In the Jisc AI team we spend a lot of our time exploring different AI tools and services, assessing their potential to enhance education. With that in mind we have launched Product Notes, a new blog series, to share our knowledge and findings with all of our members. In each entry, we’ll be looking in-depth at a particular AI tool or service, exploring its potential for education. Where possible we’ll be talking to institutions currently using these tools to share their experiences too.
In this Product Note we look at Tutello, a human plus AI tutoring platform that provides students with blended support in consolidating their understanding of a given topic.
About Tutello
The Tutello platform integrates human and AI support to offer a comprehensive tutoring experience. Students start by asking an AI chatbot questions on a given topic, which responds using its extensive knowledge base. This knowledge base is created by teachers and lecturers, who can upload relevant material, including word documents, website links and pdfs. When using the AI chatbot, students can ask follow-up questions. And if further help is needed, teachers can step in, having access to the student’s interaction history with the chatbot to provide precise, personalized support.
The teacher’s first role is to set up knowledge bases for their students, which are called ‘contexts’. They can do this by uploading relevant learning resources, such as PDFs, web links, YouTube videos or lecture recordings (Tutello can be integrated with Panopto). Teachers also have control over several aspects of the chatbot. They can choose between GPT 3.5, GPT 4 and GPT 4 turbo for the underlying AI model. And they can set the temperature to a value between 0 and 1. (A temperature of 0 indicates that the chatbot will behave in a deterministic way – the same prompt will always yield the same response. A temperature of 1 means a high degree of randomness in the response the chatbot provides. In practice, you’d expect to see widely different responses each time you reissued the same prompt.) Teachers can also decide whether they want the chatbot to adopt a Socratic approach to teaching, or whether they want interactions to be more student-led.
After setting up the context, teachers’ next interact with the platform to observe student-chatbot conversations: the platform’s dashboard gives teachers complete oversight of individual student sessions. Teachers can also be alerted if students have requested further support. And they have available to analyses of student interactions. For instance, the ‘classify’ feature gives them a break down of the categories of questions students are asking.
Product insights
To explore Tutello’s features, we assumed the dual roles of teacher and student. Acting as teachers, we first set up a knowledge base on the topic AI in education. We sourced a number of reports (in PDF format) and uploaded these to the Tutello platform. We then set the AI model as GPT 3.5, the style to Socratic, and the temperature to 0.1 (almost deterministic).
As students, we asked the chatbot a range of questions, including – how can AI be used in education, what are the risks of using AI in education, and what are different stakeholders’ perspectives on the use of AI in education?
The answers were relevant, coherent, accurate, and reflective of the content in the source material we had used.
Impressed by this initial trial, we asked the team at Tutello if they had any ready-made knowledge bases we could make use of. They set us up with a ‘Maths Primer’ context, which largely focused on calculus. As a former maths teacher myself, I was keen to sink my teeth into this topic as a student. I wanted to stretch the chatbot with some tricky questions.
I asked a range of questions, ranging from basic differentiation to topics that are highly relevant to machine learning – such as Gradient Descent and Legrangian Multipliers. In all cases the chatbot responded with highly informative, accessible and accurate answers. I was very impressed.
Acting as teachers again, we could view the conversation – and we had the option of adding comments/notes to send to the students.
Key points
- The platform can give high quality answers to queries. When asking the chatbot maths questions, the quality of answers was particularly high. We foresee students with specific questions in mind getting satisfying answers from the chatbot.
- Strong knowledge bases are key. To answer a wide range of queries with precision, the chatbot needs to have access to a strong knowledge base. While it is anticipated that most institutions will already have repositories of resources, they will need to put thought into identifying and selecting the best resources to upload to the platform.
- On-demand support for students. From our wider conversations with students, it is clear that they value on-demand support. Many students have additional commitments outside of education, and all students have their own unique patterns for effective studying. As such, Tutello’s capacity to provide students with tailored and high quality responses could potentially have a really positive impact.
- Streamlining how teachers support students. Tutello’s design truly appreciates the importance of teachers. Giving students access to both AI and teacher-led support means that teachers can focus on particularly challenging questions students may have, and thus use their time more effectively.
- Tutello is easily integrated into Canvas, Moodle, D2L, Insendi, with other LMS integrations on their way based on demand.
(Most) Suitable for
- Models of learning that value learner agency and independence, and pinpointed intervention from teachers
- Teams that want to streamline the process of responding to educational queries from students
- Teams with access to a bank of high-quality learning material
Pricing
Tutello recognises that each institution, or faculty, or course instance will be slightly different. Their standard pricing models are based on student enrolments for an academic year. However, they also work with short courses.
Find out more
To find out more about Tutello, please reach out to rosie.loyd@tutello.com
