Agents of Education: Using Generative AI to build agents to support teaching and learning at different levels of education

Generative AI prompting is not one-directional but?can allow interactions with third parties.?It can be used to create agents that can support the teaching and learning of learners on your behalf. In this talk, I will discuss how through the use of simple prompting a specialised agent persona can provide teaching and learning support on behalf of parents. Similarly, it can be used by teachers to provide learning outside of the classroom. These agents can?function by interacting with learners specifically around your requests, by teaching them things parents and teachers want them to learn, and by reporting back to you about every conversation they have had with the learner. We?created an agent using the LLM OIMISA-7B.?I will elaborate on this through studies I have conducted with teachers and parents respectively.?The agent was designed to be a conversational Teddy bear that is a study buddy for children who plays with children and in-between play engages them with activities, talk, and learning, as per the prompt request of their parents. All of this without the child feeling like it is an enforced activity. It also reports back to the parent or teacher so that they can get the information needed to personally, or through the agent, further support and engage the learner at a highly personalised level. Finally, I will talk about the importance of building specialised language models, in contrast to using generalised models, to power such agents especially for vulnerable groups like young children, and end the talk by showing our framework to build such agents with a focus on discussing how it can be used for different levels of education.