By Nick Jensen
My first face-to-face with an AI conversational avatar was Christmas 2023. I was with my extended family and my brother showed me what he had been working on. He is an actor, fresh from four years playing Hagrid in Harry Potter and the Cursed Child show in Melbourne, but during that period he and a techie mate had brought this new concept to life.
I sat down, and was transported into a 3D representation of a doctor’s office. A young woman sat in front of me, and started talking to me about how she needed a doctor’s certificate for her university exams as she had been unwell. I talked back; doing my best to sound wise and authoritative as a medical professional. It was surreal.
After the experience, my brain started to whir. I had been working in education and training for the previous 15 years (schools and universities mostly), and the potential use cases started flooding in. They could now create any person, in any environment, in any scenario, on any device! Historical figures, teacher training, onboarding scenarios, professional development, legal debates, school registrars - the only limit seemed to be my own imagination.
Fast forward a year and I had stepped out of my previous job to start my own company based on this technology. I launched ACTS Education (Avatar Conversational Training Simulations) at the start of 2025, an educational arm of my brother’s Evolve Simulations. My vision was to see this technology piloted in schools, universities and training institutions.
The idea is simple. All the avatars are connected to a large language model, and then additional layers of prompts are added to the personality framework. The 3D avatar is designed, then linked to voice, lip synchronisation, movement, demeanours and guardrails. The result is what genuinely feels like a conversation with a digital human representative.
The more I immersed myself in the technology however, the more I realised that avatar technology was not just an interesting product providing fun new ways to simulate education and training, but in many ways a significant pedagogical evolution. The impact on engagement, assessment, scale and the ethical questions that arose from it were all venturing into entirely new and unexplored territories. Allow me E-A-S-E you into it (I am so sorry!):
The challenges for learners in training are similar to the schools and higher education contexts. Student engagement is vital for genuine knowledge retention. Additionally, the gap between theory and real-world application is sometimes hard to cross. The challenge is how we turn knowledge into applied wisdom. Learning about classroom management is one thing, stepping into a classroom and figuring out the best applications is another.
For example, I recently completed a short, online course in child safety and protection, and I have to confess it left a lot to be desired. The content was fine and relevant, user friendly and even visually appealing, and at the end of each section there was a quiz. I was able to breeze through the content in a day, and ace the course, but honestly don’t remember a thing.
I can’t help but wonder how different it would be if instead of a quiz, there was an avatar simulation which required application of the content. It might be, for example, a simulation of a child who has just let me know that they feel unsafe at home. This type of simulation is almost impossible to run through with a real person. What should I say? What should I do? How should I apply what I have just learnt about support and mandatory reporting? My mind would be forced to make these connections and the experience with the avatar would have a good chance of sticking in my mind if that real situation ever arose.
There are only minimal studies as to how this technology might improve engagement. In 2020, PWC did a study on soft-skill training involving basic, pre-scripted (non-AI) avatars. The study found that it improved focus, confidence, emotional connection and learning speeds by three to four times when compared to classroom and eLearning.
Build on that with the new feature that the avatars can now have a life-like conversation with a student with social cues and emotional responses and one can see that the technology opens up a whole new horizon of engagement that has yet to be explored.
At the end of an avatar simulation, one receives a transcript of the conversation. Now this may not seem groundbreaking at first glance. Isn’t it simply a combination of oral and written assessment? Let’s break down some of the factors though.
Firstly, an oral assessment is largely done by a human assessor. Scores are provided but no automatic transcript is collected (unless the student submits their talk) which would provide the exact wording. Additionally, unless questions are asked there is no clear evidence (especially in our new AI-generated content world) that the student can do anything more than read. This is a similar issue in written submissions, with increasing problems with AI usage.
However, a back-and-forth conversation with an avatar requires the student to be able to rapidly deliver the required applied knowledge - with the various interactions all captured in the transcript. Critical thinking is tested on the spot, communication is trialled, and any questions or segways are easily addressed and brought back to the core competencies.
Additionally, if the simulation was designed well, automatic feedback could be immediately provided, and the simulation tried again if needed. This development could even be connected to assessment or competency rubrics uploaded into the system.
Also consider the data collection advantages. Not only can the transcripts be linked to individual learners profiles with improvements tracked over time, but the mass collection of transcript data would provide incredible insight from the conversations. For example, currently very little data is collected on school leavers and their potential career pathways. Put 10,000 students through a career advisor simulation and you would quickly be able to analyse trends around youth career objectives.
It also may not be too long before additional measurements are added that could have impacts for assessment. There are already some advances around AI recognising facial expressions, hand movements, tone, and eye movements, all of which add a fascinating element to soft skill and social assessment (and some ethical questions - see below).
Throughout history, we all know the gold standard of education: a one-on-one interaction with a highly capable tutor (think Alexander the Great learning from Aristotle). However, since the globalisation of mass education, ensuring education is a human right rather than just an opportunity for the rich, one-to-one learning becomes increasingly out of reach. Instead we are left with a significant shortage of quality teachers and trainers alongside side over-populated courses and schools.
However, with avatar simulations, all people at once could have access to an individual, personalised, knowledgeable tutor. There is no question that this would be a poor imitation to a one-on-one with a quality trainer, but the advantages more than make up for what is lacking. These avatar tutors would be available 24-7, on any device, and potentially even have perfect long-term memory of all learning interactions the learner had previously had.
Additionally, there is a significant cost advantage. Some courses already require simulations, and they are expensive to run. Feedback from the industry is that using a staff member for example, the cost (including time, administration, preparation and so on) is at least $26 per simulation. If using a paid simulation actor, it is closer to $35. However, Avatar simulations that are accessible and consistent, are now running at around $2 to $3 per simulation. This shift in the ability to scale and save is dramatic.
Lastly, it needs to be recognised that this new type of pedagogical experience with avatars brings with it a number of ethical questions. Some of these questions we, as humans, have never really had to address before. Being in the trenches of development here are a few that have arisen:
Although just the tip of the iceberg, these are questions that can’t be ignored. There is not time to explore these questions in detail, but there is one guiding principle that should be at the centre of our use of various AI tools in education and training. That is:
Will this tool bring benefit to human relationships, or will it bring detriment?
This is a foundational starting point, and I hope to write about this aspect in more detail in a future paper.
Hopefully this introduction to the pedagogical impact on avatar simulations has been a genuine exploration of the potential significance of this new technology. The reality is that this landscape is changing, and how we choose to utilise these new technologies like avatar simulations, both functionally and ethically, will significantly shape the future of education and training.
Nick is the Managing Director of ACTS Education, which ‘Brings education and training to life’ through conversational avatar simulations in partnership with schools, universities and training institutions. He has worked in the education sphere for 15 years, most recently as the Director of Government Relations at Alphacrucis University College and prior to that as the Inaugural Director of the Lachlan Macquarie Institute. He holds a Research Masters degree from Charles Sturt University, and is passionate about developing new ways of engaging learners, as well as the foundational ethics required for effective technological innovation in education and training. For further information, visit: https://acts-education.com.au.