Talk, tutor, and learn: how intelligent tutoring is key to boosting confidence of primary students

By Amie Liebowitz.

Maths impacts our lives on a daily basis, from child to adulthood. Yet for some, the subject may be the most difficult challenge they experience throughout their schooling.

According to the UK Government, maths is the most popular subject during A Levels with almost 25% of all students choosing to study it. However, how many more students could be studying maths post-16 if they felt comfortable or more reassured in their abilities from a younger age?

Dr Alice Hansen, a research associate at the UCL Knowledge Lab explains that,

iTalk2Learn is a 3-year EU-funded project that was created in response to PISA studies that highlighted disparities in mathematics learning for primary school students (ages 5-11) across Europe. The program aims to meet students’ needs and individual requirements that are currently sought after in the classroom.

The UCL Knowledge Lab at IOE was a part of an extensive team that brought together researchers and developers from four countries (Austria, Germany, Italy, and the UK), four universities (Birkbeck, Ruhr-University Bochum, University of Hildesheim, and UCL) and three companies (Sail Labs Technology, Testaluna, and Whizz Education).

Having a broad spectrum of talents between these researchers including artificial intelligence, digital game development and educational psychology, the iTalk2Learn project created a platform that enables the sequencing of activities designed to help students’ conceptual and procedural understanding of fractions using personalised feedback with the aid of artificial intelligence.

How iTalk2Learn works

What makes iTalk2Learn so unique in comparison to other educational technology is the overall intervention model that creates a unique experience for the participant.

• First, a student will be given an exploratory task in the newly developed learning environment Fractions Lab, where the focus is on conceptual learning, Hansen explains that “Fractions Lab is an exploratory learning environment in which children can manipulate fractions and explore various activities using fractions within that environment… [in contrast to] traditional approaches using pencil and paper that are static.”
• Meanwhile, a Student Needs Analysis component (SNA) monitors and analyses the student’s performance, their speech and their affective state (their emotions), to determine whether they were under-, over- or appropriately challenged by the task
• This information is used to determine whether the student is given another less or more challenging exploratory task; or whether they switch to the structured practice environment (e.g. Whizz Maths), where the focus is on procedural learning, to practice and consolidate what they have just explored
• The SNA then determines the next appropriate step for the individual student, and the fractions learning experience continues.

An interesting aspect of iTalk2Learn is the software’s ability to detect a student’s emotions through speech. The speech recognition system speaks to the student and encourages the student to reply. Using this technique is based on evidence that vocalising learning strategies helps the retention of the information at hand. Having collected an array of German and UK student voices, the system can respond intelligently to the student’s voices.

Building self-confidence

Jonathan Leeming, a specialist maths teacher that uses iTalk2Learn’s Fractions Lab explains that “children enjoy mathematics if they understand it. We all enjoy doing things that we are good at, and Fractions Lab enables and facilitates children to understand mathematics, understand fractions on a deep level.” 

Building self-confidence in a pupil’s maths abilities through new technologies could prepare them for secondary school and better employment opportunities. 

Proven outcomes

After being evaluated on 387 children across Germany and the United Kingdom, it was clear that the various components of the software – the combination of structure practice and exploratory tasks as well as speech recognition – were all necessary to receive such excellent feedback.

The project involved a large number of innovations. These included:

• extending the focus of typical intelligent tutoring systems by integrating exploratory learning activities, to support both structured practice and constructivist learning;
• new methods for automatic intervention selection, based on machine learning models that exploit observations about previous achievements in a systematic way;
• extending the capabilities of typical intelligent tutoring systems by integrating speech recognition and speech production, to enhance the quality and appropriateness of task selection and intelligent feedback;
• extending the capabilities of typical intelligent tutoring systems by enabling student affect detection, to enhance task selection, intelligent feedback and the overall student experience;
• an open-source platform for intelligent support systems, by integrating results from these interventions within a flexible, plugin-based architecture; and
• multi-dimension models of fractions learning and of intelligent feedback, to enhance the student experience and to promote robust learning of fractions.

With the success of the iTalk2Learn prototype, UCL Knowledge Lab’s Dr Manolis Mavrikis believes that due to its success that “it would be very interesting to see [the design] applied in different topics, both in mathematics, such as algebra, and statistics as well as other work we have done here at the UCL Knowledge Lab.”

Key facts

• Department: Culture, Communication and Media
• Centre: UCL Knowledge Lab
• Project dates: November 2012 - October 2015
• Project partners: Birkbeck, Ruhr-University Bochum, Sail Labs Technology AG, University of Hildesheim, Testaluna, UCL, Whizz Education
• Funder: European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement 318051

Related links

• iTalk2Learn
• Affective learning: improving engagement and enhancing learning with affect-aware feedback