Cognitive load theory provides limited support for the use of collaborative learning. If a task is sufficiently complex, it will generate a high intrinsic cognitive load that may overwhelm an individual. However, if this cognitive load can be shared between members of a group then the intrinsic cognitive load experienced by each member will be reduced. The process of sharing will introduce its own cognitive costs associated with communication between group members and so, in theory, the conditions that favour group work should be those where these ‘transaction costs‘ are smaller than the intrinsic cognitive load that they replace.
Sharing cognitive load in this way has become known in cognitive load theory as the ‘collective working memory effect’. It is relatively new and there is not yet a huge amount of experimental data that tests the idea. A couple of key studies (here and here) were conducted by Kirschner, Paas and Kirschner in the biological domain of gene inheritance.
In the first of these studies, high school students either worked individually or in threes. The students who worked individually were each given all of the information. The students who worked in threes were each given only an third of the information required to solve the problem. For instance, one student may be told that the mother’s eye colour is blue, a different student may be told that the father’s eye colour is brown and the third student may be told that the gene for brown eyes is dominant over blue. Importantly, the students were not able to make a note of this information. Instead, they had to combine it to come up with a solution. Finally, the students were tested individually on both recall and transferring their understanding to a range of different situations.
In this study, students who worked in a group expended less mental effort to achieve the same outcomes on the transfer test when compared with those who worked individually – this comparison of effort to outcome was described as learning ‘efficiency’. However, the pattern reversed for the recall test and students who worked individually were more efficient. In the second study, working individually was more efficient for low complexity tasks, but working as a group was more efficient for high complexity tasks.
Retnowati, Ayres and Sweller tested the effect of collaborative learning for solving mathematics problems. They compared working individually with working as a group when both conditions involved learning from worked examples. In this case, working individually was superior. However, when they replaced the worked examples with problem solving, working as a group now led to better performance. Nevertheless, studying worked examples was still superior.
These experiments support the predictions of cognitive load theory, but it is important to bear in mind that there is a much wider literature on collaborative learning from outside of the cognitive load theory field. When compared to individualised learning, collaborative learning has repeatedly been shown to provide an advantage. However, considering the experiments above, it is worth returning to these studies and asking whether the concepts being learnt were simple or complex and whether the comparison condition made use of worked examples.
And everyday uses of collaborative learning are probably very different than the controlled versions used in studies. For instance, in my experience of real-world group work, students usually decide for themselves how to divide-up the task. The phenomenon of social loafing, where some group members slack-off and allow others to taken the burden, means that cognitive load will not necessarily be shared equally across group members and this may further erode any advantages of group work.
When faced with a complex learning tasks, the best approach may be to try to break it down into smaller components and develop worked examples. However, in situations where this may not be possible, such as in a complex professional learning environment, then the use of collaborative learning may be beneficial.
Filling the pail 
You missed this one from John, Femke, Jimmmy and me: From Cognitive Load Theory to Collaborative Cognitive Load Theory https://link.springer.com/article/10.1007/s11412-018-9277-y
Sorry
Strangely enough, the corporate world is crying out for graduates who are skilled collaborators–or at least this is what we hear from the National Careers Service. It’s not as though educators have been slow to respond–collaborative learning has been a staple ingredient of CPD for quite a while now.
Considering that collaboration is a biologically primary skill, all this concern seems a bit odd. In ‘Does Education Matter?’ Alison Wolf explains that those employed by organisations like the National Careers Service or the CBI are not business people themselves, but are making careers in a given specialisation such as tax law, training or education. As such, their advancement depends upon staying within the confines of the relevant group-think.
In a rational world, I’m not sure why anyone would want to go to the trouble of crafting the sort of collaborative learning environment outlined in Paul’s recent paper. For a start, unless pupil behaviour is near the top of Terry Haydn’s 10-point scale, at best one will find that loafing takes over, and at worst, behaviour will degenerate even further down the scale. One suspects that fashionable ideas are seldom trialled in schools where pupil motivation and behaviour is not at least adequate, and even then I’m not sure that we always consider the Hawthorne Effect.
Were this not so, it would be impossible to explain the failure of AfL. In his 2013 inaugural address at Durham’s CEM, Robert Coe lamented that “It is now a rare thing, in my experience, to meet any teacher in any school in England who would not claim to be doing Assessment for Learning. And yet, the evidence presented above suggests that during the fifteen years of this intensive intervention to promote AfL, despite its near universal adoption and strong research evidence of substantial impact on attainment, there has been no (or at best limited) effect on learning outcomes nationally.”
Everyone thinks they are doing AfL but hardly anyone has read the original papers (which aren’t long) not is it clearly defined before discussion. It has simply become the good bits of your teaching. Even simple arguments like telling people that mini+whiteboards are not intrinsically AfL but medley a tool for feedback and it is the nature of the questions you ask that matters are lost. But I checked to see if they could answer the question or I asked how they felt rather then trying to discern if they can distinguish between a misconception and a correct solution or checking the boundaries of there knowledge by restating. Good AfL flows naturally from a clear curriculum clear lesson objectives (as in what idea you are teaching not that wordy nonsense in the board) and a clear set of question examples. All domain specific and all perfectly trainable (for the teacher). As always what we say we do is not what we really do and the devil’s in the details. Scalability is a pig.
Pingback: Skills for the 21st Century, Parkland and more in this week's news roundup - Psych Learning Curve
Reblogged this on The Echo Chamber.
There seem to be two different potential benefits to collaborative learning, one where high expertise students can enhance their learning with low transactive costs, and another where high expertise students can guide low expertise students like a scaffold effect.
In this study: Nihalani, P., Mayrath, M., Robinson, D., & Graesser, Arthur C. (2011).’ When Feedback Harms and Collaboration Helps in Computer Simulation Environments: An Expertise Reversal Effect’, they found that for complex tasks at undergrad level a collaborative group outperformed an individual condition for high prior knowledge students. They speculated that: ‘high prior knowledge students who worked in collaborative groups with feedback appeared to benefit from the opportunity to discuss the content, learning task, and possible solutions.’
Whereas in this study Zhang, L., Lei, C., Kalyuga, S., & Lee, C. (2016). ‘Effectiveness of collaborative learning of computer programming under different learning group formations according to students’ prior knowledge: A cognitive load perspective’, they found that for complex learning a mix of high/low prior knowledge students produced the best learning. They put this down to ‘The guidance in solving problems these learners might have received from more experienced learners presumably reduced their involvement in unguided search processes that usually cause high levels of cognitive load.’
These two studies aren’t directly comparable but are these both ‘collective working memory effects’?
It seems like teachers could make use of either in the right scenario.
Pingback: The Alpha-Phonics Blog