Integrating artificial intelligence (AI) within the education system is a responsibility shared by both educators and the broader system. When assigning tasks to students, if those tasks can be completed by AI, one must consider whether engaging in them without AI is worthwhile.

In an interview with “Vijesti,” Milutin Pavićević, the chief creative director of “Alicorn” and executive director of the Montenegrin Association for Artificial Intelligence (MAIA), shares insights on the potential incorporation of AI tools in Montenegrin and global education systems.

Pavićević emphasizes that teachers’ primary role is to equip students for the “real world,” where large language models (LLM) and AI will play an increasingly significant role as students transition out of the educational environment.

LLMs are deep learning algorithms designed to process and create human-like text. These models are trained on vast datasets that include texts from diverse sources like books, articles, websites, customer feedback, social media posts, and product reviews.

The main objective of LLMs is to comprehend and anticipate patterns in human language, enabling them to produce clear and contextually relevant text.

In late December, “Vijesti” published a questionnaire aimed at probing potential misuses of AI tools in education while identifying effective applications and benefits of these tools. A total of 130 participants responded.

While not a comprehensive study, a common theme emerged from most responses—students are utilizing AI to complete assignments for them. To shift this trend and effectively implement AI in education, a concerted effort from state institutions and the educational community is essential.

Don’t overestimate the capabilities of technology

Pavićević notes that the general public is beginning to recognize the significance of AI tools across various life sectors. He points out that other facets of AI are already in use within education, but LLMs have proven particularly advantageous today.

“In recent years, their usability has surged, demonstrating immense potential in transforming content and automating certain processes,” Pavićević states.

However, a lack of understanding regarding LLM capabilities and limitations leads to inflated expectations and misplaced trust, resulting in significant risks associated with relying on inaccurate information and the desire to exclude human input from decision-making—a perilous choice for stochastic AI.

He illustrates that while LLMs excel at addressing straightforward and previously resolved complex mathematical problems—even achieving responses comparable to doctoral-level insights in initial queries—over extended interactions (20-40 messages) and in more intricate issues, their performance may regress to below elementary school standards.

“They excel at content transformation yet struggle with finding innovative solutions,” Pavićević adds.

“They falter when dealing with numerous parameters, even if each is relatively simple; when faced with about 200 parameters to analyze, they become ineffective.”

“Resilient” tasks required

The “Vijesti” questionnaire respondents identified a prevalent misuse of AI tools—students using AI to solve tasks on their behalf.

Examples of such misuse included: “Children replicate answers;”; “They simply copy information provided by questions.”; “Creating written assignments in their native language without reading anything, demonstrating vocabulary at the level of mere statistical error.”; “Interrupting the research process, acquiring instant answers, negatively affecting critical thinking development, misrepresenting others’ opinions as their own, simply copying content for homework, cheating on tests.”

Pavićević asserts that it is the responsibility of educators and the system to foster critical thinking by presenting students with tasks that cannot be solved by AI.

“Every subject contains elements of enigma requiring a ‘drill’ task to build the routine necessary for critical thinking. It is our responsibility to craft these tasks in ways that AI cannot navigate,” emphasizes Pavićević.

He adds that professors must verify whether submitted work is plagiarized or generated by AI.

“Plagiarism detection systems can now accurately identify such occurrences. We have methods for evaluating diploma, master’s, and doctoral theses, and today’s technologies can effectively discern text produced by AI with a 98 percent accuracy rate—for each paragraph of text, they can identify whether it was written by an AI,” he explains.

Studying differently at colleges and schools

AI is already included as a subject at Montenegrin universities, such as in the undergraduate studies at the Faculty of Electrical Engineering (ETF) and the Faculty of Science and Mathematics (PMF), as well as in master’s studies at the Faculty of Information Systems and Technologies (FIST) of the University of Donja Gorica (UDG).

When asked whether AI could be a subject in elementary and secondary schools, Pavićević noted the advanced mathematical understanding required for university-level AI study but acknowledged the possibility of educating younger students about AI through existing subjects.

He questions whether this education should occur in media literacy classes or if a dedicated AI course is essential.

“Understanding its everyday relevance suggests that perhaps children and students in other curricula should learn how to use AI, aware of its risks and benefits… This constitutes a vital part of their education,” Pavićević asserts.

Responses from the “Vijesti” questionnaire regarding how to encourage students to use AI tools as aids in completing tasks rather than relying on them entirely provided various strategies:

“AI should assist in processing topics requiring internet sources, not during students’ independent work”; “Should be utilized for result verification and interactive learning”; “Show them various applications and adequately respond to any misuse. Teachers can recognize AI-assisted work”; “Transform school and educational practices in Montenegro, guiding lectures and assessments towards project-based demonstrations where students can freely utilize AI tools.”

Development and regulation need to be harmonized

On the topic of ethical AI tool use and potential regulations, Pavićević highlights two possible paths.

“One approach is legal. We acknowledge that the European Union is taking proactive steps towards ethical AI use. Meanwhile, a significant race is currently underway; Europe and America were leading this two years ago, but now America and China are progressing much faster, with China still lacking comprehensive regulations,” Pavićević elaborates.

He believes that strict regulation may hinder AI development.

“Halting advancement to facilitate ethical oversight is a major risk. However, developing ethical guidelines must proceed at a quicker pace to keep up,” Pavićević concludes.

Most “Vijesti” questionnaire respondents suggested that stricter regulations and increased educator involvement are vital to promote ethical AI use within education.

“Engage professors to review papers and ensure they identify non-original work. Without accountability, students may resort to copy-pasting without developing critical thinking skills.”; “Provide thorough training for proper usage to yield better outcomes.”; “Until current educators are adequately trained in detection tools, at least at the rectorate level, implementing automated checks for seminar/exam papers should be a priority.”; “Establish legal regulations governing AI use.”; “Incorporate AI utilization in references when composing essays and seminar papers. Encourage creative applications of AI for foundational ideas that students can enhance. Recognizing honesty regarding AI usage should be rewarded (within reasonable boundaries).”; “Limit AI applications to clarify processes rather than produce answers,” are among the suggestions put forth.

Pavićević: Don’t let the machine value humans

Milutin Pavićević stresses that artificial intelligence should not be involved in evaluating students, although it holds significant potential for personalizing education.

“To nurture critical thinking, it is essential to encourage students to articulate concepts in their own words. Teaching this using a large language model poses considerable risk, as it could inadvertently disadvantage the most creative students,” he states.

He underscores that machines should never be permitted to evaluate humans without human oversight.

“Machines tend to be susceptible to biases and might establish feedback loops resulting in discriminatory practices,” cautions Pavićević.

He adds that AI could greatly enhance personalized learning by:

“Identifying knowledge gaps within a large student body, signaling to professors where improvement is needed in their explanations. Transforming tedious topics into engaging educational activities or games that captivate students’ interest. This is where LLMs can truly shine…” the interviewee asserts.

Abuse is an everyday occurrence

The “Vijesti” questionnaire revealed that nearly all respondents, including primary and secondary school teachers, had witnessed AI tool misuse in educational contexts.

They described the constant struggle with AI misuse:

“As a teacher of vocational subjects in an electrical school, this happens daily. In mobile application programming, they regularly resort to AI tools”; “In elementary school, students write entire essays using Chat GPT for assigned topics”; “As a high school educator, many students submit papers – not authored by them – generated through AI”; “In a theoretical subject at university, approximately 70 percent of the students submitted AI-generated texts for exam essays. The professor noted this was not reflective of their abilities”; “Yes, as an educator, I encounter this issue frequently.”; “I’ve discerned from the nature of assignments that AI tools facilitated the work”; “I’ve heard numerous accounts from colleagues in education that students increasingly rely on AI for written assignments, primarily due to a lack of self-confidence in their abilities. They turn to AI for assistance.”; “Students in elementary school requested and received assistance from AI for composition. The teacher realized that the output was not their own,” reads several of the comments.

News