AI has been getting much attention in recent years because of its various uses in commercial applications like face recognition, speech recognition, personal digital assistants like Apple’s Siri Amazon’s Alexa, and self-driving cars. The reality is that we now have the right combination of computing power, the Internet of Things (IoT) and powerful machine learning algorithms to greatly enhance everything we do in business today. Naturally, we can think that if businesses are already leveraging AI; then, sooner or later Learning and Development (L&D) will need to follow suit. Let’s take a look at various concepts of AI and how they can enhance the reach of your L&D function with automation and impactful analytics.
Much of the credit given to AI’s abilities today really belongs to Machine Learning (ML). Often described as a subset of AI, Machine Learning focuses on the way algorithms learn from data, experience, and inference. For example, enabling computer systems with ML algorithms and assessing data from video cameras, sensors, radar and GPS signals, allows self-driving cars to recognize road signs and hazards. ML can be applied in various forms such as supervised, unsupervised, reinforced and deep learning. Since ML is about systems that can learn from data, then it makes sense to explore how it fits the purposes of L&D. Next, we will explore how each type of ML can be applied in learning analytics.
Supervised learning is a very simple concept of machine learning. Supervised algorithms can analyze data and known results of that data to predict future outcomes. The more data that’s available, the more accurate the system would be. Supervised learning is used in several industries like insurance, customer service, and medicine. For example; a supervised system can analyze worker performance and course results to recommend courses associated with levels of high performance. The same can be done with assessment and competency data. These algorithms can analyze the metrics of your top-level performers and make predictions about how well the rest of the organization will perform in comparison.
Unsupervised learning is a variant of ML that’s really helpful to classify data and establish groups by finding similarities in the data. This is why unsupervised learning is widely used by companies to better market their product by identifying and establishing customer segments from their customer data. How can unsupervised learning be used in L&D? An unsupervised system can analyze data on your learning content and determine which training content type produces best results i.e. ILT, eLearning, video, informal, etc. The main advantage of using these algorithms is that they won’t be skewed by human bias or what we think we already know.
Reinforcement learning is the closest thing to experiential learning for computing systems. Reinforcement algorithms learn by experience by performing trial and error tasks. These algorithms need to “practice” several times to maximize their chances for success while avoiding failures. Prime examples of this type of machine learning are robots learning how to walk or open doors. However, this approach is not limited to physical tasks. It can be used for strategic decision-making as well. Carnegie Mellon University used it for Libratus, an AI system that beat several top poker players this past January. L&D can use reinforcement learning to have systems that can autonomously decide what training approaches are more likely to achieve desired results. For example; imagine a system that can send case scenarios about a specific task to analyze worker responses. The system would then correlate worker performance on those tasks to their performance on the case scenarios. The reinforcement learning can then adjust the training content to maximize the level of worker successes vs. failures to meet performance standards.
No other type of machine learning design is as fascinating than deep learning. This is because deep learning algorithms work on a schema of calculations known as neural networks. These networks have been compared with the way our brain neural connectomes work, but we can consider them to be just the most basic form of that. Deep learning is still based on inputs and outputs, but it’s what’s in between the input and output layers that make a difference. There can be three or millions of hidden layers that would assign weights to variables and the more layers involved the deeper the data connections that can be made. Deep learning is used for complex tasks in systems that use facial recognition, speech recognition and haptic sensory. The hidden power of deep learning is in detecting anomalies or outliers in data. For example; credit card companies use deep learning to prevent and detect fraud by analyzing transactions and purchasing behaviors. So, if you always buy apples at one market location, but now your credit card is used at another distant market and no apples are purchased; then, these algorithms can flag this behavior as potential fraud. There are tremendous possibilities for L&D and performance support using deep learning. Imagine sales call centers that can analyze conversations, facial expressions and keystrokes to identify the behaviors that lead to closing deals more effectively. Factory floors can have camera systems and sensors to flag risk behaviors compromising workplace safety.
Implementing AI in L&D
So far, you have read about all the wonders of AI with machine learning, but there are some realities L&D must deal with before even thinking about pursuing them. First, you need the right mindset to venture into AI. It’s not a thing that will replace you. The technology is just a tool, it’s what you do with it that makes a difference. Second, AI with machine learning requires some computing power and this means several servers with high-performing processors. So, find out if AI is been used in your business as it would be easier to use existing computing resources than trying to procure your own, especially for learning purposes. Third, think beyond all you know. What you think you know right now it’s very likely to be biased, so think about a vision of possibilities beyond what you can currently do.
The aim of this article was to bring learning professionals closer to AI as the powerful tool it can be. Machine learning is a subset of AI responsible for most of the greatest advances seen in computing systems performing human-like tasks. Supervised learning can predict outcomes from historical data with known results while unsupervised algorithms are great at finding similarities in data to classify groups. Reinforcement learning applies experiential approaches to learn from mistakes and successes and deep learning truly fine tunes a systems ability to detect outlying information without assumptions. This is the realm of AI, you learn from it as it learns from you. It will change the way you think and provide learning solutions.