Transformando o aprendizado na era da IA: integração de um LLM no ensino fundamental

Abstract

With the evolution of conversational Large Language Models (LLMs), such as ChatGPT, and the availability of an Application Programming Interface (API) to access this technology, there emerged a need for greater control over the functioning of this tool. Prompt engineering, a field of study focused on enhancing methods and techniques to better control the responses of these models, played a crucial role in this context. This approach involves the inclusion of additional information in the user’s initial call, such as rules and roles that the model should assume, as well as details about the user’s profile. In this work, the GPT 3.5 model was employed in the creation of a web application aimed at educating children in the third grade of elementary school, aged between 8 and 9 years. The project focused on developing a controlled prompt pattern that assumes the role of a teacher, accompanied by a set of rules to model the interaction with the LLM. This allowed the children to ask the tool questions on specific topics. The responses were tailored to the children’s inquiries, and the effectiveness of the application was evaluated by experts in the field. The conclusion of this process demonstrated the viability and efficiency of prompt engineering in personalizing the educational experience for the child audience, however, many aspects still can be improved in the future.