We’ve all seen it. The deep fake political ads. The CEOs using AI-generated avatars to host company meetings. And now, every new Ed Tech vendor in L&D wants us to try AI video avatars because they’re the NEXT BIG THING that will solve all your organization’s learning, retention, and performance problems.
I call bullsh*t. This reminds me of when every new hire onboarding was going to be AR/VR powered (spoiler alert: I’ve sat through more than one virtual onboarding with good ol’ PowerPoint slides since then). It also reminds me of when EVERYONE was using those videos with the hand that writes things on screen. I have the same reaction to that as I do to the outfits I wore as a college freshman – some things are better left in the past.
But AI video avatars are not necessary or even helpful during learning, and from a research-backed standpoint, we’re going to talk about why:
AI Video Avatars Are NOT More Effective
Despite their increasing use, there's limited evidence that AI-generated avatars significantly improve learning outcomes. A study on the use of AI avatars in a business ethics course found that while students accepted and acknowledged the technology, there was no clear indication that it enhanced their learning experience. The authors say, “Yet, for the most part, the students interviewed had neither strongly positive nor negative attitudes to seeing and hearing their subject content embodied in an AI avatar.” The avatars, though realistic, failed to replicate the relational and social dynamics critical to human learning. One study even showed that participants scored higher when content was presented by a human professor than they did with an AI professor. That study found that it was the content and how it was presented by a real human that made the difference – not the avatar. Across the board, no significant difference has been found in the studies done.
There was one study from 2011 that claims “positive effects of an avatar activity that engaged the authors and their peers during two graduate-level collaborative online classes.” They argued that the use of pedagogical agents, or avatars, helped increase the social presence and community as well as the trust in using the online class space – all because of avatars. However, THOSE avatars looked like this:
And I have to admit – if Butters from South Park was narrating my information security compliance course, I would probably enjoy it a lot more… but if this is all it takes to engage learners, then we don’t need the cost of replicating real humans.
AI Video Avatars Increase Cognitive Load
Cognitive Load Theory suggests that the additional visual and interactive elements introduced by avatars may increase cognitive load, potentially making learning more difficult rather than easier. Because there’s no significant evidence that they make a positive impact on the learning experience or on retention, these avatars can be defined as seductive details, which I wrote about avoiding for TD Magazine. Seductive details are anything that is designed to be more interactive and engaging for the learner but has no direct correlation with the learning outcome. So, if you’re using an AI video avatar because you think the course needs a pedagogical agent and it’s cool and engaging – unfortunately, you’re just adding more cognitive load for your learners.
Research Focuses on the Response to AI Avatars, NOT the Impact
There's a notable gap in rigorous research comparing the effectiveness of AI avatars versus traditional human-guided learning. Although some studies explore the potential of pedagogical agents (characters guiding learning), they often fail to conclusively demonstrate that these agents outperform human instructors or even non-character-based digital tools in terms of learning outcomes. Many of the studies out there right now are being sponsored by these Ed Tech companies and are specifically looking at the response to the AI avatars and not their effectiveness when compared to other delivery methods. They’re like “hey, people really think this is cool.” But “cool” and being able to use them as a tool to do your job better are two VERY different things (I’m reminded again of my college freshman outfits here - “cool” but not functional…)
Perhaps you’re thinking, “Okay Heidi, then what do you suggest?” Instead of clinging on to trendy, underdeveloped technology, let’s stick with what we know works:
Bring Humans Back in Focus
Emphasizing human facilitators or mentors who can provide personalized feedback and adapt to learners' needs in real-time has been shown to be more effective in fostering deeper understanding and engagement. Create something asynchronous, and then provide virtual workshop sessions where you can dive more deeply into topics and have real human mentors provide the amount of nuance appropriate for the subject.
Simplify Your Design
Reducing cognitive load through well-structured, clear, and concise multimedia content without unnecessary visual or interactive elements can enhance learning outcomes. Keep it simple. I know I keep saying it, but it keeps needing to be said: “The most engaging thing learning can be is useful.” It doesn’t need all the bells and whistles and seductive details to be effective.
Use Better Learning Strategies
Incorporating active learning strategies, such as problem-solving tasks, scenario-based learning, social learning, and hands-on activities, are often more effective than passive learning methods involving pedagogical agents or AI video avatars. Contrary to popular belief, you don’t need AI video avatars for your scenarios either – stock photos or amazing descriptions will suffice. And if you don’t believe me that a well written description is enough in a scenario, just go ask your favorite book snob what they think of the latest movie adaptation of their favorite book.
Although AI avatars are novel and interesting and all that everyone is going to be talking about during the upcoming conference season, they are not yet a proven necessity in corporate training or education. More research is needed to fully understand their impact, and L&D pros should be cautious about replacing human interaction with AI until there's stronger evidence to support their effectiveness. In the meantime, let’s continue to use proven useful ways to help people work and learn better.
