Education

Artificial Intelligence in Education: Exploring Real-World Practices

Artificial Intelligence in Education: Exploring Real-World Practices

Learn: Producer of online courses from scratch to PRO

Learn more

Artificial intelligence in education is not just a concept of the future, but real tools and technologies that are already in use today. Rather than focusing on fantastical scenarios, it is important to consider the current state and potential of AI in education. Real possibilities include adaptive learning systems, personalized curricula, and automated assessments that improve the quality of education and make it more accessible. What are the specific benefits of artificial intelligence in education, and how can they change the approach to learning?

In June, SberUniversity held a seminar entitled "AI in Education: What Can Technologies Do Now?" The goal of the event was to identify the balance between real possibilities and fantastical expectations of artificial intelligence technologies in the educational process. Experts from SberUniversity, HSE University, and University 2035 participated in the seminar, which provided an opportunity to discuss current trends and prospects for using AI in education.

During the discussion, participants addressed several key issues. They examined relevant topics related to current industry trends and exchanged views on best practices and new technologies. Participants focused on the importance of innovation and adaptation to rapidly changing market conditions. The role of teamwork and effective communication in achieving shared goals was also highlighted. At the end of the discussion, the participants came to the conclusion that continuous learning and development are necessary for successful advancement in their field.

  • What AI algorithms are used in education right now;
  • Real cases of implementing AI in education;
  • In what educational projects does it make sense to use AI;
  • Why is it important for those who want to implement AI in an educational project to keep ethics in mind.

What AI algorithms are used in education right now

Andrey Komissarov, Director of the Data-Based Human Development Department at University 2035, and Andrey Petrovsky, Executive Director of Data Research at the Sberbank Artificial Intelligence Lab, shared information about the three main types of artificial intelligence that are actively used in the educational process. These technologies help improve the quality of education, adapt educational materials to the needs of students, and analyze their progress. Artificial intelligence in education is becoming an important tool for creating a personalized approach to learning and increasing student engagement.

Ensembles of algorithms based on logic of varying complexity. This type of AI is responsible for creating individual learning paths or analyzing the effectiveness of a lesson plan.

  • Pre-trained neural networks are created to solve a specific problem and trained on a large amount of data. Such neural networks can, for example, be responsible for automatic language understanding or recognizing emotions on students' faces. Experts noted that this typically refers to opensource models, such as the foreign GPT-3 and BERT, or the domestic YaLM 100B from Yandex or RUGPT-3 from SberDevices.
  • Neural networks, which belong to the field of shadow deep learning. In this case, a pre-trained neural network is further trained on a smaller amount of data and adapted to solve a more specific problem. For example, when a language neural network is specialized to check open-ended assignments for a specific course.
  • Andrey Komissarov emphasizes that currently, the main area of ​​artificial intelligence application in education is related to data science. This includes data analysis to form hypotheses, testing them, and using the tested hypotheses to achieve results. In his opinion, in the future, we will witness a wider application of big data-based learning. However, at present, there is a problem with a lack of large-scale data, which limits the opportunities for the full use of artificial intelligence technologies in educational processes.

    Examples of AI Use in Education in Russia

    During the discussion, each speaker presented practical examples of the implementation of artificial intelligence in the educational process. We have systematized these cases by technology use criteria, allowing us to better understand their effectiveness and scope of application.

    Andrey Komissarov shared two practical examples concerning soft skills that can be effectively applied in education. The first case concerns team selection, where it is important to consider not only the professional qualities of the participants but also their ability to interact. The second case concerns communication skills assessment, which allows us to identify strengths and weaknesses in communication and develop individual programs to improve these skills. These approaches can significantly improve the effectiveness of the educational process and facilitate the development of essential competencies in students.

    As part of the National Technology Initiative accelerator, a team selection system using artificial intelligence was developed. The main goal for the scientists was to form teams from program participants capable of generating ideas and launching startups based on collaborative efforts. This innovative methodology is aimed at optimizing the process of creating effective teams, which contributes to the development of startups and the implementation of new technologies.

    Photo: Jacob Lund / Shutterstock

    The organizers conducted a preliminary Participants were diagnosed and five key characteristics/harmonies were identified, which formed the basis for team formation. These characteristics included a harmony of knowledge, roles, personality traits, values, and interests. Each of these harmonies ensured the necessary combination of hard and soft skills, which contributed to the effective solution of assigned tasks. This approach allowed for the creation of balanced teams capable of maximizing the potential of each participant. Komissarov emphasizes that this approach contributes to the creation of more effective teams than simply working with expert tutors, which in turn increases their chances of successfully completing the accelerator. He also emphasizes the need for further team monitoring, including analysis of digital interactions between participants and educational content. This will help identify key factors influencing the development and success of teams during the learning and adaptation process. The expert notes the significant potential of using artificial intelligence to form effective learning groups where students are able to better understand and support each other. This is especially relevant in corporate training, where effectiveness is determined by business metrics. In this context, the cost of mismanaged employee time becomes especially tangible. Using AI to optimize group learning can significantly improve productivity and reduce costs, which is key to successful company development.

    Communication skills assessment using artificial intelligence can significantly improve the effectiveness of training and development for sales managers, educators, and other communications professionals. Andrey Komissarov cited a project that analyzed audio recordings of meetings and conferences to assess participants' communication skills. This approach allows for an objective assessment of interactions, identifying strengths and weaknesses in communication, and developing strategies for improving communication. The use of AI in this area opens new horizons for professional growth and improving the quality of interactions in various fields.

    Experts claim that artificial intelligence can analyze the complexity, coherence, and clarity of human speech. AI also evaluates vocabulary diversity and the number of empty phrases, making it possible to identify criteria for well-developed communication skills and identify common communication problems. Such analysis facilitates a deeper understanding of speech characteristics and helps develop effective communication strategies.

    Analyzing students' communication skills can be a useful tool for assessing their current level. It allows for the recommendation of optimal development paths and the evaluation of the effectiveness of the educational process by comparing results before and after training. This approach promotes more targeted communication skills development and improves the overall effectiveness of education.

    Denis Federiakin, research fellow and lecturer at the Institute of Education at the National Research University Higher School of Economics, presented two ideas demonstrating the use of artificial intelligence technologies in educational assessment. The first idea concerns the creation of assignments, and the second concerns the automatic checking of open-ended responses. These initiatives highlight the potential of AI in optimizing educational processes and improving the quality of student assessment. The use of such technologies can significantly facilitate the work of teachers and improve feedback for students, which, in turn, contributes to more effective learning.

    The automatic development of assignments in Russian language and literature for schoolchildren was implemented in collaboration with scientists from the Pushkin Institute of the Russian Language. Research has demonstrated the feasibility of using two types of artificial intelligence for this purpose. This approach enables the creation of learning materials tailored to students' knowledge levels, which facilitates more effective subject acquisition. The use of modern technologies in the educational process opens new horizons for students and teachers, improving the quality of learning and increasing interest in subjects. Simple algorithms: These help generate tasks that test students' understanding of Russian language rules. For example, selecting a letter representing a voiced sound from several options. This method of creating assessment tasks is suitable for situations where there is a clear list of Russian language rules that can be described to the AI. Advanced neural networks: These are capable of generating tasks that test language literacy and text comprehension. For example, arranging jumbled sentences in a text in order or choosing the appropriate phrase to fill in a gap in a sentence. The RUGPT-3 language model, pre-trained on a Russian language corpus and non-specific to any age or topic, was used as a basis. Powerful machine learning was applied to a smaller data set, enabling the model to generate Russian language and literature assignments for schoolchildren. Denis Federiakin emphasizes the significant potential of artificial intelligence in assessing open-ended assignments and essays. The subjectivity inherent in expert evaluation of open-ended assignments presents a serious problem. Psychometrics has been developing methods for many years aimed at eliminating individual biases that can arise during assessment. Recommendations for the use of various data processing methods help make the assessment process more objective and fair.

    Photo: Gorodenkoff / Shutterstock

    The speaker claims that the psychometric principles applied in the work With people, can be effectively used in training neural networks. This can lead to improved results, as integrating psychological aspects into machine learning algorithms can increase their efficiency and accuracy.

    Revised text:

    Be sure to read the additional materials:

    A program has been developed that can check English essays more effectively than a teacher. This technology uses modern natural language processing algorithms to analyze the structure, grammar, and style of texts. With this program, students can receive instant feedback and recommendations for improving their work. The program takes into account various aspects of writing, making it a useful tool for learning and self-improvement in English.

    Federiakin confirmed his hypothesis by citing an example from a related field. Researchers from Harvard created a neural network for analyzing social networks and detecting hate speech. When further training the model, they used the principles of developing rating rubrics for experts. As a result, the neural network demonstrated high performance in identifying inappropriate comments, outperforming both human moderators and Google's neural network trained using traditional data science methods. This case highlights the importance of innovative approaches to model training and their potential advantage over existing content moderation solutions.

    Andrey Komissarov and Alexey Zaytsev, founders of the educational project "01Mathematics," presented three effective ways to apply artificial intelligence in learning analytics. These solutions are aimed at improving the educational process, optimizing individualized learning, and enhancing the quality of learning. Using AI in learning analytics allows for analyzing student data, identifying their strengths and weaknesses, and offering personalized recommendations for successful learning. Thus, artificial intelligence is becoming an important tool for educational institutions striving for innovation and improved learning effectiveness.

    The Moscow Electronic School has implemented predictive analytics, which, according to Alexey Zaytsev, allows teachers to estimate how much time students will spend on various activities both in the classroom and at home. This includes assignments, tests, and practical exercises. Teachers can use these predictions to optimize lesson planning and monitor how effectively the class is progressing through the curriculum. Thirty-seven percent of Moscow schools are participating in the project, providing artificial intelligence with sufficient data for analysis and accurate predictions. This solution significantly improves the quality of the educational process and helps teachers better adapt their teaching approaches.

    The 01Mathematics project has implemented analytics for the quality of educational materials. Artificial intelligence analyzes data on the percentage of successfully completed tasks, the average speed of their completion, and the statistics of student errors. As a result, the AI ​​provides the methodologist with a hypothesis about tasks that contain errors or are poorly understood by students. This analysis is key, especially in the first weeks after the course launch, noted Alexey Zaytsev.

    Andrey Komissarov presented the digital reflection analytics implemented in the "Digital Professions" program from the Ministry of Digital Development, which is attended by approximately 300,000 people. The main goal of digital reflection is to determine what knowledge and skills students acquire after completing each module, as well as to understand how they plan to apply this knowledge in practice. The process involves collecting and analyzing data on how students understand the educational material, which helps improve the educational process and increase its effectiveness.

    • After each module of the program, the student was asked to describe their comments in free form, to talk about what was useful in the module and how they plan to use it;
    • We collected an array of data on these digital reflections of students, filtered out erroneous answers (for example, those where the user, instead of their review, simply copied and pasted the course description or sent a meaningless set of characters);
    • We analyzed the informativeness of the reflection, that is, we assessed how many words and phrases related to the course topic are retained in the digital reflection after training.

    This became possible thanks to a neural semantic network, the speaker said. Neurosemantic networks are an advanced technology that allows you to process and analyze large volumes of data, which opens up new possibilities in various fields, including artificial intelligence and natural language processing.

    Andrey Petrovsky from the Sber Artificial Intelligence Laboratory explained how to organize lecture notes using open models. The process involves two main steps: first, speech is converted to text using the Speech to Text speech recognition model, and then, using language models such as BERT, key sentences that best reflect the lecture's content are extracted from the resulting transcript. This technology can be particularly useful for methodologists for automating the creation of memos and teaching materials for students. The use of modern artificial intelligence tools can significantly improve the efficiency of the educational process and enhance the quality of learning.

    Alexey Zaitsev and Denis Federiakin shared their experience implementing adaptive learning for the Moscow Electronic School based on the 01Mathematics project. This project implemented an adaptive fading system that initially guides students and then gradually empowers them to develop independently and make decisions throughout the learning process. This approach creates an individual educational trajectory, which significantly increases the effectiveness of material assimilation and promotes the development of independence in students.

    Read also:

    To improve the SEO optimization of your content, it is important to consider keywords and phrases that can attract the target audience. Understanding the needs and interests of your audience will help you create more relevant and useful content. Use natural language and avoid keyword overuse to keep your text readable and engaging. Also, remember the importance of internal and external linking, which will help improve your website's indexing and search engine visibility. Adaptive learning is a method that personalizes the learning process based on the needs and knowledge level of each student. This form of learning uses modern technology to analyze student performance and tailor learning materials to their individual needs. Why is adaptive learning important? Firstly, it promotes more effective information acquisition, as each student receives content tailored to their level. Secondly, this system helps maintain learning interest by allowing students to work at their own pace. Third, adaptive learning helps identify knowledge gaps and quickly address them, significantly improving overall academic performance.

    In a rapidly changing world and constantly updated information, adaptive learning is especially relevant. It not only improves the quality of education but also prepares students for the real-world challenges they may face in the future. Thus, adaptive learning is an important tool for creating a more effective educational environment.

    Advances in big data analytics have enabled in-depth studies of student learning. Collecting information on the speed of video lesson completion, error rates during interactive tasks, and student interactions with the educational system (including the use of additional materials and working with graphs) opens up new opportunities for optimizing the learning process. This approach helps identify each student's individual needs and tailor educational materials, promoting more effective learning.

    According to representatives of 01Mathematics, the platform's artificial intelligence can detect when a student is struggling with a particular learning module and provide timely assistance. This may include additional materials on a relevant topic, simplified learning tracks, or review of previously covered concepts that may impact current understanding. The technology analyzes the digital learning footprint of each individual student and compares it with data from tens of thousands of other students, allowing over 90% of students to successfully complete the modules they study and move along an individually tailored educational path, Zaitsev noted.

    In which projects does it make sense to use AI?

    Expert research shows that in learning tasks that can be algorithmized and are repeated, artificial intelligence technologies demonstrate results comparable to, and sometimes even superior to, human abilities. Denis Fedyarkin noted that the use of AI in such areas opens up new opportunities for improving the efficiency and quality of task performance. This demonstrates significant progress in the development of technologies that can significantly simplify and optimize processes that previously required human involvement. The use of artificial intelligence in education is becoming increasingly relevant and in demand, which underscores its importance in the modern world.

    Artificial intelligence effectively copes with the automation of routine intellectual work. If you notice repetitive actions in a process, this indicates an opportunity to implement artificial intelligence technologies. Automating routine tasks with AI not only increases productivity but also frees up time for more creative and strategic tasks. Using artificial intelligence helps optimize workflows and reduce the likelihood of errors, which is an important step towards improving overall business efficiency. According to the expert, successful automation of educational processes requires access to large volumes of data. For example, using artificial intelligence to generate educational assignments is justified within the framework of a project covering all schoolchildren in Moscow. At the same time, such a practice would be ineffective for small, proprietary online schools with a limited number of students. Should those without sufficient data ignore the possibilities of artificial intelligence? Andrey Petrovsky argues that this is not the case. Indeed, classical language models such as BERT and GPT-3 require enormous amounts of data for training – from petabytes to hundreds of gigabytes. This amount exceeds the capabilities of a typical personal computer. However, further training of open neural networks for specific educational tasks is possible even with smaller data sets. For example, a dataset of 12,000–16,000 examples is sufficient for successful training. These indicators serve as a benchmark for those planning to use artificial intelligence technologies. The expert recommends considering these nuances when integrating AI into educational processes.

    Data volume is an important, but not the only, criterion for assessing the feasibility of using artificial intelligence. Time and cost are also significant factors. Denis Federiakin noted that the common belief that significant progress can be expected in just one or two years after implementing AI is misleading. It's important to consider that implementing AI requires time for adaptation, model training, and integration into existing processes.

    Vyacheslav Yurchenkov, Head of the Educational Technology Center at SberUniversity, noted that implementing artificial intelligence in education requires a realistic approach. He emphasized that even with immediate implementation of the technology, one shouldn't expect immediate results. The amount of work won't immediately decrease, as data must be collected and properly labeled for system training. Only then will artificial intelligence be able to provide useful results for both teachers and students. It is important to understand that the process of integrating technology into the educational process requires time and effort to achieve effective results.

    Four Calls for Those Wanting to Implement AI in an Educational Project

    At the end of the seminar, experts presented four ethical recommendations for educational project managers considering integrating artificial intelligence into their activities. These recommendations will not only help improve the effectiveness of projects but also ensure the ethical and responsible use of technology in education.

    When implementing any technology, attention must be paid to its testing and verification, which requires advance planning of the project time. In education, the operation of artificial intelligence is monitored by methodologists, teachers, and training and development (T&D) specialists, depending on the context of the technology implementation. This ensures the high-quality and effective use of AI, allowing it to be adapted to educational processes and improve their effectiveness.

    Andrey Komissarov noted that no educational organization in the world using artificial intelligence does so without oversight. This statement is true even for those institutions that have large volumes of data and powerful models, the development and training of which have required significant resources. It is important to understand that the use of artificial intelligence in education requires careful monitoring and evaluation to ensure the effectiveness and safety of learning processes.

    Read also:

    Recommendations for the implementation of artificial intelligence in the educational process have been created. These recommendations aim to optimize learning, improve teaching effectiveness, and personalize the educational experience for students. Artificial intelligence's ability to analyze data, predict student performance, and provide personalized recommendations makes it a valuable tool in modern educational institutions. The introduction of AI into education opens new horizons for improving the quality of learning and preparing students for the demands of the modern world.

    The use of artificial intelligence should be limited to areas where its decisions and the reasons for them can be clearly tracked and analyzed. This is especially important in educational projects targeting children and adolescents. Errors or misinterpretations of data can have serious consequences for the future of young people. Therefore, it is essential to ensure transparency and understandability of AI to ensure its safety and effectiveness in education.

    Denis Federiakin cited an example supporting his thesis: a neural network judge created by scientists systematically discriminated against the Black population, and its developers did not immediately notice this. There's also a well-known case involving Amazon's neural network, which discriminated against women when screening resumes for the HR department. Such incidents are not isolated and highlight the importance of thoroughly checking algorithms for bias. These examples serve as a reminder of the need to consider ethical aspects when developing and implementing artificial intelligence technologies, which is critical to ensuring fairness and equality in the use of automated systems. Denis Fedyarkin noted that there are many pitfalls in transparency that can go unnoticed. This highlights the importance of careful analysis and a deep understanding of the processes involved in transparency to avoid unexpected complications and misunderstandings.

    Still: The Matrix / Warner Bros.

    Alexey Zaitsev proposed replacing teachers with artificial intelligence to save time and money. While this proposal seems appealing, experts warn that such a replacement could be misguided. Social interaction and the teacher's personality play a key role in ensuring a quality education. Therefore, artificial intelligence should not be viewed as a replacement for teachers, but as an effective assistant in the educational process.

    The optimal solution is to use technology to fill the gaps in education that teachers cannot fill. According to experts, if technologies are developed by professional methodologists, they can significantly improve the accessibility and quality of education in regions where access to qualified teachers is lacking. This emphasizes the importance of integrating modern educational technologies into the education system to ensure equal opportunities for all students.

    The digitalization of education is becoming an increasingly relevant topic, but its true purpose – helping students unlock their potential – is sometimes forgotten. Experts emphasize the importance of maintaining a human touch in projects using artificial intelligence. Andrey Komissarov noted, "We need to take into account our hesitations in choices, mistakes, and emotions so that artificial intelligence can contribute to a more effective and efficient educational process." It's important to remember that technology should serve as a tool for enhancing the educational experience, not a substitute for human interaction. Optimize your text for SEO and improve its content, avoiding unnecessary elements and maintaining the topic.

    Read also:

    • What is digitalization of education and why is it needed?
    • Virtual reality is coming to schools.
    • How schools and universities teach using virtual and augmented reality.
    • Cool ideas: a selection of interesting foreign EdTech projects.
    • VR, Big Data, and artificial intelligence: how will these technologies change education?

    Online Course Producer from Scratch to PRO

    You will learn how to launch profitable courses and webinars with minimal investment. You will understand how to turn existing EdTech projects into profit and make money on online learning.

    Find out more