Learning Science as a Tool to Drive Outcomes for Your Training Programs
At Sana Labs, we are passionate about improving how people learn. In fact, we consider it to be one of the most important problems to solve globally. Learning is a meta problem: if we improve how we train people and in turn teach them how to apply what they learn, we can improve everything else. To be concrete: if we improve how nurses learn, we can improve how they treat patients and in the end save lives. If we improve how researchers learn, we can improve how medicines are developed and in the end help cure diseases, and so on.
Cognitive and learning science has long been a lighthouse for professionals wanting to adopt evidence-based learning design. However, the prevalence of training programs intentionally designed with a learning science approach is still moderate. Current research is powering a new mindset that is increasing adoption day-by-day.
Implementing learning science into your training programs can be a powerful way to improve how your organization learns and to boost learning outcomes. This white paper aims to introduce learning science concepts that can be used (as a smorgasbord) when designing training programs for your workforce. It also sheds light on important vocabulary that acts as a common communication framework for learning professionals.
Prerequisites for learning: motivation, novelty, control, challenge
In order for (effective) learning to happen, there are four main prerequisites that need to be met: (1) motivation, (2) novelty, (3) control, and (4) challenge. Your learning teams can impact all of these prerequisites, but they also depend on the mindset of your learners.
1. Motivation: One of the key drivers of learning is motivation, which is also one of the common areas where learning programs fall short. An unmotivated learner might, at best, waste time ticking boxes, and at worst, negatively impact their job performance.
A motivated learner will acquire skills and apply it in their everyday job to drive business performance.
In order for a learner to be motivated, there are prerequisites for the learning experience and the learning content. The learning experience must be aligned with their skills gaps, career ambitions, work application, and timeliness. It also must be an engaging and meaningful experience, and by delivering it in a delightful product will increase motivation.
There are two types of motivation:
- Extrinsic motivation: Extrinsic or external factors driving motivation may include earning a certificate for completing a course or learning that gives the learner credit towards a promotion or ranks the learner on a public leader board. Gamification is often associated with this type of motivation. It is also related to a ‘push’ or mandatory type of learning.
- Intrinsic motivation: Intrinsic or internal motivation is often associated with the learner’s belief that completing a course will positively impact their job performance or improve an area of personal interest or simply the self satisfaction of accomplishment. This is also related to a ‘pull’ or curiosity-based type of learning.
Whether or not to optimize the learning with extrinsic or intrinsic motivation depends on a large variety of factors including your organization’s learning culture, the individual learner, regulatory factors, and more. Ken Bain (2014) studied the concepts, and concluded that offering external rewards to intrinsically motivated learners might negatively impact their performance.
2. Novelty: What’s worse than an unmotivated learner is a learner that is forced to sit through learning they already know. Our definition of learning is the “process of gaining knowledge or skill by studying, practicing, being taught, or experiencing something.” That is, in order for learning to happen, the learner needs to experience something new (Schomaker, Meeter, 2015).
3. Control: Weiner (1986) showed that learners who feel that they have more personal control of their learning are expected to do better than learners who do not feel in control of their learning. Factors might include the ability to influence the available learning assets, the ability to browse learning assets and self-enroll in courses, the ability to influence which instructor-led-training programs are offered, etc. A learner who has decided to enroll in a course based on their own motivation might be more likely to learn more.
4. Challenge: Learning content needs to be at an adequate difficulty for learning to happen. Also, it is important that learners know the prerequisites of a course before they start it. Delivering content that is too difficult will more likely result in learners churning before they complete the course. Delivering content that is too easy will more likely result in learners becoming bored and churning before they get to fill their knowledge gaps. Although it is highly dependent on context, we’ve found that the optimal difficulty is when learners have approximately a 70% likelihood of correctly answering formative assessment questions.
Process of learning: schema theory, cognitive load theory, retrieval practice
In order to optimize the process of learning, make sure to implement evidence-based learning design.
Firstly, pay attention to the schemas.
Schemas are cognitive structures that organize knowledge stored in our long-term memory, often forming unconsciously.
A schema contains groups of linked memories, concepts, or words. This grouping of things acts as a cognitive shortcut, making storing new things in your long-term memory and retrieving them much quicker and more efficient.
When designing training programmes, explicitly defining schemas can be a way to expedite the creation of schemas in the mind of the learner. Do so by providing an overview of where the learning fits into a larger context, and continuously show where the learner is within the course curriculum and how the course curriculum fits together. If your training involves abstract concepts, facilitate the creation of schemas by providing concrete examples — it will help the learners create schemas in their memory and associate the concepts with their own experiences.
Secondly, optimize the cognitive load. The Centre for Education Statistics and Evaluation in New South Wales, Australia (2017) summarizes the concepts well: The brain’s working (or short-term) memory has a limited capacity. Overloading it reduces the effectiveness of learning. There are 3 types of cognitive load: intrinsic (how complex the task is), extraneous (distractions that increase load), and germane (linking new information with what’s already stored in long-term memory). The germane load is the cognitive load that makes us learn; the load associated with schemas being formed in our mind. If the training requires too much intrinsic and extraneous load, there is no space for the germane load to happen.
Hence, it is a good idea for a content creator to decrease the intrinsic and extraneous load of the content. Dual coding theory (Sweller, 1988) suggests that using a combination of images, a small amount of text and narration (visual and verbal stimuli) is the most efficient way of reducing extraneous load. The coherence principle argues for identifying only the necessary parts and distilling them to reduce the amount of information in each step of learning. The signaling principle argues for highlighting the most important information by arrows or circles to create a visual memory.
Thirdly, use retrieval practice as an active learning technique to boost learning. Retrieval practice, at its core, is when a learner recalls a piece of information without having it in front of them. This technique has been shown by cognitive science to improve memory as concepts become easier to retrieve in the future. It also strengthens or expands existing schemas, and helps learners recognize knowledge gaps.
There are multiple ways to practice retrieval, including formative assessments, flashcards, concept maps (start from a keyword in a topic, and then adding keywords related to the topic), brain dumps (writing down everything a learner knows about a topic), elaborative interrogation (a learner asks “why” and “how” questions on a topic, and attempts to answer). Many of these tools are used today as summative assessment tools, but they should be more commonly used as learning tools to improve outcomes. Read more in “How to use retrieval practice to improve learning” (Agarwal et. al., 2020).
Consolidation of learning: spaced repetition and interleaved learning
In order for your training to be effective, it is important that it translates into long-term memory. While one can hope that learners will remember the information over time, potentially by applying it on the job, there are several techniques that will maximize the likelihood of translating the output of training into long-term memory. Two of these techniques are spaced repetition and interleaved learning.
One important concept in spaced repetition theory is called knowledge decay. Knowledge decay was first described by the psychologist Hermann Ebbinghaus (1885), who found that the memorization of knowledge decreases exponentially if not reviewed periodically.
Paradoxically, forgetting can be a good thing. If one partially forgets concepts in-between reviews, they will develop stronger long-term memories than if one reviews concepts immediately after completing the training.
Ebbinghaus found that memorization was more efficient if he spaced out short review sessions over time, rather than cramming reviews all at once. Landauer and Bjork (1978) showed that expanding review is better than uniform review in two well-controlled experiments. By strategically reviewing content at optimal intervals, significant improvements to long-term memory can be achieved.
Interleaved learning, in short, is switching between topics in a study session as opposed to focusing only on one topic at a time (Carvalho, Goldstone, 2019). This is the opposite of “blocking practice,” i.e. focusing on one topic at a time, the most commonly-used method for structuring learning. Interleaved learning has been shown to help with concept learning and generalization. In particular, it helps you differentiate between multiple concepts and apply them correctly in a given setting. It does relate to spaced repetition, and in an optimal scenario spaced repetition and interleaved learning work in parallel. In a study completed by the Sana Labs research team, we found that breadth over depth in topic matters is associated with higher exam success rate (Cristus et al., 2020).
How the Sana platform makes use of learning science to deliver more efficient learning for your workforce
Our learning platform, Sana, uses personalization and adaptive learning, coupled with a delightful user experience to overcome some of the common challenges of meeting learning prerequisites, augmenting the process of learning, and consolidating the learning.
- Adaptive learning makes sure to keep every learning interaction at the optimal difficulty.
- Spaced repetition is built in as a core part of any course.
- Retrieval practice is a core component of any learning programme in Sana embodied by the formative assessment questions driving the adaptivity.
- Sana optimizes for breadth over depth by covering all chapters before recommending mastering the content.
- Integrated nudging gives the learner a feeling of control by getting indications on areas that they’ve struggled with.
Building adaptive learning content inherently satisfies many of the properties advocated by learning science. For example, by breaking the learning down into granular pieces, Sana makes sure to decrease cognitive load. By automatically serving up image recommendations from high-quality stock libraries, Sana makes use of dual coding theory.
Given the complexity of learning science practices, learning professionals need tools that support them in delivering effective training programmes backed by research. We are passionate about being at the forefront of improving how the workforce learns, and we keep learning science principles at the top of mind with every feature we develop in Sana.
Thank you Ken Hubbell, Christian Horne, and Ulf Änggårdh for reviewing drafts of this. If you want to learn more, please send a note to samuel@sanalabs.com
