Abstract:

Significant advances have been evidenced in the field of Artificial Intelligence in recent decades. Such advances are related to an evolution of several algorithms in applications of machine learning, pattern recognition, natural language processing, automation and robotics. Several software began to integrate AI algorithms in their codes, covering a multitude of applications and providing numerous solutions. These softwares can store large volumes of data. With advances in AI in recent months, specifically with the emergence of ChatGPT, new challenges have been imposed on teaching environments. An important factor is the integration of AI software interfaces with humans, becoming increasingly friendly. From various software, especially ChatGPT, the user began to interact with the machine in a more friendly way, allowing the AI to answer a multitude of questions. However, to what extent is there effective learning of the contents proposed in a teaching plan given the easy access of students to AI tools such as ChatGPT? How will ethics in an evaluative environment deal with this issue? Thus, this study intends to analyze and discuss the impacts of AI, specifically ChatGPT, in the classes of three Engineering courses at a federal university in Brazil. The great challenge is to measure student learning in an entire academic semester, after the emergence and use of ChatGPT, notably from November 2022. To achieve the objectives, a web page was developed to apply and evaluate the proposed scenario. Complementarily, a case study divided into two parts was carried out: to analyze and measure learning in a Computer Programming discipline; and analyze and measure learning in two theoretical disciplines. 56 activities were carried out, distributed among the disciplines of Programming Language, Computer Organization and Computer Organization and Architecture, totaling 881 responses in evaluative questionnaires. At the beginning of the school semester, a survey was carried out with the students about the evaluation methods of the three disciplines, with the options to choose from: formative evaluation, activities; summative assessment, tests; sum of the two methods. Most students, 84%, chose formative assessment as the best assessment method. This choice occurred due to the integration of the contents taught with the applications of these contents. All classes were divided into two parts: application of the concepts listed in the teaching plan and formative assessment. In addition to the formative assessment, at the end of the semester a summative assessment was adopted. The results showed that, in Engineering courses with a strong technological bias, it is necessary to combine forms of teaching using AI with the contents required in the teaching plans. In some activities, students used ChatGPT to consult the answers. The use of ChatGPT was accepted for research purposes, to analyze the impacts of its use on teaching and learning. The contribution of the work is to discuss how to combine the new AI tools with teaching and at the same time have effective learning. Therefore, it appears that new teaching strategies become necessary for the assimilation of contents without AI tools causing dependence on teaching and learning. It is necessary to discuss ethical issues of AI tools in education as these tools only tend to increase.

Keywords:

Artificial intelligence, ChatGPT, University education, Engineering Education.