, delayed transfer, and delayed comprehensive test. The analysis revealed an interaction, Wilk’s λ = .86, F(3, 145) = 20.25, p < .01, f = .37. To follow up the multivariate interaction effect, separate univariate two-way ANOVAs were conducted for each of the three learning measures.

Immediate Transfer Test

The day one transfer test over the tutorials was analyzed using a two-way ANOVA. There was no main effect of collaboration, F(1, 145) = .17, p > .05, MSE = .04, f = .03. However, there was a main effect for cognitive load, F(1, 145) = 14.61, p < .01, MSE = .04, f = .33. There was also an interaction, indicating the effect of cognitive load on immediate transfer scores was dependent on collaboration, F(1, 145) = 20.25, p < .01, MSE = .04, f = .40.

Tests of the simple effect of collaboration (individual vs. collaborative) within each cognitive load condition (low vs. high) were conducted to follow up the interaction effect. For students who received the low cognitive load instructional sequence, those in the individual condition (M = .74, SD = .15) outperformed those in the collaborative condition (M = .57, SD = .22), F(1, 74) = 14.41, p < .01, MSE = .04. Conversely, for students who received the high cognitive load sequence, those in the collaborative condition (M = .60, SD = .23) outperformed those in the individual learning condition (M = .46, SD = .22), F(1, 74) = 9.58, p < .01, MSE = .04. Figure 5 illustrates the disordinal interaction between collaboration and cognitive load.

Figure 5.

Collaboration by Cognitive Load Interaction on the Immediate Comprehension Test.

Delayed Transfer Test

There were no main effects for collaboration, F(1, 145) = 1.09, p > .05, MSE = .03, f = .08, or cognitive load, F(1, 145) = 1.26, p > .05, MSE = .03, f = .10. However, similar to the immediate transfer test, there was an interaction, F(1, 145) = 4.16, p < .05, MSE = .03, f = .17.

Tests of the simple effect of collaboration within each cognitive load condition were first conducted to follow up the interaction but yielded no significant effects. Thus, instead we tested the simple effect of cognitive load condition (low vs. high) within each collaboration (individual vs. collaborative). There was a simple effect of cognitive load within the individual learning condition, F(1, 69) = 5.76, p < .01, MSE = .03, indicating that students who received the low cognitive load sequence (M = .62, SD = .15) outperformed those who received the high load sequence (M = .54, SD = .15). In contrast, there was no simple effect of cognitive load within the collaborative learning condition, F(1, 76) = .38, p > .05, MSE = .03; thus, there was no difference between the low and high cognitive load conditions (M = .60, SD = .17 vs. M = .62, SD = .19, respectively). This interaction is displayed in Figure 6.

Figure 6.

Collaboration by Cognitive Load Interaction on the Delayed Transfer Test.

Delayed Comprehensive Test

There were no main effects of collaboration, F(1, 145) = 2.05, p > .05, MSE = .02, f = .12, nor cognitive load, F(1, 145) = .37, p > .05, MSE = .02, f = .05, or interaction, F(1, 145) = .59, p > .05, MSE = .02, f = .06.

Cognitive Load Scale

To help explain the findings involving the dependent measures, we also looked for differences among groups on the cognitive load scale (split-half reliability = .78). A collaboration by cognitive load interaction was found, F(1, 145) = 4.37, p < .05, MSE = .92, f = .17. There was no simple effect of collaboration within the low cognitive load condition, F(1, 74) = .27, p > .05; thus, cognitive load ratings for the second contract did not differ between the individual and collaborative learning conditions (M = 2.33, SD = .93 vs. M = 2.44, SD = .79, respectively). However, there was a simple effect of collaboration within the high cognitive load condition, F(1, 74) = 5.75, p < .05. Students assigned to the individual learning condition (M = 2.57, SD = .92) reported greater levels of cognitive load while completing the second contract than students in the collaborative learning conditions (M = 2.05, SD = .94). Figure 7 displays this interaction.

Figure 7.

Mean Ratings of Perceived Cognitive Load on Contract Two.

Discussion

This study explored how two elements of an interactive, digital game environment may affect different learning processes. First, the theory of situated cognition states that optimal learning occurs within authentic contexts or communities of practice (Brown & Palincsar, 1989; Lave & Wenger, 1991). In the present study, communities of practice were represented by the student groups who worked together to collaborate (Johnson, 2003). Thus, “situatedness” or necessary knowledge was the capability to engage in social interaction reflective of effective collaboration. Second, models of collaboration have typically not considered levels of group processing that best facilitate individual learning. Only relatively recently have researchers begun to identify optical levels of group and individual difficulty (e.g., F. Kirschner et al., 2009a). Taken together, theories of group learning have often overlooked the dynamic relationship between individual- and group-level cognitive processing and performance. For example, collaborative models that consider the group, as opposed to the individual learner, as the primary unit of analysis fail to specify the conditions under which collaborative efforts are more effective than individual efforts The assumption under which they operate is that collaboration is simply advantageous.

The findings of this study add to the growing literature demonstrating that both individual-level and group-level performance varies across different cognitive load sequences by examining this relationship within a digital game environment. Research examining collaborative processes has also not examined cognitive processing that occurs during learning phases relative to processing that occurs during tasks (or test phases). Results from the present study are consistent with the notion that groups that receive high cognitive load instruction make use of collective processing capacities through sharing the cognitive load imposed by the instructional content. Individual learners, in contrast, have only the option of relying on themselves, and thus experience extraneous processing in a high load sequence, allowing them only to retain surface level information from the tutorials. Our results also confirm that when groups receive low cognitive load instruction, individual learning suffers compared to receiving individual instruction.

Cognitive Load Reduction Versus Desirable Difficulty

Two somewhat similar but competing theories might help to explain our findings. The first is CLT which concludes that sometimes it is helpful to reduce unnecessary, extraneous load for students to optimize learning (Sweller, 2020). In contrast, the notion of desirable difficulties concludes that sometimes it is helpful to introduce difficulty into learning to improve long-term transfer (e.g., McDaniel & Einstein, 2005). These two approaches can perhaps be reconciled if one considers the amount and type of effort expended during learning to hold a “sweet spot” that is neither too much nor too little to maximize performance. In other words, difficulty level should be “just enough” to engage learners as well as not too much cognitive load so as not to frustrate learners. Consider again the cholesterol analogy. For difficulty level to be desirable, the ratio between difficulty that engages (intrinsic load) vs. that which does not (extraneous load) should be high.

But what about notions of optimal levels of group or collective cognitive load and desirable difficulty? Is there a similar sweet spot for both? Kirschner and his colleagues (P. A. Kirschner, 2001; Kirschner et al., 2009a, 2018; P. A. Kirschner & Erkens, 2013) have investigated optimal collaborative learning using a cognitive load framework. They have concluded that individual learning becomes less effective than group learning when task complexity increases. Conversely, when cognitive load is low, individual learning can be better than collaborative learning. These conclusions match up nicely with the present findings.

The desirable difficulty literature is comparatively silent on the issue of collaborative learning. But it is easy to imagine a similar relationship between individual and collaborative learning processes. Increasing difficulty eventually becomes undesirable for individuals and more desirable for group learning. Likewise, some learning tasks can be handled well by individuals and do not reach a level of difficulty to be desirable for groups. Thus, group, in addition to individual, desirable difficulty should be considered when designing learning tasks that may involve collaboration.

Limitations

As with any study, these conclusions are tempered by limitations. First and foremost is that the learning conditions in this study, particularly the collaborative groups, are a limited reflection, an artificial implementation, of learning environments as a result of the controlled laboratory nature of the study. By nature, collaboration involves a social psychological dimension related to the socio-emotional aspects of a group (Kreijns et al., 2003), including the development of a sense of trust or community with group members. Those processes typically require group members to interact with one another on several occasions before achieving true collaboration. It seems reasonable to predict, then, that the complex patterns of interactions between cognitive, motivational, and social factors found within a real world classroom environment are not exactly duplicated by the conditions tested in this study. In that sense, it is not clear to what extent the present results can be generalized to a real learning environment, classroom or workplace, that involve collaboration.

Implications

The results presented in this study hold implications for the design of collaborative learning environments, especially those that involve digital games. Bereiter (1997) suggested that “the main weakness of situated cognition is, it seems, precisely its situatedness” (p. 286). Students certainly need opportunities for engaging in collaboration if they are to develop skills needed for effective collaboration. More importantly, if collaboration is to be beneficial for individual learning, design of learning environments should also consider the cognitive load demands of specific tasks. Some tasks are conducive to collaboration whereas others are not. Even within a motivating and engaging digital game environment, such considerations are important to optimize learning.

Many digital games inherently support collaborative environments through their complex artifacts that embody multiple representations giving them a clear advantage for developing collaborative skills. To this end, the use of Aspire in this study demonstrated the potential for engaging students in collaborative and individual learning activities. But an important consideration when allowing students to collaborate is the complexity of the learning activity. As revealed in the present study, collaboration on tasks that have low cognitive load demands may result in poorer learning than if students worked individually. Likewise, when task complexity increases, preventing students from collaborating may also be detrimental. This finding is not new (e.g., F. Kirschner et al., 2008). In fact, a research agenda for Collaborative Cognitive Load Theory has been recently presented (Janssen & Kirschner, 2020). But the present study is the first to examine this relationship between collaboration and cognitive load in a game environment and suggests potential considerations when designing more effective collaborative learning environments leveraged on a digital game. Instructional sequences should distinguish learning tasks that benefit from collaboration from those that do not.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Daniel H. Robinson

Author Biographies

Priya K. Nihalani is a product leader in the privately-owned education and training industry. She has led the design and development of learning technologies for the United States Air Force, Macmillan Learning, Scantron Corporation, and numerous higher education institutions. Her research interests lie in the everyday, real-world applications of educational technology in both formal and informal learning environments. Priya received her doctorate in Educational Psychology from The University of Texas at Austin and currently resides in Minnesota’s Twin Cities.

Daniel H. Robinson is Associate Dean of Research and the K-16 Mind, Brain, and Education Endowed Chair in the College of Education at the University of Texas at Arlington. Dan served as Editor of Educational Psychology Review from 2006-2015 and as Associate Editor of the Journal of Educational Psychology from 2014-2020.

References

Basden

B. H.

Basden

D. R.

Henry

(2000). Costs and benefits of collaborative remembering. Applied Cognitive Psychology, 14(6), 497–507. https://doi.org/10.1002/1099-0720(200011/12)14:6<497::AID-ACP665>3.0.CO;2-4

Bereiter

(1997). Situated cognition and how to overcome it. In Kirshner

D. I.

Whitson

J. A.

(Eds.), Situated cognition: Social, semiotic, and psychological perspectives (pp. 281–300). Lawrence Erlbaum Associates.

Binkley

Erstad

Herman

Raizen

Ripley

Rumble

(2009). Developing 21st century skills and assessments. White Paper from the Assessment and Learning of 21st Century Skills Project.

Brown

A. L.

Palincsar

A. S.

(1989). Guided, cooperative learning and individual knowledge acquisition. In Resnick

L. B.

(Ed.), Knowing, learning, and instruction: Essays in honor of Robert Glaser (pp. 393–451). Lawrence Erlbaum Associates, Inc.

Chatham

R. E.

(2007). Games for training. Communications of the ACM, 50(7), 36–43. https://doi.org/10.1145/1272516.1272537

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates.

Cranney

Ahn

McKinnon

Morris

Watts

(2009). The testing effect, collaborative learning, and retrieval-induced facilitation in a classroom setting. European Journal of Cognitive Psychology, 21(6), 919–940. https://doi.org/10.1080/09541440802413505

Dillenbourg

(2002). Over-scripting CSCL: The risks of blending collaborative learning with instructional design. In Kirschner

P. A.

(Ed.), Three worlds of CSCL: Can we support CSCL? (pp. 61–91). Open University of the Netherlands.

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191. https://doi.org/10.3758/BF03193146

10.

Foreman

Arthur, ‐Kelly

Pascoe

Smyth King

(2004). Evaluating the educational experiences of students with profound and multiple disabilities in inclusive and segregated classroom settings: An Australian perspective. Research and Practice for Persons With Severe Disabilities, 29(3), 183–193. https://doi.org/10.2511/rpsd.29.3.183

11.

Gardner

H. E.

Csikszentmihalyi

Damon

(2002). Good work: When excellence and ethics meet. Basic Books.

12.

Gibson

Aldrich

Prensky

(2007). Games and simulations in online learning: Research and development frameworks. Information Science Publishing.

13.

Janssen

Kirschner

P. A.

(2020). Applying collaborative cognitive load theory to computer-supported collaborative learning: Towards a research agenda. Educational Technology Research and Development, 68(2), 783–805. https://doi.org/10.1007/s11423-019-09729-5

14.

Johnson

D. W.

(2003). Social interdependence: The interrelationships among theory, research, and practice. The American Psychologist, 58(11), 934–945. https://doi.org/10.1037/0003-066X.58.11.934

15.

Johnson

D. W.

Johnson

R. T.

(1999). Human relations: Valuing diversity. Interaction Book Co.

16.

Kirschner

Paas

Kirschner

P. A.

(2008). Individual versus group learning as a function of task complexity: An exploration into the measurement of group cognitive load. In Zumbach

Schwartz

Seufert

Kester

(Eds.), Beyond knowledge: The legacy of competence (pp. 21–29). Springer.

17.

Kirschner

Paas

Kirschner

P. A.

(2009a). A cognitive load approach to collaborative learning: United brains for complex tasks. Educational Psychology Review, 21(1), 31–42. https://doi.org/10.1007/s10648-008-9095-2

18.

Kirschner

Paas

Kirschner

P. A.

(2009b). Individual and group-based learning from complex cognitive tasks: Effects on retention and transfer efficiency. Computers in Human Behavior, 25(2), 306–314. https://doi.org/10.1016/j.chb.2008.12.008

19.

Kirschner

P. A.

(2001). Using integrated electronic environments for collaborative teaching/learning. Learning and Instruction, 10(1), 1–9. https://doi.org/10.1016/S0959-4752(00)00021-9

20.

Kirschner

P. A.

Erkens

(2013). Toward a framework for CSCL research. Educational Psychologist, 48(1), 1–8. https://doi.org/10.1080/00461520.2012.750227

21.

Kirschner

P. A.

Sweller

Kirschner

Zambrano

(2018). From cognitive load theory to collaborative cognitive load theory. International Journal of Computer-Supported Collaborative Learning, 13(2), 213–233. https://doi.org/10.1007/s11412-018-9277-y

22.

Kreijns

Kirschner

P. A.

Jochems

(2003). Identifying the pitfalls for social interaction in computer-supported collaborative learning environments: A review of the research. Computers in Human Behavior, 19(3), 335–353. https://doi.org/10.1016/S0747-5632(02)00057-2

23.

Lave

Wenger

(1991). Situated learning: Legitimate peripheral participation. Cambridge University Press.

24.

Leahy

Sweller

(2019). Cognitive load theory, resource depletion, and the delayed testing effect. Educational Psychology Review, 31(2), 457–478. https://doi.org/10.1007/s10648-019-09476-2

25.

Linn

M. C.

Davis

E. A.

Bell

(Eds.). (2004). Internet environments for science education. Lawrence Erlbaum Associates.

26.

Kalyuga

Sweller

(2020). Altering element interactivity and variability in example-practice sequences to enhance learning to write Chinese characters. Applied Cognitive Psychology, 34(4), 837–843. https://doi.org/10.1002/acp.3668

27.

Mayer

R. E.

Moreno

(2003). Nine ways to reduce cognitive load in multimedia learning. Educational Psychologist, 38(1), 43–52. https://doi.org/10.1207/S15326985EP3801_6

28.

McDaniel

M. A.

Einstein

G. O.

(2005). Material appropriate difficulty: A framework for determining when difficulty is desirable for improving learning. In Healy

A. F.

(Ed.), Experimental cognitive psychology and its applications (pp. 73–85). American Psychological Association.

29.

Meudell

P. R.

Hitch

G. J.

Boyle

M. M.

(1995). Collaboration in recall: Do pairs of people cross-cue each other to produce new memories? The Quarterly Journal of Experimental Psychology. A, Human Experimental Psychology, 48(1), 141–152. https://doi.org/10.1080/14640749508401381

30.

Moreno

(2005). Instructional technology: Promise and pitfalls. In PytlikZillig

Bodvarsson

Bruning

(Eds.), Technology-based education: Bringing researchers and practitioners together (pp. 1–19). Information Age Publishing.

31.

Moreno

(2006). When worked examples don’t work: Is cognitive load theory at an impasse? Learning and Instruction, 16(2), 170–181. https://doi.org/10.1016/j.learninstruc.2006.02.006

32.

Moreno

(2009). Constructing knowledge with an agent-based instructional program: A comparison of cooperative and individual meaning-making. Learning and Instruction, 19(5), 433–444. https://doi.org/10.1016/j.learninstruc.2009.02.018

33.

Moreno

Mayer

R. E.

(2002). Learning science in virtual reality multimedia environments: Role of methods and media. Journal of Educational Psychology, 94(3), 598–610. https://doi.org/10.1037/0022-0663.94.3.598

34.

Moreno

Mayer

R. E.

(2007). Interactive multimodal learning environments. Educational Psychology Review, 19(3), 309–332. https://doi.org/10.1007/s10648-007-9047-2

35.

Murphy

P. K.

Alexander

P. A.

(2005). Understanding how students learn: A guide for instructional leaders. Corwin Press.

36.

Nihalani

P. K.

Mayrath

Robinson

D. H.

(2011). When feedback harms and collaboration helps in computer simulation environments: An expertise reversal effect. Journal of Educational Psychology, 103(4), 776–785. https://doi.org/10.1037/a0025276

37.

Peker

Tekcan

A. İ.

(2009). The role of familiarity among group members in collaborative inhibition and social contagion. Social Psychology, 40(3), 111–118. https://doi.org/10.1027/1864-9335.40.3.111

38.

Plass

J. L.

Homer

B. D.

Kinzer

C. K.

(2015). Foundations of game-based learning. Educational Psychologist, 50(4), 258–283. https://doi.org/10.1080/00461520.2015.1122533

39.

Puncochar

J. M.

Fox

P. W.

(2004). Confidence in individual and group decision making: When two “two heads” are worse than one. Journal of Educational Psychology, 96(3), 582–591. https://doi.org/10.1037/0022-0663.96.3.582

40.

Scardamalia

Bransford

Kozma

Quellmalz

(2010). New assessments and environments for knowledge building. White Paper from the Assessment and Learning of 21st Century Skills Project.

41.

Sweller

(2020). Cognitive load and educational technology. Educational Technology Research and Development, 68(1), 1–16. https://doi.org/10.1007/s11423-019-09701-3

42.

Sweller

Chandler

(1994). Why some material is difficult to learn. Cognition and Instruction, 12(3), 185–233. https://doi.org/10.1207/s1532690xci1203_1

43.

Sweller

van Merriënboer

Paas

(1998). Cognitive architecture and instructional design. Educational Psychology Review, 10(3), 251–296. https://doi.org/10.1023/A

44.

van Merrienboer

Sweller

(2005). Cognitive load theory and complex learning: Recent developments and future directions. Educational Psychology Review, 17(2), 147–177. https://doi.org/10.1007/s10648-005-3951-0

45.

Weldon

M. S.

Bellinger

K. D.

(1997). Collective memory: Collaborative and individual processes in remembering. Journal of Experimental Psychology. Learning, Memory, and Cognition, 23(5), 1160–1175. https://doi.org/10.1037/0278-7393.23.5.1160