- Статьи
- Society
- Scoring the slot machine: how students mastered the course with the help of an AI professor
Scoring the slot machine: how students mastered the course with the help of an AI professor
For the first time, Russian students completed a training course entirely under the guidance of artificial intelligence instead of a regular teacher. The course "Biology of Human Behavior" was taught by an AI assistant, it was developed by the AI Laboratory of the Skolkovo School of Management and specialists from Tyumen State University. As the results showed, the students studied more subjects than those who studied with a living professor, learned to use scientific terms more accurately and reproduced the key ideas of the course more fully. About how he taught AI and in what disciplines he will be able to replace traditional lecturers, see the Izvestia material.
How did the experiment go?
An educational experiment with the introduction of artificial intelligence was conducted in Russia — the course "Biology of Human Behavior" for students of the joint bachelor's degree at the Skolkovo School of Management and MIPT was conducted by Robert, an AI assistant.
"Despite the active penetration of AI into the educational process, so far its use has been limited to auxiliary functions," the report on the research says. "However, a full-fledged AI teacher who assumes the role of a lecturer and becomes the only source of knowledge is a fundamentally new stage."
The traditional and innovative courses were held in an intensive format, the duration of the training was six seven-hour academic days, there were 42 first-year students in both groups. None of whom had previously passed the Unified State Exam in biology and had only situational experience working with AI.
In the traditional format, the course included lectures given by the teacher, as well as face-to-face seminars. During the course with Robert, students interacted with AI through various formats of individual and group work. Robert was assisted by a human mediator who directed the process, but did not have expert knowledge of the discipline being studied and, accordingly, could not insure the AI.
The digital professor was designed as a classic chatbot with which you can communicate through questions.
"The AI teacher is able to respond from different theoretical frameworks," Alexander Didenko, head of the AI Laboratory at the Skolkovo School of Management, told Izvestia. — In fact, there are several "personalities" in it who conduct an internal dialogue, commenting on each other as a neuroscientist, an animal behavior specialist, an endocrinologist, a geneticist, and so on. This is an interesting finding that will be useful to us wherever we need to give a complex interdisciplinary answer or comment on a situation from different points of view. An ordinary bot will not cope well with such a task, it will simplify the task for itself.
Sofya Matveeva, a mediator at the Center for Educational Research Based on Artificial Intelligence Technologies (based at the TSU School of Advanced Studies), said that the central element of the interaction was "active questioning and productive distrust of AI."
— This allowed students to constantly clarify details from the AI person and complicate their thoughts while working on tasks from the mediator. The assignment formats included both free discussions based on AI personas, as well as structured formats such as debates, case analysis, close-reading seminars, and plenary sessions, she said. — In all these and other formats, the AI person acted as an intellectual partner: she helped to reveal the implicit details of biological processes, clarify concepts, form arguments and deepen the analysis of the issues discussed.
What results did the experiment show?
According to Alexander Didenko, the organizers recognized the result of the experiment as successful, and the participating universities will use this format in the future.
"The results turned out to be promising: students who studied with the help of AI showed better results than their predecessors who studied the subject using traditional methods," the report says. "Moreover, the study allowed us to determine the optimal language models that ensure the best balance between the quality of content and the cost of its generation."
At the end of the course, personal interviews were conducted in both groups, during which students were asked to answer questions about acquired knowledge, a general understanding of human behavior biology as a scientific discipline, and awareness of the influence of biological mechanisms on behavior.
The answers of the students who were taught by AI related to a larger number of topics from the course material than the answers of those who studied with a regular teacher. In addition, the accuracy of using scientific terminology was higher in the AI group, and students reproduced the key ideas of the course more fully and in detail.
"We have seen that AI is transforming the very essence and logic of learning: the era in which cramming and strict adherence to instructions were valued is coming to an end," said Alexander Didenko. — It is no longer important what a person knows, but how he thinks.: is he able to interpret the data, build causal relationships, adapt, formulate solutions in conditions of uncertainty?
The participants of the experiment interviewed by Izvestia confirmed that they had a need to look for and check something, which motivated them to reread lectures and books on their own.
"There is always doubt whether the chatbot is right, whether you can trust it," said Ksenia Novgorodtseva, a first—year student at the Skolkovo Moscow School of Management. — It makes you learn on your own. And in general, there was probably also an important factor — you need to enter your own training track, go deeper, and ask questions. There is a need to verify the knowledge that we have already received and draw our own conclusions based on it. You're being taught to learn.
Another student, Ksenia Shirokova, also believes that the learning format with an AI teacher improves skills that will then be useful during independent study.
— This is also the ability to ask questions correctly, specifically and accurately. The wording directly depends on which answer your AI teacher will give you," she said. — The ability to critically analyze information and understand where I'm deceiving you is improving: despite the fact that, according to the feedback from the creators, the neural network was trained on reliable scientific materials, no one cancels technical problems, so we learned how to work with information from the bot, checking it additionally in other sources.
Ksenia Shirokova said that part of our educational process in the classroom was based on working in small teams.
"We made presentation projects on the topic, delved into it on our own and then talked in front of each other," she described the process. — When you do not passively perceive information during a lecture, but search for it yourself, process it, formalize it and tell it, information is remembered effectively.
But such a format can be intimidating, Varvara Murashova from the School of Advanced Studies admitted — she took part in testing an AI teacher.
"Maybe it's just because people may not fully understand how it all works," she said. — The uniqueness of such a professor lies in the fact that he can accommodate much more than a typical teacher. And here a paradox arises: in fact, a machine is much weaker than a human, but in reality no professor can contain such a volume of information.
In which disciplines will AI help
Artificial intelligence is most effective in mathematics and technical sciences, information technology and software engineering, according to Ilya Krasilnikov, Vice Rector for Digital Development at Moscow Polytechnic University.
"Here, AI helps analyze large amounts of data, automate task verification, identify knowledge gaps, build individual learning trajectories, and predict student success," he said. — And in language disciplines, AI assistants can help develop reading and writing skills. AI also works well in social sciences and sociology, where it helps analyze large amounts of textual and statistical information, identify trends and patterns.
Ilya Krasilnikov is confident that a complete transition to AI learning is possible only in strictly formalized areas where tasks can be unambiguously algorithmized.
"For example, these are basic programming courses or separate sections of mathematics," he said. — However, even in these areas, AI often acts as an assistant, rather than a full-fledged substitute for a teacher. It automates routine processes, but it is not able to completely replace the role of a mentor, especially in developing soft skills and motivating students.
According to him, AI is widely used in universities — these are chatbots for operational consultations of students and applicants, virtual assistants and tutors who help in adaptive learning and support in the study of complex topics, predictive analytics systems - analyze academic performance, identify the risks of expulsion, recommend individual development trajectories. AI is also used to automate knowledge testing in tests, essays, and image analysis.
—AI scaling is a modern trend today that makes education more personalized, accessible and effective, but a neural network will not replace a teacher, but will become his assistant," the expert said. — The teacher will focus on strategic, creative and communicative tasks, and the AI will take over the routine processes.
Konstantin Bakulev, Deputy head of the Department of Engineering Cybernetics at NUST MISIS, believes that the expansion of the use of AI in universities should lead to an increase in both teaching staff and IT specialists, which will require increased funding for the implementation of neural networks in the educational process.
"Using AI in training is a serious complex process that usually requires more man—hours and, consequently, finances to ensure the reliability of the system and keep it up to date," he said.
If students learn how to use modern tools to optimize the educational process, they will free up time to solve problems that require greater involvement, said Leonid Sanochkin, technical leader of the NLP team at MWS AI. For example, for students of areas related to machine learning, it is quite acceptable for part of the code to be prepared using AI, if they have already mastered this skill.
Also, he added, there are areas where the use of neural networks is still limited. For example, when studying rare or low—resource languages, modern language models are not sufficiently trained on such data.
The introduction of artificial intelligence into education requires a balance," Anastasia Belozerova, Deputy Director of Research at VisionLabs, is convinced. — AI should not replace teachers, but it can become their effective tool. On the one hand, it opens up opportunities for personalizing learning, increasing student engagement and cognitive support. On the other hand, there are still risks: from reducing the development of critical thinking and creativity to issues of ethics, confidentiality and authorship of works.
She also noted that in humanitarian fields that require creative thinking and emotional engagement, the role of AI is limited — it is not yet able to replace live discussion or in-depth analysis.
—A biology course with a lot of structured data is well compatible with AI, but the interpretation of literary texts is not," the expert said.
In schools, artificial intelligence, she recalled, helps teachers create educational materials, update programs, analyze student success and form individual learning trajectories.
The use of artificial intelligence in higher education is a logical step in the development of educational technologies, says Sergey Karaev, head of the Department of Information Technology at the New Look School.
"And in schools, AI is more often used as an assistant or tutor — to train skills, help with homework preparation and study complex topics," he said. — However, the most important task here is to teach students to work with AI consciously, understanding its limitations. Critical thinking and digital literacy are becoming the foundation of computer science courses today and, in fact, the foundation of modern education. A striking example is the use of AI in web development training.
For example, neuromodels can generate code or an interface based on a description, but in practice, students face difficulties: it is not always clear how the proposed solution works, why an error occurs, or how to make changes correctly. This leads to an awareness of how important it is to master basic skills on your own.
Переведено сервисом «Яндекс Переводчик»
