We are living in the world when many disciplines became interconnected and learning shifted from its traditional form. With the speed that the new information is thrown at us due to advancement in a technological development, our brain is using shortcuts to capture, process, and retain that information due to a limited cognitive load and working memory capacity. The human working memory can handle seven pieces (plus or minus two) of information at a time. Hence, techniques derived from the Cognitive Load Theory (CLT) are employed and one of these techniques is chunking, which is a natural processing, storing, maintenance, and retrieval mechanism where long strings of stimuli (e.g. information) are deconstructed and grouped into smaller segments, clusters, or chunks. Consequently, our working memory can handle more information (Portrat, Guida, Phénix, & Lemaire, 2016).
Chunking is a form of sequential learning, which is an important component in self-directed learning. In particular, it simplifies the process of acquiring new information and skills, and task and working memory performance. It consists of three following subprocesses: recognition, reactivation, and maintenance, and it can be strengthened by bottom-up processing (Portrat, Guida, Phénix, & Lemaire, 2016). Fonollosa, Neftci, and Rabinovich (2015) described chunking as a non-linear dynamic hierarchical multi-layer neural network model. They also found that the average chunk size is determined by the parameters of the learning dynamics (Fonollosa, Neftci, & Rabinovich, 2015).
Notwithstanding, chunking falls under the CLT paradigm. Kalyuga and Singh (2016) explored the role of CLT in instructional design, especially in the context of complex learning. According to the traditional perspective of CLT, an explicit instruction and guidance are required prior to completing task for a novice learner to succeed (Kalyuga & Singh, 2016). However, this approach is only sufficient for simple learning. What happens when there is complex learning involved such as solving an advanced physics problem that incorporates myriad skills and concepts, and prior knowledge? Therefore, Kalyuga and Singh (2016) offer an alternative approach, which is more personalized and tailored to specific learning goals. As in, they propose exploring and inquiring about a novel problem prior to seeking guidance and instruction, which requires the retrieval of relevant prior knowledge and schemas from the long-term memory (Kalyuga & Singh, 2016). Consequently, the cognitive load will be drastically reduced when acquiring this new knowledge since it only requires making connections between topics, updating schemas, and chunking information as opposed to viewing that information as an unrelated entity. Moreover, there are three types of cognitive loads and they are intrinsic, extraneous, and germane. The extraneous load, which occurs when “interrelated elements of instruction are separated over distance or tine”, can be reduced by ordering objects in a logical sequence (Syn & Batra, 2013, p. 5). The intrinsic load, which is a number of interacting elements involved in a task, can be decreased by grouping the parts involved in the task into meaningful clusters (i.e. chunking) (Syn & Batra, 2013). The germaine load, which is responsible for the construction of the schemas to improve the task performance, can be reduced by presenting a concept in a form of a pattern of information, which is essentially a schema (Syn & Batra, 2013).
The role of chunking and CLT, in general, can be goal-oriented, and are especially important when learning a new subject or software system, which is a requirement in many fields and industries. For instance, Syn and Batra (2013) conducted a study to investigate whether using CLT and schema-based technique CHOP (chunking, ordering, and patterning) would improve novice analyst performance in modelling a sequence diagram (SD), which is a Unified Modeling Language (UML) diagram. They found that the CHOP technique significantly enhances the novice analysts’ performance in modelling SD, and is significantly more useful than the worked-example approach, where a problem, accompanied with a solution, is provided to the learner (Syn & Batra, 2013). Although for both the worked-example approach and the CHOP technique, the intrinsic load is high, for those novice analysts that used the CHOP technique, the intrinsic load was lower due to the hierarchy-based chunking approach (Syn & Batra, 2013). Furthermore, the extraneous load was much higher for novice analysts who used the worked-example approach due to the absence of schema results, which resulted in the learner using the mean-ends approach, which requires heavy extraneous load (Syn & Batra, 2013). The germane load, on the other hand, was lower for the novice analysts that used the worked-example approach than those analysts that used the CHOP technique since not as much effort was spent on learning and applying new schemas (Syn & Batra, 2013). The results of this study suggest that the worked-example approach requires a heavier working memory capacity than the CHOP technique.
These concepts can also be applied to machine learning, in particular, unsupervised learning, and deep learning. These, in turn, are used in educational settings to create and improve the educators’ and students’ learning experience, academic performance, and need and time for intervention.
Fonollosa, J., Neftci, E., & Rabinovich, M. (2015). Learning of chunking sequences in cognition and behavior. PLoS Computational Biology
(11). Retrieved from http://link.galegroup.com.tc.idm.oclc.org/apps/doc/A437059235/AONE?u=new30429&sid=AONE&xid=9e29e1a5
Kalyuga, S., & Singh, A.-M. (2016). Rethinking the Boundaries of Cognitive Load Theory in Complex Learning. Educational Psychology Review
(4), 831–852. https://doi-org.tc.idm.oclc.org/10.1007/s10648-015-9352-0
Portrat, S. S. P. f., Guida, A., Phénix, T., & Lemaire, B. (2016). Promoting the experimental dialogue between working memory and chunking: Behavioral data and simulation. Memory & Cognition
(3), 420–434. https://doi.org/10.3758/s13421-015-0572-9
Syn, T., & Batra, D. (2013). Improving sequence diagram modeling performance: a technique based on chunking, ordering, and patterning. Journal of Database Management
(4), 1+. Retrieved from http://link.galegroup.com.tc.idm.oclc.org/apps/doc/A369550306/AONE?u=new30429&sid=AONE&xid=1f2ea270