How will AI impact learning?

In 2016, the US Defense Advanced Research Projects Agency (DARPA) trialled the use of ‘virtual’ tutors to train up new recruits. Based on AI (artificial intelligence), the tutor would replicate the interaction between an expert and a novice, reducing the time it would take new navy recruits to get up to speed in technical skills. The trial was hugely effective – recruits who learned from the digital tutor were found to frequently outperform experts with 7-10 years of experience, both on written tests and when solving real-world problems.

AI plays an increasing important role in our daily lives, whether it’s voice assistants such as Alexa sourcing a weather report or an algorithm in Netflix suggesting a new series to watch. On a basic level, AI algorithms – such as ChatGPT – trawl through vast amounts of data to provide responses to questions or carry out simple tasks such as adding an item to a shopping list. Most learning and development teams will be some way off developing AI tutors to lead corporate training programmes, but that’s not to say that AI does not have an emerging role in how organisations share knowledge. This is already happening in a multitude of ways, such as:

• Suggesting eLearning modules an employee might find useful (based on what they’ve done before or other parameters)
• Adapting content or assessments based on how the employee performs (a quiz moves on to a different question depending on the previous answer, for example)
• Generating captions for people who are hearing impaired, or providing text to screen readers for the visually impaired
• Auto-tagging learning content, making it easier to manage and curate
• Providing answers to questions via virtual assistants (such as chatbots and Alexa-type devices)
• Offering learning at the point of need based on context (the system sees the employee has a sales meeting, so offers a short course on sales techniques)

Far from seeking to completely replace the human aspect of training, in many ways AI can augment it or at least make it more cost-effective. “In the race to deliver impact in a post- pandemic world, organisations will need to differentiate themselves from their peers,” says Sean Farrington, senior vice president for EMEA at technology skills platform Pluralsight. “Intelligent technology skill development, backed by AI, will provide an opportunity to expedite go-to-market times and deliver a competitive advantage.” This can happen in two ways, according to Jonathan Crane, chief commercial officer at IPsoft, manufacturers of AI tool Amelia. On the one hand, a digital ‘colleague’ can take over mundane tasks and simple transactions, using machine learning to improve its performance; on the other, it can enable employees to learn through doing because they are supported. “If you want to get better at something, you take the people who know what they’re doing. AI is constantly learning so can pick out material where you might not know the answer, or help you to prioritise a response. Collaboratively you’ve learned and augmented your skills, and this is the biggest benefit of AI,” he says.

Personalisation is a key element of this. Similar to the way consumer sites such as Amazon and Netflix make recommendations to us based on items or programmes we have viewed before, LMS software can use AI to offer recommendations of courses or content that might be useful. The delivery mechanism for this could be in the form of a chatbot or in the way the courses are presented, with a ‘recommended’ stream prominent on the opening page. Toby Gilchrist, head of implementation services – LMS at Ciphr says this can help employees understand where their role fits into the business, and how learning can support that role. Digits LMS, for example, gathers data on employees’ skills through a questionnaire, which benchmarks them against others in their department or in similar roles across the business, which all contributes to personalising individuals’ learning experiences. “The guidance is important for the self-led aspect of learning – recommending what might be relevant for them and nudging them into building their skills,” he explains. Businesses can introduce parameters for recommendations based on insights from connected talent management and HR systems: so if someone is on track for a promotion or has shown interest in a particular career path, this can feed into the courses they see. AI can trigger reminders that someone needs to complete annual data protection or other compliance training, saving time and administration for the HR and learning teams.

Adaptive learning is another way in which AI can personalise content and ensure the right knowledge gets to the right employees at the right time. This could be as simple as a quiz that uses a basic algorithm to ask more advanced questions based on a certain response (or to reinforce knowledge where someone has failed to grasp it), or a sophisticated gamified learning programme that offers tailored scenarios depending on decisions made in previous stages. In both cases, the algorithm has learnt something about the user’s level of knowledge and is able to offer appropriate learning content at this level. Laura Baldwin, president of O’Reilly, an online learning platform, says this can deliver knowledge at the point it is needed rather than all learners having to start from scratch. “We believe in structural literacy,” she says. “If you know [programming language] Python you probably know basic functions in other programming languages. Rather than assuming everyone has to start at the beginning, a good AI platform will know this.”

Any interaction with learning systems will of course produce data, and AI tools can use this data to further personalise course content and predict what might be useful in the future. In the background, algorithms produce data on learner behaviour that can help organisations see where learning has been consolidated and where particular individuals and teams need more support or are speeding ahead. It can also help learning teams see how the design of courses affects how employees interact with them, enabling them to tweak where necessary.

Pluralsight’s algorithmic engine, Iris, uses data to build a fuller picture of individual learning needs but also those of the whole organisation. Farrington adds: “With every assessment and course completed, [Iris] absorbs information about the state of technology skills, collecting feedback and adapting as learners seek skills in new technologies. It uses this data to inform technology strategies and recommends what skills are needed to keep pace with change.” Crucially, it can ensure learning investments are targeted more effectively. “When businesses do not use technology to underpin their employees’ skill development, they often embrace a one-size- fits-all approach. This tends to result in training being delivered in classroom scenarios or through presentations by external consultants. It’s costly, inefficient, and immeasurable,” he argues. Through intelligent recommendation, the ability to tag courses more easily, and predictive analytics from data, teams can curate bespoke learning experiences more simply – and at lower cost.

AI is already supporting accessibility to learning – think of the autogenerated captions on YouTube – and this is likely to become even more advanced in years to come. One of the pioneers in this area, Microsoft, has developed an app called Seeing AI that ‘narrates’ the world around someone in a range of languages so visually impaired people can ‘recognise’ colleagues or read text and websites. Algorithms can also distil information for employees who may have reading or cognitive difficulties, for example summarising lengthy articles or providing snapshots of articles for those that feel overwhelmed by information. We have become used to voice assistants such as Alexa and Google answering questions in our homes and these tools are beginning to be deployed in a workplace context, too. The natural language processing and algorithms that support voice assistants to answer our questions are evolving, so arguably the natural next step is to develop ‘robot’ teachers that can deliver courses as we saw with the US Navy. Developments in public education are moving more quickly on this than in business. Swedish educational technology company Furhat Robotics recently piloted the use of a ‘social robot’ to educate school children on how to code, while Colombian start-up Van Robotics created an educational robot, ABii, that functions as a personal tutor to children with learning challenges. The robot monitors children’s attention cues and adapts to their learning habits, collecting data on their progress as it does so. Time magazine named it one of its top inventions of 2020, and it is now used across schools in 20 US states.

Such developments will seem terrifying to many, who will have concerns over losing the crucial human element of learning and the ethical implications of allowing a robot to determine educational outcomes. It will be some time before we see such technology deployed in corporate classroom learning, but some industries such as healthcare and engineering are already exploiting virtual and augmented reality as a safe way to offer learners immersive experiences of intricate tasks – and this is powered by AI. “Learning through experience, rather than watching as a bystander, brings about real behavioural change,” explains Thompson from VirtualSpeech. The algorithm monitors how employees behave?in?certain?theoretical scenarios and adapts the learning based on behaviour. “It measures audience perception so if you’re coming across as argumentative it might suggest an in-app tool on developing better eye contact,” she adds. “AI helps guide the learning and personalise it.”

Looking to the future, AI will play a bigger part in scanning the horizon for potential skills gaps and augmenting what human teams do, rather than replacing them, argues learning consultant Nigel Paine. “It will always be used to help define individuals’ needs and preferences, but there will also be a much bigger role,” he explains. “It will provide early indicators of skills issues before they arise, package learning for individuals, and help them to make better decisions with more access to data. It will prompt us and help us to operate smarter and faster – but it won’t replace us.”

Five key takeaways

• AI’s usefulness for L&D teams is already being demonstrated, with its importance set to grow in the coming years
• Don’t regard it as a threat: think about how AI can help the human members of an L&D team make more intelligent training interventions, and cut down the time they spend on repetitive or labour-intensive tasks
• Remember, AI is only as useful as the data it is trained on. If there are flaws or biases in your datasets, algorithms will only serve to reinforce these problems
• Expect to see AI enabling more personalised learning, and adaptive learning, soon
• As AI’s importance at work more widely grows, your people will need training on how to leverage it most effectively, too

This is an extract from?Good Work, Great Technology: Enabling strategic success through digital tools, published by leading UK?HR software provider Ciphr. For more insight into how technology can change work for the better,?download the complete book for free, now