YOLI, a Danish company, specialises in creating innovative and “smart” educational games for children. YOLI has been our trusted partner for several years, and together with them and other partners, we are working on a European Union grant project focusing on adaptive learning. In the first year, we developed relation data-based and the first iteration of an AI game using Chat-Gpt that will be used for testing purposes. These will be used as building blocks for our partners to develop a custom AI model.
YOLI, based in Denmark, specialises in creating innovative and “smart” educational games for children. They focus on game design that seamlessly blends digital and traditional board game elements, fostering learning through interactive play without reliance on screens.
YOLI collaborates closely with kindergartens, recognising the pivotal role of early education. The games are designed to captivate children's interest by offering various games covering many topics and themes that cater to different learning preferences, including physical activity and creative imagination. Through playing YOLI games, children learn essential skills such as teamwork, storytelling, and problem-solving required for a better start to their school life.
With a firm belief in the potential of children to shape a better tomorrow, YOLI is committed to advancing the UN Sustainable Development Goal to “ensure inclusive and equitable quality education and promote lifelong learning opportunities for all.”
To learn more about the company, click here.
YOLI consistently seeks innovative methods to captivate children's interest, striving to develop games that engage children in learning while accommodating individual learning needs. With this objective in mind, YOLI was looking for reliable and trustworthy partners to collaborate on a European Union grant project centred around innovation and adaptive learning by using state-of-the-art AI models.
In 2022, we first crossed paths with YOLI when they were searching for a software development partner to update their existing solutions and develop new features for the Yoli app. More information about the first project can be found in this case study. Our collaboration with YOLI proved highly rewarding, with both parties satisfied with the project outcomes. Over time, what began as a simple collaboration blossomed into a strong partnership.
Leveraging this solid foundation, YOLI invited us to expand our cooperation by participating in a significant European Union grant project alongside ITU (IT University of Copenhagen).
One of the primary objectives of this three-year project is to deliver innovative solutions for adaptive learning aimed at children. To achieve this, our initial emphasis was on planning and estimating the scope for the first year. We focused on pinpointing partner dependencies and agreeing on the scope requirements. The planning process, coupled with our close collaboration with the partners, proved pivotal in ensuring the successful initiation of the project.
To accomplish the main goal of the project, ITU will focus on making a custom AI model. But before they can do that, they require a test game for data collection and a robust database to store game data. Therefore, our starting point was to deliver the products that ITU requires to start with their custom model.
Firstly, we designed and developed a relational database and an API to monitor and evaluate adaptive learning experiences. The development of this data structure was guided by several key requirements: real-time processing and storage of data, retention of raw logs for ITU's analysis, and integration of search functionality to enable data analysis for YOLI.
Secondly, we successfully crafted the initial iteration of the first YOLI AI game. This involved utilising OpenAI connection services to implement the ChatGPT-3.5-turbo functionality. Completing the first game iteration enabled YOLI to test the AI game with their test group. The data generated from this testing phase will be essential for ITU in developing an adaptive learning model.
Throughout the project, our team faced several challenges. One of these was associated with the collaborative nature of the project, involving multiple partners. Ensuring alignment on the project’s vision and requirements was critical. To address this, we prioritised transparent communication and fostered close collaboration among all partners. These efforts were instrumental in laying a strong foundation for a successful first year of the project.
Another significant challenge revolved around designing a data storage solution capable of handling diverse data types and allowing data search. Firstly, we needed to ensure the various data types could be parsed. This required a thorough analysis of numerous log files and conducting detailed meetings with our partners to grasp the nature of the data fully. Secondly, we dedicated extensive time to analysing and testing various data storage solutions, striving to develop an optimal design that aligned with the project's objectives.
Lastly, recognising the potential and risks of ChatGPT, together with YOLI, we have prioritised safety when designing and developing this game. User input is restricted to predetermined content, with AI-generated output rigorously monitored through a detailed logging system. Based on these insights, the input was fine-tuned to the model to ensure the accuracy and appropriateness of the content from ChatGPT. Moreover, a kindergarten teacher must initiate and oversee gameplay on a mobile device, remaining present throughout. This year, we are also developing tools for direct teacher feedback on content creation and children's engagement levels. Together with Yoli, we've taken proactive steps to ensure a safe and enriching experience, reflecting our commitment to responsible AI development.
We successfully completed the first year of the project, delivering the database and the initial iteration of the adaptive game for testing alongside other essential components. These achievements lay a solid foundation for the upcoming year, during which we will further collaborate with YOLI and ITU on adaptive educational games.
.jpg)