Dave and Justin discuss the possibilities of AI assisted content creation in education and skills.
As well as the ethical perspective of AI content creation in learning, what are the concerns of learning businesses themselves?
An obvious one is the advancement of large language models. Imagine being able to ask ChatGPT to educate you on level-one nursing, as it becomes highly trained in such content. This scenario could potentially put learning platforms out of business, with OpenAI becoming the owner of that space. While this concern looms large for companies, we’re not there yet. Language models have yet to be extensively trained on such specialised content, and the cost of keeping them updated still poses a significant hurdle.
So how can learning businesses instead take advantage of the advancement of AI, and specifically large language models?
AI can excel in personalising course content for learners with specific skill gaps and goals. By tailoring modules and learning approaches to individuals’ needs, it enables a more effective and targeted educational experience. Currently, many learning platforms are very prescriptive. If you signed up to a course on how to sell through Amazon for example, you’d have to work through all the modules, even if you knew much of it already. In order to work out what you don’t know, you generally have to work your way through all of it. Now there is the alternative of using a large language model to assemble the content from a library of content based on what it knows about you, and the questions you’ve asked.
And is this an area of personalisation that businesses are already exploring?
Smart ones - yes. Often, an employee will be signed up to a learning platform by the company they work for. That company has a set plan of what they want the employee to learn, having identified their skill gaps. The platform would already have that information, so you’d be passing that, plus a catalogue of your content, and it would come back with a recommended learning pathway through that content for the specific individual. It’s creating a roadmap of learning content, and learning tasks for you to undertake, and then guiding you through it. Also, if you’re submitting written work or answers to tests, it’s giving you feedback on where you need to improve. That’s the level of integration, and interaction with a learning assistant.
What are your thoughts on the advancement of virtual assistants in careers’ guidance?
It’s exciting. Imagine a virtual assistant integrated with a Learning Management System (LMS) equipped with extensive knowledge about various career qualifications. This assistant could craft a tailored pathway for you. For instance, a 15-year-old interested in nursing would receive information about available courses and suggested work experiences, guiding them towards the necessary qualifications.
And a virtual assistant serving as a career advisor isn't just limited to eLearning platforms?
No - it could also benefit schools, colleges - even businesses perhaps focusing on corporate social responsibility (CSR) and internal career development. For instance, take a career assistant specialised in pathways to telecommunications careers, in areas such as programming or networks. This assistant might ask your age, existing knowledge and qualifications, then provide a tailored list of steps to prepare for job applications. Additionally, it could suggest relevant work experience opportunities aligned with your goals.
What format suits this type of interface?
Well it could be an avatar for spoken or typed interaction, or a simple text-based chatbot. Experimentation with different environments is well underway. For instance, employing game characters within a gaming environment, where you communicate with a character through your own avatar, might cater to a younger demographic. This contrasts with the traditional conversational chat interface, offering a more engaging experience.
Sounds expensive?
Yes and no - definitely within the realms of possibility. Several ongoing projects are already integrating AI to generate 3D game model environments and characters. They generate output compatible with platforms like Unity, enabling the creation of in-game video footage. This is already happening, and Unity is the kind of platform capable of rendering virtual reality or augmented reality environments in a similar fashion. Educational studies have indicated that learning outcomes improve significantly when individuals engage with a human face rather than audio recordings. Visual cues, such as facial expressions, foster empathy and enhance concentration, leading to a more impactful learning experience.
Helping our learning and skills clients utilise AI, what’s the most important skill we’re providing?
It involves understanding how to fine-tune and effectively prompt the model to understand the required task and submitted content from the learners. Additionally, it’s about sanity-checking and cleaning up work uploaded in all manner of ways; some students structure their work better than others. Furthermore, it's about cleaning up potential training data and ensuring there is enough data for training and testing responses. This data is then used to refine the model's understanding of specific subject areas. Going forward, the model can better infer a student's understanding of a subject based on their work, having learned from previous examples.
What currently most interests you in the work we’re doing?
Simulation-based learning. I found the following charts within a research paper titled "Educational applications of artificial intelligence in simulation-based learning: A systematic mapping review." The paper examines the current focus of educational institutions on the utilisation of AI and smart technologies. It reveals a greater prevalence of intelligent tutoring systems compared to mixed reality and virtual reality simulations. Concerning direct AI applications, the intersection of virtual agents and assessments appears to show a current sweet spot, aligning with both user demand and AI capabilities.
Which offers the most practical solution, would you say?
In terms of value - intelligent tutoring systems. Utilising virtual reality or mixed reality requires audience access to expensive equipment, like headsets, which many either can’t afford or aren’t yet ready to embrace. While theoretically offering an exceptional learning experience, the reality is that the audience is limited, and implementation requires significant time and financial investment. This is why widespread adoption hasn’t yet happened. However, as people are becoming more accustomed to co-piloted or virtual assistant setups in their daily routines, particularly in professional and educational contexts, there's a growing acceptance of this technology. Having a virtual assistant guide individuals through their learning journey, providing advice and reviewing work enhances user and customer experience. It also improves operational efficiency by reducing the need for human intervention and its associated costs.
The third chart illustrates that combining virtual agents and assessments yields a consistently high level of performance across various learning types. The yellow line indicates a strong correlation with effective learning outcomes. A standalone virtual agent demonstrates less strength in certain cognitive aspects; however, it performs well in scenarios involving social learning, where conversational engagement is prominent.
What’s your conclusion?
Well, it might seem an obvious use case, but the crucial question is whether individuals are actively implementing or discussing its implementation. From a learning standpoint, can we confidently assert that learners will indeed benefit from this approach? The evidence presented in this third chart suggests that they can indeed expect an improved learning experience.
Originally published on LinkedIn
