THE INFLUENCE OF INTEGRATING ARTIFICIAL INTELLIGENCE INTO THE SELF-LEARNING PROCESS ON STUDENTS’ ACADEMIC ACHIEVEMENTS IN KAZAKHSTAN

In the last couple of years, AI has become one of the key aspects of the development and evolution of almost every field of activity and has also become a concern for the education system, inevitably penetrating its work.

However, the help of AI-powered tools for learners in individual learning paths is getting less attention, even though a significant part of users of these tools are students. It is crucial to investigate the possible outcomes of integrating AI into self-learning, as it has the potential to improve students’ academic performances.

The aspect of AI that makes it significant for learning is the ability to adapt to learners’ personal needs. AI-powered applications can act as virtual assistants, providing guidance, pointing to students’ weaknesses, and learning gaps [3].

However, the current stage of their development may not be capable of education and have low levels of personalization. As they take the role of knowledge providers, there is often a case where students prompt chatbots outright solutions or information instead of giving direction for learning unless they are taught how to properly write prompts, making them study without actively engaging in the learning process [1].

In addition to that, most of the AI-powered applications were trained on a large amount of data, and the results offered answers can be biased or even false. Moreover, those answers are based on pre-assembled data, which jeopardizes queries on current events by incorrect or outdated answers [2].

To gather the primary data there was conducted an experiment with an experimental and a control group consisting of 6 students of 11 grade of Nazarbayev Intellectual School each. The control group was given 30 minutes for an independent study of the given topics (they were not familiar with them previously) which primarily implied research on websites. By the end of this time, they were given a test with 5 multiple-choice questions and 5 written-answer questions on the given topics for 15 minutes. The experimental group had the same task, despite that now they had to use AI for their independent study.

Despite the effort to make the research as valid as possible, the sample chosen for the study may not be representative of the whole studied population and the findings of the study may be easily outdated. It can be clearly seen in Figure 1 that the control group generally did better than the experimental one, getting an equal number of right and wrong answers (30). In contrast, the experimental group got more wrong answers (37) than right answers. According to this difference, it may be claimed that the usage of AI-powered tools for the self-learning path was not as beneficial and effective as manual research on the internet.   

Figure 1.  Number of right and wrong answers for all questions

Although the overall result of the control group was better, in most of the questions (6) the number of right answers of both groups did not have a difference of more than 1 right answer. It may mean that even though there are several gaps within the functionality of AI-powered tools when using them for educational purposes, generally those tools can provide information that might be needed for self-learning, however, are they contributing to the retention of the newly comprehended information is another question.

Despite the objectively better performance of the control group, there is a question in which the experimental group scored significantly more (difference of more than 2 right answers) than the control one. 

Figure 2 demonstrates how AI-powered tools advantaged the experimental group when answering question 7. This may lead to the claim that AI-powered tools might provide additional information for an expanded understanding of the topic.  

Figure 2. Number of right and wrong answers for question 7

In conclusion, the integration of AI may cause not active engagement in learning, jeopardize information reliability, and expose it to bias, which gives wrong ideas and weakens student’s academic performances. However, it has potential for personalization and contributes to the extended view of the studied topic, this positive influence seems to increase in the future.