How can AI help build a hybrid learning solution? Find out with Dr Pauldy Otermans and Dev Aditya.
In the ever-evolving landscape of education, we are continuously at a crossroads. Traditional classroom learning methods are becoming increasingly inadequate for supporting students’ entire learning journeys and preparing them for modern-world challenges. Furthermore, we also know the world also has a severe and growing shortage of teachers: currently, 44 million. 1 So, to truly revolutionise education and meet its current needs, we must harness the power of technology.
Despite the improved accessibility provided by MOOC (Massive Open Online Course) platforms and online lessons, learning outside of the classroom remains largely in one direction. Typically, students either read or consume passive audio-visual content like recordings and videos, seeking clarifications independently and applying their learning without guidance. Moreover, teachers often lack insights into students’ learning progress, comprehension and challenges outside the classroom.
Introducing AI tutors
Recognising these challenges to transform the educational experience is key and this is where AI can really support educators. An example of this is the use of AI tutors. An AI tutor can be designed to act as a one-to-one tutor for students during their outside-of-classroom learning. Imagine having a personal tutor available 24/7, capable of breaking down complex content, teaching, motivating and testing you. AI tutors can do exactly that, bridging the gap between traditional and digital teaching.
The framework of this innovation is grounded in the need for a more interactive and responsive educational environment outside classroom hours. Traditional methods often fail to address individual student needs, leaving many to struggle in silence when faced with significant doubts. AI changes this dynamic by providing immediate, personalised teaching, feedback and support.
From a user-interface perspective, learners interact with a human-like avatar powered by a Generative AI Large Language Model (LLM). This is done to give the AI a more human-like appearance as compared to chatbots, that act like learning assistants and co-pilots. While teaching, the AI tests learners and provides AI-generated feedback to improve their answers. If the AI detects that a student hasn’t fully grasped a topic, it re-teaches the material in a simpler, more accessible way. Students can ask any number of questions, allowing for instant doubt clearing and engagement in scenario-based activities which are personalised to each student. Even for learners with varying language proficiencies, the AI can adjust the language level or switch languages entirely to facilitate better understanding.
Outcomes
In one such case, the results were found to be extremely positive. Researchers have seen over 60% completion rates for programmes with a minimum of nine lessons, compared to completion rates of 7-14% for non-compulsory learning on traditional MOOC platforms. Additionally, students ask the AI teacher an average of four questions per hour of learning, indicating a high level of engagement. 2 It was clear that these continuous interactions ensure that students remain motivated and clear in their understanding.
However, everything cannot be AI-driven, and we personally believe that having the human in the loop and playing the central role remains important today. In this spirit, the AI can also report back to human teachers and academics on student progress and challenges. For instance, they can now also see what questions students ask the AI during their learning, helping them to understand where students may need some further support in class.
This feedback loop enables educators to provide better personalised support during contact hours and classroom sessions, seamlessly integrating AI and human teaching to enhance each other’s strengths.
Benefits versus challenges
Benefits
Personalised learning. The strongest benefit of using an AI tutor is personalised learning tools. The AI can tailor lessons to fit each student’s needs, learning pace and learning style. This personalisation can enhance learner retention and accommodates to diverse learning needs.
Scalability and availability. In addition, AI teachers are very scalable and can address vast numbers of students simultaneously, thereby making quality education accessible (even in remote areas). In addition, the AI teacher is available 24/7. Linking back to personalisation, the AI can provide instant feedback on activities, thereby helping students understand their mistakes and supporting them with areas they struggle with in the moment.
Data. Another important area is data. Through the use of AI, individual student data can now be tracked and analysed in real time. This can offer insights that can help teachers as well as parents to support learning more effectively.
Admin. Finally, the AI can grade activities, prepare lesson plans and many other administrative tasks, which can free up lots of time for human teachers to focus on more meaningful interactions with each student.
Challenges
Empathy. AI is not yet very good at truly empathising with learners, which is a crucial component for creating a positive learning environment.
Bias. Another element is bias in AI systems. The AI is only as good as the data it has been trained on. If the data is very biased, most likely the AI’s output will be as well.
Access: Implementing AI systems requires access to technology, which may not be accessible to all students. However, with the rise of mobile phone usage and the spread of internet, this challenge will reduce in the next few years.
Privacy. Finally, an important element is privacy and data security. AI systems collect extensive data on student performance and student interactions with the AI. This can raise questions about data security, data privacy and the ethical use of personal information. With the move towards Edge AI, these concerns are less problematic, as with running the LLM on your device, personal data no longer has to leave the device the user is using.
Recommendations
To create a successful AI-driven educational system, you can start by focusing these three key areas:
- Choose robust open-source models: Use strong, bias-checked models like Mistral 7B or Meta’s Llama 2 and 3. These provide essential conversational skills efficiently.
- Enhance with proprietary data: Tailor your AI with specific datasets, including student interactions and feedback. Fine tune your model with this data for maximum effectiveness.
- Implement de-biasing techniques: Finally, ensure fairness and accuracy by applying standard de-biasing methods during model training. This is a critical ethical step.
Conclusion
So, we conclude with this: embracing AI technology is not just an option but a necessity. Tools like AI tutors represent a significant step forward in creating a more interactive, responsive and personalised learning environment. By combining the strengths of AI and human teaching, we can better prepare students for the challenges of the modern world and ensure a brighter future for education.
References
- UNESCO (2024) ‘Global report on teachers: What you need to know’. Available at: https://www.unesco.org/en/articles/global-report-teachers-what-you-need-know (Accessed: 11 November 2024)
- Aditya. D, Silvestri. K and Otermans. P (2024) ‘Can AI teach me employability? A multi-national study in three countries’. Available at: https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2024.1461158/full (Accessed 17 December 2024)
Authors
Pauldy is a female tech leader in the UK. She is a neuroscientist and psychologist by academic background and a female leader of AI technology. Pauldy helped build the first digital human teachers powered by AI in 2021 called OIAI. She is also active in the research community.
Dev is an expert in Generative AI from London, UK. He led the team that built the first Digital Human AI teacher (OIAI) with the mission to upskill 750 million learners globally by 2030. He is also an active researcher in HCI and AI who publishes in the space.
