Visualising Deep Learning

Part of the Visualising Education Blog Series

The idea of DEEP learning has been batted around for mainly years but few people truly understand the implicit knowledge which is being inferred to when it is used. The word DEEP is an acronym based on cognitive brain research, which discovered for learning to be retained for a period of time it needs to achieve one or more of the following: Distinctive (Eysenck & Eysenck (1980)), Elaborate (Craik & Tulving (1975)), Effort (Taylor et al (1979)) and Personalised (Rogers et al (1977)). I will purposefully avoid talking about fast and slow thinking, which is also referred to as parallel and serial thinking. However, DEEP learning and thinking are much more than the long-term ability to recall surface knowledge such as facts, it is how we use this knowledge to critically engage within the world around us.

The brain does not remember information and events in our spoken language, rather it uses pictures, sounds and meaning (semantic) (McGill 2016). Therefore, every time we speak our brain has to translate from the stored format to a language format which creates a pattern specific to the person’s native language.  This is empathised by the statement, ‘we know more than we can say and we can speak more than we can write.’ This statement is often attributed to David Snowdon from the Knowledge Management domain but there are references to similar statements from other domains too. However, the point I am making is that the brain brings together stored memories or information chunks which is used to construct meaning, thought or a response. If the pattern (sequence of sounds i.e. word) is correct the connections between the information chucks, become stronger increasing recall potential. It is important to acknowledge that others have proposed different approaches to understanding long-term memory such as Tulving (1985) indexing, who suggested that memories are group by semantic, episodic and procedure.

There is an argument that the learning process is explicit (easy to explain) in the beginning, as the connections need reinforcement before they become implicit (hard or impossible to explain). For example, when learning to ride a bike we learn to hold the handlebars and push down on the paddles. However, as this action is reinforced and practiced it is refined by considering how hard we hold the handlebars and how much force is applied through the peddles. Then, somewhere along the way, we stop thinking about it and it becomes autonomic.   This is the same for learning to write or draw. These patterns are typically referred to as schemas (the name can vary between domains), which firmly situates the learning process within the cognitive domain.

Piaget (1952 and 1964) is often accredited with the development and application of schemas, however, philosophers such as Aristotle (BC350) and Immanuel Kant (1724-1804) also refer to similar concepts. (Top tip, if you want to know about schemas ask someone with a childcare degree.) In very simple terms a schema is used to recognise objects and events. For example, ‘cup’ is a schema and, for argument sake, a cup has a handle and holds liquid. However, there are different types of cups i.e. mugs, tea cup, fine china cup and so on. The problem is that as individuals we all have our own personalised schemas.  For example, is a plastic bag a cup? It comes down to how the different types of cups are categorised. For a person like me a cup is anything I can drink out of and I have little or no interest beyond this. However, for my wife, there is an extensive range of cups and etiquette around using different cups.  The argument could be made that my lack of specific terms for ‘cups’ suggests that a) my schema is limited by my vocabulary or b) I do not know the difference between the different types of cups. This brings us back to how information is stored in patterns within the brain.  Where the schema is a visual representation of the pattern within the brain and the idea that the brain or knowledge and understanding can be represented in a three-dimensional spider diagram.

For the schema adjustment (change based on new information) or assimilation (integrating new information) process to take place, learners need to challenge their own understanding (effort) through active engagement with the learning process, which Piaget referred to as active learning.  By making the learning or content distinctive and elaborate the learner is more likely to remember and engage within active learning and apply effort, helping the learner to personalise the content. However, to support schema formation and cross-referencing, between different schemas, targeted vocabulary needs to be reinforced. Critically, problem-solving success is linked to a person’s ability to merge or cross-reference multiple schemes (Chi et al (1981), Marshall (1995, pg62)).

In summary, I have outlined the core components of deep learning and situating it within the cognitive domain of learning through the use of schema adjustment or assimilation. I have suggested how teaching and learning can be shaped to consider this approach to promote deep learning.

Link to info Proc Model: