Interview with a digital tutor for schoolchildren: "Our goal is to develop thinking and speech" / ITech content

Contents:

• Why "Digital Tutor" and not "AI Teacher"
• How the tool will develop
• Why you shouldn't be afraid of communicating with a digital tutor

Head of Digital Products at the Samoletum educational ecosystem, which is part of the Samolet Group of Companies, is also a leading expert in the field of digitalization of education at University 2035. Specializing in the development and implementation of innovative solutions, he contributes to the transformation of the educational process, making it more accessible and effective. Expertise in digitalization of education enables the implementation of modern technologies and approaches, which in turn supports the strategic goals of educational institutions and helps them adapt to changes in the digital world.

At Samolet Group, Andrey Komissarov is developing an innovative digital tool for educational institutions based on artificial intelligence technologies. This product, called "Digital Tutor," is aimed at simplifying the homework process for schoolchildren without interfering with the educational process. We discussed this tool with Andrey, how it will benefit students, and what he expects from parents regarding the use of the AI ​​tutor. We also discussed the appropriateness of children interacting with artificial intelligence.

Why a "Digital Tutor" and not an "AI Teacher"

The AI ​​tutor project arose in response to the growing need for personalized learning. Modern educational approaches require adaptation to the unique needs of each student. The AI ​​tutor combines modern artificial intelligence technologies and machine learning algorithms, allowing it to analyze the learning style, preferences, and knowledge level of each user. This makes the learning process more effective and accessible. The idea behind creating such a tool was to provide each student with the opportunity to receive support and recommendations tailored to their individual needs. As a result, a project was developed that aims to improve the quality of education and increase student engagement.

In July of this year, I began working in the education vertical of the Samolet group of companies. This is a new area for the developer, and one that is of great importance, as Samolet not only builds residential complexes but also schools. One of the most pressing problems in modern school education is overcrowded classrooms. Under these conditions, students are deprived of an individualized approach to learning, leading to difficulties with homework, accumulation of learning gaps, and the need to turn to tutors. This, of course, impacts family budgets.

We proposed developing a digital solution to address these issues using an ensemble of algorithms based on generative neural networks. This ensemble was initially created in my lab to support corporate training.

We chose to develop a digital tutor because it offers a personalized approach to learning that is difficult to implement in a traditional system. The digital tutor allows students to receive personalized recommendations and adapted learning materials, which contributes to a deeper understanding of the subject. Unlike a teacher's assistant, who can assist with classroom management or administrative tasks, a tutor focuses on the needs of each student. This approach significantly improves the quality of learning, increasing student motivation and engagement. A digital tutor is becoming an indispensable tool in the educational process, providing access to knowledge anytime, anywhere. Integrating generative algorithms based on artificial intelligence directly into the educational process may be inappropriate and may even negatively impact learning. Interfering with the teacher's work should not be an option. They should be able to carry out their duties without fear of losing their position. We strive to make the teacher's job easier by providing detailed data on the student's interaction with the AI ​​tutor during homework. This allows teachers to better understand the learning process and adapt their methods to the needs of each child.

Benjamin Bloom's Two Sigma Problem stems from the significant gap between the educational outcomes achieved with tutoring and those of students in large classes in standard schools. This gap is two standard deviations, meaning that children in the school system perform twice as poorly. By high school, many students have accumulated pedagogical problems and errors in understanding the material. How can this gap between individualized and mass approaches to education be bridged? We are focusing our efforts on solving this important problem.

Artificial intelligence in the role The tutor will be used by various categories of users, including students, professionals, and educational institutions. This tutor will be able to provide personalized learning materials, adapting to the knowledge level and needs of each user. Interaction with the AI ​​tutor will include an analysis of current knowledge, setting learning goals, and developing an individualized program. Users will be able to receive feedback, take tests, and monitor their progress in real time. This approach will make learning more effective and accessible, which is especially important in the modern educational environment.

In our pilot project, we offer tutoring support in social studies for students in grades 8 and 9. The primary interaction will be between teenagers and teachers. Teachers will be able to register with the AI ​​tutor, available as a website and chatbot in messengers. As part of this project, teachers provide students with homework assignments that can vary in difficulty. We offer standard assignments, advanced assignments for the Unified State Exam (USE), and challenging assignments for Olympiad participants. Teachers can choose the level of assignments for each student and upload their own additional materials, combining them with those provided by the AI ​​tutor. Students will receive homework through the new tool and can complete it directly within the app. This will ensure convenience and improve learning efficiency, as all assignments will be collected in one place. Working with the digital tutor will be possible in two modes. The first mode is a conversation, which involves interacting with the AI ​​tutor on a topic assigned by the teacher. The program poses problematic questions related to the topic, which encourages active thinking in the student. The main goal is not simply to retell the material, but to use guiding questions to help the student independently comprehend and understand the material being studied. This approach develops critical thinking and promotes deep learning.

The conversation mode will be available in both audio and text formats. This will allow users to choose the most convenient communication method depending on the situation and preferences.

The AI ​​tutor operates in audio format, which encourages children to speak their responses rather than write them down. I have been studying the use of AI algorithms in educational environments for a long time, and my lab's research has shown that visual elements in such tools are not in demand—users are not interested in "digital characters." Furthermore, children often show no desire to write or print lengthy texts, and oral communication skills are underdeveloped at school. Our main goal is to stimulate the development of thinking and oral communication in children.

The first mode of working with the tutor is conversation. What is the second mode?

The second element of the system is homework. The AI ​​tutor's architecture includes an "Examiner" module, which allows students to submit completed assignments for review. In response, students receive formative assessment and feedback. If the assignment was written, the student reads the written text, and the digital tutor asks questions for in-depth analysis and reflection. This process promotes critical thinking and improves learning.

Will the dialogue continue at this stage?

The AI ​​tutor adapts to the child's thinking and speech, asking additional questions and taking into account the answers received. It is based on an ensemble of neural networks, primarily Llama 2, trained on a specialized dataset. A key part of its operation is advanced prompt engineering, which allows the tutor to accurately determine which questions it can and cannot answer. For example, it refrains from providing answers to homework questions so that the student can complete them independently. This approach promotes critical thinking and independence in students, making the AI ​​tutor an effective tool in the educational process.

The tool provides a grade ranging from two to five, accompanied by detailed formative feedback. If a child is dissatisfied with the result, they can retake the homework up to seven times. Thanks to the adaptability and flexibility of large language models, the artificial intelligence doesn't repeat the same questions during retakes, but rather offers them in a rephrased form. This allows the child to better understand the material and improve their knowledge.

How will the tool develop?

Are there any similar tools to your AI tutor? If so, is there any evidence from experience that confirms that learning with a digital tutor produces results comparable to working with a human tutor?

Our digital tutor has much in common with the Khanmigo chatbot, developed for Khan Academy. Khan Academy is a leading online platform for school education. However, a full evaluation of the effectiveness of the Khanmigo AI tutor will require more time and analysis. It is important to note that such technologies can significantly change the approach to learning, providing students with personalized help and support.

An interesting application of artificial intelligence is the game-based learning simulators developed with support from DARPA (the Defense Advanced Research Projects Agency). These simulators are effective in areas such as mathematics, physics, and chemistry, where tasks can be presented as puzzles. They operate in a flipped classroom format: first, the student attempts to solve the problem independently, and then the GPT-based system provides an explanation of theoretical aspects that may have been unclear. It is important to note that, as with Khanmigo, this process involves accumulating a research base and collecting data on educational experiences. It is premature to draw definitive conclusions at this time.

Are there any conclusions yet?

The Chinese tool Squirrel AI has demonstrated significant results in preparing students for the Gao Kao, a national exam similar to the Russian Unified State Exam. The system begins teaching second-grade students, providing a standardized educational path that effectively replaces the traditional school curriculum using neural network technologies. Despite the lack of an individualized approach, this method delivers significant improvements in educational outcomes. Students using Squirrel AI achieve 30-40% higher grades than those who did not use this tool. This is especially significant given the highly competitive nature of the exam and its importance for the future of Chinese schoolchildren. Your AI tutor does offer limited freedom of action for the child, as the teacher selects the difficulty level of the task from three possible options. This raises questions about how individualized learning can be and how much autonomy the student receives. Determining the difficulty level of a task should take into account not only educational standards but also the unique abilities and interests of each child to ensure learning is as effective and engaging as possible. Our AI tutor effectively interacts with the thinking and personality of each child. It analyzes how each student copes with tasks and provides the necessary support. This not only improves the learning process but also tailors learning to the unique needs of each student, fostering a deeper understanding of the material and skill development.

We also plan to introduce gamification elements into our digital tutor. Specifically, we will implement a progress bar similar to those used in video games. As the bar fills to a certain level, users will be able to earn various prizes. These could include merchandise, trips, unique opportunities, or even cash scholarships. This approach will not only increase user motivation but also make the learning process more engaging and interactive.

Gamification of the educational process can be applied to assignments of varying difficulty. If the teacher assigns a standard homework assignment, a student wishing to demonstrate their abilities can request a more challenging task from the AI ​​tutor. This will allow teachers to evaluate not only the quality of the completed assignment but also the student's initiative, which will contribute to the development of their motivation and interest in learning.

We plan to introduce the AI ​​tutor into schools through a phased integration process. First, we will train teachers so they can effectively use technology in the educational process. Then, we will develop interactive platforms that will allow students to interact with the AI ​​tutor in real time, providing personalized recommendations and assistance in studying subjects. An important step will be testing the system in several educational institutions to evaluate its effectiveness and make any necessary adjustments. We will also collect feedback from teachers and students to further improve the AI ​​tutor's functionality. Our goal is to create a safe and productive educational environment in which technology will contribute to the improvement of learning and student development.

The pilot project will be presented at the AI ​​Journey conference at the end of November. The students in this project will be students from schools built with the participation of the Samolet company. These include School No. 51 in Lyubertsy, the Start School in Vidnoye, the Lyubertsy Engineering and Technology Lyceum, the Vershina School, and the Samolet Private School in the Sputnik residential complex. This project aims to introduce modern technologies into the educational process and improve the quality of instruction.

In the future, we plan to expand the use of the AI ​​tutor to all academic subjects, including primary school. In the first phase, we will focus on a pilot project that will identify potential shortcomings, collect UX tests, and analyze user behavior. This will help evaluate the system's effectiveness and identify vulnerabilities. The pilot period is expected to last approximately six months, during which we will openly share progress reports.

Neural networks will be trained in parallel across various subjects. To effectively train the Llama model, it requires extensive and diverse datasets. This allows it to develop its capabilities and improve the quality of its information processing.

Sizes can vary depending on the specific object or category. For example, in the case of animals, sizes can range from a few centimeters to several meters. Buildings, for example, can range in height and area from small structures to skyscrapers. Plant sizes in the natural world also vary, from dwarf species to giant trees. For a more accurate answer, it is important to clarify the specific object or category in question.

The collected dataset on social studies comprises 1.8 gigabytes of high-quality text, carefully divided into key topics. All materials are current, no older than 2020, and include a variety of content types: textbooks, creative assignments, articles, and regulations. Six experienced methodologists collected this dataset, followed by a manual prompt engineering stage that cannot be automated. This approach ensures the high quality and relevance of the collected information, which is particularly important for educational purposes.

Social studies was chosen as the subject for the pilot project because it allows for the effective integration of digital tools into the educational process. Although digital technologies are often associated with routine skills such as mathematics and foreign languages, social studies offers a unique opportunity to develop critical thinking and analyze social processes. The use of digital resources in social studies contributes to a deeper understanding of sociocultural phenomena and enhances the learning process. In this way, we can not only improve learning but also prepare students for contemporary societal challenges.

As a researcher, I believe it is important to conduct a feasibility study before beginning serious work. In this context, a feasibility study involves examining all possible vulnerabilities of large language models. One of the key vulnerabilities is the so-called "dementia" of artificial intelligence, which manifests itself in a decrease in the quality of the generated material. This can create the illusion that the neural network is losing its intellectual capacity.

Reasons for problems in Artificial intelligence can have different applications. One of them is training on datasets created by other neural networks instead of people. However, the most common cause of the so-called "dementia" of AI is a phenomenon known as "dataset leakage." This occurs when data from different topics is mixed in a single dataset. Social science, as a multifaceted field, includes sociology, philosophy, religious studies, economics, and other disciplines. That is why it can be used to analyze the feasibility of technologies in various fields and test how they cope with the diversity of topics and data.

Why you shouldn't be afraid of communicating with a digital tutor

You noted that teachers are not interested in integrating artificial intelligence-based tools into the educational process. Do you think parents will accept a tool that helps children with homework with a positive attitude?

We are conducting an experiment to study attitudes toward an AI tutor. This complex system requires preliminary preparation, which includes collecting data on the educational experience and digital footprint. In our schools, we begin by analyzing each student's digital footprint to understand their needs and adapt the educational process using technology.

We do this because we believe the primary goal of school education is to offer children the opportunity to find socially useful activities that bring them happiness. This is achieved through the harmonious realization of their personal potential. School education should not only teach but also help children discover their talents, which subsequently contributes to their successful integration into society and professional life.

It is necessary to collect information about how the student thinks and behaves, as well as their interests. This will allow us to more accurately describe and visualize their potential. Analyzing this data will help identify strengths and areas for development, which will contribute to more effective learning and personal growth.

Are you collecting additional data on children's learning?

Our experiment focuses on the use of an AI tutor in the educational process. In the future, this tool will become an important source of data, but currently, schools record insufficient information in digital format. To address this issue, we developed a Telegram chatbot designed to collect data on formative, role-based, and criterion-based assessments that teachers provide to their students. The chatbot is called "Teacher's Navigator" because it will analyze the collected data and provide teachers with additional recommendations for each student. This will improve the quality of education and individualize the approach to learning.

The experiment in collecting digital footprint data is taking place in three schools and is carried out only with the consent of students' parents. As part of this project, we are analyzing criterion-based assessment and formative feedback provided by teachers in audio messages. Artificial intelligence processes every message, allowing it to generate recommendations for further learning and student development.

Many parents are unaware of exactly what data is being collected and for what purpose. It's important to understand that until a clear career path is formed that aligns with your child's unique potential, it's difficult to truly appreciate the value of the information being collected. That's why I share my projects at various conferences to explain the significance of this data and its role in children's development.

Students' repeated interactions with artificial intelligence can have various psychological consequences. First, students may develop a dependence on AI as a source of information and support, which can reduce their ability to think independently and critically. Second, constant interaction with AI can impact social skills, as students may prefer interacting with machines over interacting with peers. This can lead to isolation and a lack of emotional connections. Furthermore, students may begin to perceive AI as a more reliable source of knowledge, which can lead to a distorted view of reality. It is important to consider these aspects and find a balance between the use of AI and traditional teaching methods to minimize negative consequences and develop students' essential communication and critical thinking skills.

Research shows that a lack of communication in children can lead to serious psychological consequences. Experts such as neuroscientists, speech therapists, and educators note that limited interaction negatively impacts emotional and cognitive development. Children who do not have sufficient communication may face difficulties in developing social skills, as well as in the formation of language skills. Therefore, it is important to provide children with opportunities for communication and interaction, which will contribute to their harmonious development.

Humans are social beings, and this statement is based on scientific facts. Our brain primarily processes social interactions, including other people's appearance, their body language, and facial expressions. Regardless of our wishes, a significant portion of the brain's computing resources is focused on analyzing these social aspects. Therefore, individualized learning is often less effective, as it ignores the importance of interaction and collaborative learning. It is important to recognize that social connections and shared experiences play a key role in the learning process and personal development.

Learning in a group setting is an important aspect of children's development. There are interesting educational systems, such as the Jiddu Krishnamurti Foundation Schools in India, that eliminate the traditional classroom system. In these schools, learning occurs spontaneously and in groups, which promotes deeper learning. Graduates from these institutions become highly qualified specialists. Similar conclusions were reached in our work at the "School of Digital Transformation" at University 2035. Research has shown that adults studying online in a group setting understand the material 15% better than those studying individually. Thus, group learning promotes more effective knowledge acquisition and skill development.

Communication with artificial intelligence should not be seen as a substitute for live human interaction. Artificial intelligence is a powerful tool that can significantly simplify the process of learning and providing information. However, interaction with AI cannot replace sincere and emotional human dialogue. It's important to remember that live interactions enrich our experiences, fostering deep connections and developing emotional intelligence that even the most sophisticated algorithm cannot offer.

The use of AI tutors in group learning processes significantly improves the quality of the educational experience. AI tutors are able to adapt to the individual needs of each student, allowing them to interact more effectively in a group. They analyze the knowledge level and progress of each participant, offering personalized recommendations and assignments. This promotes a deeper understanding of the material and increases student motivation. Furthermore, AI tutors can facilitate interactions between group members, helping them share knowledge and experiences. Thus, integrating AI tutors into group lessons opens new horizons for collective learning and the joint achievement of learning goals.

We plan to use AI tutors not only in lessons but also for homework assignments. When using this tool, the teacher receives a dashboard that provides detailed information on how students completed their homework. This allows us to significantly reduce the time spent on discussing completed assignments in class, eliminating the need for the teacher to spend 20-25 minutes on a survey about homework completion.

Teachers will be able to effectively use their freed-up time to organize group work. We will provide them with teaching materials covering various approaches, such as station rotation, project-based learning, game-based learning, STEAM, and case studies. These methods will allow students to integrate the skills they have developed with the AI ​​tutor at home and apply them in group lessons in the classroom. This will create a favorable atmosphere for interaction and the development of critical thinking among students, which in turn will improve the quality of the educational process.

Some believe that in the near future, traditional in-person education will be available only to a limited number of people, while online learning will be a mass and lower-quality format. Do you agree with this point of view?

I agree with this forecast, but with some caveats. Firstly, we should not perceive this future as gloomy. Currently, there are numerous online products and infopreneurs. With the development of neural networks, the number of such specialists will increase, and the volume of their output will increase. At the same time, the quality of the materials offered will also improve, as GPT-based tools are capable of creating content of a higher level than what is currently offered on the infobusiness market.

Secondly, the difference between mass education and a high-quality VIP product lies in the approach to the needs of each student and how we develop their meta-subject skills. The most important aspects are, firstly, thinking, which allows one to transform acquired knowledge into personal understanding; secondly, verbal skills; and finally, most significantly, the process of self-actualization. Effective education should not only transmit information but also promote the development of critical thinking, communication skills, and personal growth in the student.

Self-actualization implies not just a set of skills, but a deep process of personal development aimed at unlocking one's potential. This is the desire for self-knowledge, self-improvement, and the realization of one's unique potential. Self-actualization involves understanding one's values, goals, and aspirations, as well as the ability to apply them in life. This is a state in which a person achieves harmony with themselves and the world around them, realizing their talents and abilities to the maximum.

I believe this skill is a key soft skill that significantly influences a person's success in life. It is a combination of awareness, willpower, initiative, and adaptability. Developing this skill allows one to effectively cope with challenges and achieve goals.

Self-actualization plays a key role in shaping our identity and facilitates the active movement from dreams to concrete actions. This process involves formulating desires, planning them, implementing them, and then analyzing the results. Completing this cycle brings satisfaction from what has been achieved, which, in turn, begins a new stage in the pursuit of goals and dreams. Effective self-actualization fosters personal growth and development, enabling each individual to achieve their maximum potential.

Self-actualization is a key factor that makes us effective individuals. Developing this aspect in education is of paramount importance. I believe that elite education can be accessible and free. What matters is not the financial aspect, but how it unlocks a student's potential and influences their thinking, speech, and the process of self-actualization. These aspects are no less important than traditional educational outcomes such as knowledge and practical skills. Developing self-actualization creates a deeper understanding and promotes personal growth, which makes education more meaningful.

Today's schools and universities can be viewed through the prism of elite education. In the modern educational landscape, elite institutions are those that offer a high level of teaching, unique programs, and access to resources unavailable in regular educational institutions. Such characteristics include highly qualified faculty, innovative teaching methods, and an individualized approach to each student. Furthermore, an important aspect of elite education is the opportunity to receive international grants and internships, which opens up global prospects for students. Thus, elite education forms not only professionals, but also leaders capable of influencing the development of society and the economy.

Read also:

• What AI assistants have learned to do in education
• “The most effective AI solutions reproduce proven pedagogical practices”
• 5 key findings on the development of independence in learning
• “Neural networks help develop and use fundamentally new formats”