Imagine if companies could customize their training so that it met the exact needs of every learner, giving them only the lessons they need and making sure they learned those skills before moving on. This is the premise of adaptive learning and it is going to change the way learning leaders think about training and the content they provide to their people.
Adaptive learning uses algorithms and machine learning to adjust the learning experience on the fly, explained Candace Marie Thille, assistant professor of education at Stanford University. “It uses intelligent educational technology to predict what the learners know, and what they need to learn next.”
From college to corporate
Such customization could have a huge impact on the quality and efficiency of corporate learning environments where there is a wide diversity of skills and knowledge. Courses presented in an adaptive learning environment would respond to that diversity, allowing learners to focus on the lessons they need and skip the rest.
While adaptive learning has its roots in academia, the applications for corporate learning are immense, said Zach Posner, managing director of the learning science platforms for McGraw-Hill Education in New York. “Whether you are in a classroom or using an online module, traditional corporate learning is very one-size-fits-all,” he said. Everyone works through the same content at roughly the same pace regardless of their baseline knowledge or learning style. He sees adaptive learning as a proxy for one-to-one coaching. “It is a scalable way to personalize the learning experience.”
Such personalization delivers benefits to individual learners and the business, he said. Learners spend less time completing content, which improves productivity and lowers costs associated with time away from work. But they are also more likely to stay engaged and to learn more because the content is appropriately challenging. In return, companies get a flow of real-time data on what their people know and where they are struggling, which they can use to tweak current content, make personalized career development plans and identify where new or different programs are needed.
“Adaptive learning opens the door for a new level of sophistication,” said Dan Lovely, chief learning officer of multinational insurance giant AIG in New York. “It enhances the learning experience and gives learners what they need rather what has been deemed necessary for a cohort of learners.”
This isn’t just in theory. Companies like McGraw-Hill, Area9 Learning and Axonify Inc. are producing thousands of titles with adaptive formats and working with customers to convert their in-house content.
“The outcomes speak for themselves,” Posner said. His team recently helped a large financial services firm convert their five-hour online new-hire training course to an adaptive learning format, then they ran a pilot, tracking results in the new program against a group taking the traditional course. The average time spent in the adaptive course was three hours, with some people completing it in 40 minutes. The adaptive learners also scored an average of 25 percent higher on a final test than the traditional learners. “They saw wins across the board,” he said.
AIG is hoping to see similar results with two adaptive learning programs it is currently piloting for underwriter development and compliance training. “Both are in the early stages, but we expect to have results in six months,” Lovely said. He hopes these will be the first of many adaptive learning programs the company will deploy as the technology and content management options advance. “The good news is that the software exists off the shelf to this today, though we would need it to be more flexible to fit our needs,” he said. He imagines that in the future, AIG will partner with a vendor to co-create adaptive learning content in an effort to find that balance.
Doubling Down in Las Vegas
Along with being more flexible, adaptive learning is more fun than traditional training, said Christiana Houck, director of learning solutions for technical training at Aristocrat Technologies Inc., a Las Vegas-based gaming machine manufacturer with 3,000 employees. Aristocrat relied for years on longer online learning modules for its remote techs, who had to take the courses during breaks or after hours. “They dreaded it,” she said.
Last year, she teamed with Axonify to make the content both mobile and adaptive. They rolled out the new format in February. Now, techs get daily microlearning modules sent to their phones and they can review the content in minutes between jobs then answer a few short questions.
“The adaptive part is the test at the end,” she said. “The system uses their score to learn their strengths and weaknesses, then chooses what content to send next.” Every tech needs to master each module twice before they can move on to the next section, though some may see it five or six times before they advance. “They love it, it’s like a game,” she said.
The program initially hit a few small bumps. Houck noted that the learning culture at Aristocrat had always been that once you took a course you passed the test, and there are a lot of techs on her team who never scored below an 80 percent.
But when she rolled out the first set of modules, only 11 percent of techs got an expert rating (above 90 percent) the first time through. When they didn’t master the modules the first time around, some of them started to get nervous. “They felt like they were failing, and that their managers were watching,” she said. Once she explained that they were not expected to pass the first time around, they relaxed.
Houck closely tracks success rates on the modules and uses the data to determine where individuals might need more intensive training, or where the team as a whole has a skill gap. For example, a majority of the 140 techs struggled with a set of meter-reading questions, which is a skill they are expected to learn on the job. “We were making assumptions that they all know how to do this,” she said. When her team saw the scores and compared them to error rates in the field, they realized there was a knowledge gap in the workforce and added a meter-reading course for all new hires.
Axonify has seen many examples of clients discovering surprising knowledge gaps in their workforce, said Carol Leaman, the company’s CEO. In one example, a large retailer in the Middle East assumed front-line staff had core customer service skills, including how to upsell, greet customers and use language appropriate to the setting. But when they gave them an adaptive learning-based customer service course, they mastered less than half of the lessons. “Customer service skills are table stakes in retail, so the results were shocking,” Leaman said.
Once they started sending daily learning modules and tests and using the algorithms to determine who needed to review which core principles, their scores shot up. In a matter of weeks everyone was scoring above 75 percent.
Adaptive learning provides companies with an evidence-driven baseline of knowledge, and charts its progress, she said. “It can be really exciting for executives and employees to see the acquisition of knowledge and confidence grow over time.”
Extension of Micro-learning
Many leading learning organizations have already taken the most important first step toward preparing their learning content for an adaptive environment — chunking their content. Adaptive learning is all about creating lots of small pieces of learning and testing, then threading them together via machine learning algorithms. Lovely said that microlearning has been a trend for a long time, with companies building libraries of videos and tiny bits of content that learners can absorb on the go. “Adaptive learning intersects this trend to create a more productive learning ecosystem,” said Lovely.
The vendors — or in-house teams — may then reduce that content into even smaller chunks, breaking one lesson with three to five learning goals into 40 to 60 microlearning objectives, to get at the heart of exactly what people know and where they have gaps. “We take chunking to a whole new level,” Posner said.
Breaking it into tiny pieces lets learning leaders pinpoint the learning need. “If 100 percent of learners get all the questions right the first time, the module is too easy,” he said. Whereas if everyone fails, it might suggest they need to break that module into a series of even smaller hyperfocused lessons. “The great thing about adaptive learning is you can get that data and respond in real time.”
That analysis and response to data is critical — and it cannot be done by a computer, said Stanford’s Thille. “There must always be a human element.”
She worries that companies think these kinds of technologies can replace human interaction, or that CLOs might rely too heavily on data generated by these systems to make promotion or hiring decisions without considering all the facts. For example, if a learner takes longer to master a topic than their peers it might have to do with their learning style, cultural differences, language mastery or personal issues, all of which could mask their true knowledge of a subject. “It’s fine to use the data to support decision-making, but you must not be uncritical,” she said. As with all technologies, these systems can be biased especially when judging something as complex as a person’s abilities. “That doesn’t mean you shouldn’t use these tools, but be cautious of taking a computer’s word as truth.”