Adaptive learning system and learning program therefor

The adaptive learning system combines AI with human supervision to deliver emotionally engaging, personalized guidance, addressing the emotional gap in AI-driven education and enhancing student motivation and retention.

JP2024165997AInactive Publication Date: 2024-11-28小玉 英明

Patent Information

Application Number
JP2023093547
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-18
Publication Date
2024-11-28
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing adaptive learning systems, despite technological advancements in AI, fail to replicate human emotional interaction, leading to decreased student motivation and increased dropout rates due to reliance on inorganic machines that lack emotional intelligence.

Method used

An adaptive learning system that integrates a large-scale language model with predetermined instruction content and human supervision, ensuring AI responses mimic specific teachers while being supervised by real teachers to provide heartfelt answers, thus maintaining emotional engagement.

Benefits of technology

Enhances student motivation and learning engagement by providing personalized, emotionally resonant guidance, reducing reliance on incorrect AI answers, and minimizing dropout rates.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024165997000001_ABST
    Figure 2024165997000001_ABST
Patent Text Reader

Abstract

To provide an adaptive learning system and learning program therefor, which reduce the workload of teachers, increase the learning appetite of students, and make it feel as if a dedicated home tutor is there to take care of each student.SOLUTION: An interactive AI system S with a large-scale language model designed to automatically provide appropriate answers to questions from learners is provided, where a prompt P set with given instruction content is linked to an AI teacher installed in the interactive AI system S.SELECTED DRAWING: Figure 2
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an adaptive learning (also called "active learning") system and its learning program that realizes more efficient and effective learning by providing learning content optimized for each client (student). [Background technology]

[0002] "Adaptive learning," which could be described as the ideal form of education, has the potential to reduce the burden on teachers and enable individual instruction tailored to the individual abilities and progress of each client (student). It is expected to be more efficient and promote more effective learning than traditional e-learning, and has been implemented in schools across the country.

[0003] However, after actual implementation, as shown in the references below, while some students saw learning outcomes, various problems were reported, such as a decrease in motivation to learn and students dropping out. Additionally, due to factors such as the high initial implementation costs, many school facilities are hesitant to switch to adaptive learning.

[0004] (References) ▲1▼Masuakuriumu: "Problems with Active Learning in High School Mathematics (February 2017)" ▲2▼Ministry of Education, Culture, Sports, Science and Technology: "Handbook of Active Learning Failure Cases" ▲3▼ University Journal Online "Why Active Learning Doesn't Work" etc.

[0005] Amidst such trial and error, around November 2022, a conversational AI system equipped with a large-scale language model (such as ChatGPT4) was announced, becoming a global topic of conversation. This latest AI (such as ChatGPT4) has made great strides in language comprehension and sentence generation capabilities compared to previous AI language models, and is now attracting attention in Japan as something that can respond in human-like language. Research and development has begun in industry, government, and academia, and it is expected that it will also be used in healthcare, education, the robotics industry, and more. [Prior art documents]

[0006] [Patent Document 1] Patent Publication No. 2022-7770 [Patent Document 2] Special Publication No. 2023-513095 [Patent Document 3] JP 2023-63444 A Summary of the Invention [Problem to be solved by the invention]

[0007] Indeed, with the emergence of the latest AI models (such as ChatGPT4), it can be said that AI (artificial intelligence) has achieved a dramatic technological advancement. However, no matter how technologically advanced AI (artificial intelligence) becomes, or how capable it is of responding in human-like language, it remains an inorganic machine (robot), not a living organism. Therefore, they cannot become human beings, creatures with rich emotions. This does not mean that the latest AI models are inferior to humans, but rather that they are now superior to humans in terms of their information processing and analytical capabilities. However, what this means is that AI (artificial intelligence) and humans are fundamentally different. However, the inventors of this case speculate that the above problems have arisen as a result of advances in AI technology, which has led humans to try to entrust even roles that are traditionally performed by humans to AI. In other words, no matter how well the AI ​​model imitated human language, it was not enough to stimulate the emotions of human students and increase their motivation to learn. (See the references above.) The inventors' goal is to provide an adaptive learning system and learning program that is co-created by AI and humans to increase students' motivation to learn, making them feel as if they have a specialized private tutor who is tailored to each student. [Means for solving the problem]

[0008] In order to solve the above-mentioned problems, the present invention employs the following configuration.

[0009] That is, the adaptive learning system according to the present invention is an adaptive learning system that realizes more efficient and effective learning by providing learning content optimized for each client (student), In a conversational AI system S equipped with a large-scale language model that automatically responds with appropriate answers to questions from clients (students), A prompt P with specific instruction content is linked to the Ai teacher installed in this interactive Ai system S.

[0010] Here, "predetermined instruction content" does not have the broad meaning of responding in human-like language, but rather means responding in language like that of a specific teacher (for example, a homeroom teacher).

[0011] The inventor of the present invention compared the answers of the latest Ai (ChatGPT4, etc.) with those of Ai Sensei linked to Prompt P, which is set with the specified instruction content of the present invention, and obtained many objective opinions that the learning motivation of the client (students) was increased. Therefore, with adaptive learning in this configuration, the effects can be obtained simply by linking (including internal or external linking) to existing learning systems such as E-learning, so there is no waste in the existing learning systems and investment costs can be reduced.

[0012] The instructions for Prompt P explicitly state that Teacher Ai should "set her role," "set her answer," and "set the answer conditions." "Role setting" does not mean that Teacher Ai will answer questions proactively like Teacher XX. Ms. Ai is a supporting role (assistant) to Mr. XX, and the people (students) she is speaking to are never aware of her existence. In other words, her scope of work is clearly defined. In the "answer condition setting" section, specific answer conditions according to the adaptive learning of the present invention are set. For example, information about a real teacher (such as name, gender, skills, etc.) is set as work history (history). In addition, prohibited information is also set out, such as not giving answers that deviate from the specified educational guidelines, and not giving answers that constitute power harassment, sexual harassment, or moral harassment. It is also possible to incorporate technical techniques such as "empathy," "conformity," "praise," and "echoing" that are backed by psychology, human behavior, and neuroscience. On the other hand, information about the person seeking advice (student) (name, usual name, gender, skills, etc.) is also set. "Setting the answer" means setting the timing when Teacher Ai will give the answer. For example, the "time" setting may be such that a response is given after an input from the client (student).

[0013] Therefore, the adaptive learning according to this configuration allows the client (student) to feel as if he or she has a specialized tutor just for him or her, which can increase the student's motivation to learn. It also prevents incorrect answers (such as power harassment, sexual harassment, and moral harassment) that are common among real teachers and new teachers.

[0014] In the adaptive learning system of the present invention, the answers of the Ai teacher, which automatically replies to questions from the client (student), are given to the client (student) through the supervision of a real human teacher. Even with the latest AI (such as ChatGPT4), incorrect answers and typos are still common. However, in educational settings where there are many inexperienced students, students often acquire new knowledge by trusting the teacher's answers and guidance. Therefore, in the invention having this configuration, a reply is given to the person seeking advice (student) through the supervision of a real teacher who is a human.

[0015] Here, "supervision" refers not only to visually checking the text that Ai-sensei has automatically answered, but also to making simple corrections to the text. Furthermore, the real teacher who provides the "supervision" does not necessarily have to be the actual homeroom teacher; it can be a specialist with skills and experience equal to or greater than those of the homeroom teacher. With this "supervision," the response speed may be slightly slower than that of Ai Sensei, who can automatically respond instantly to clients' questions. However, as mentioned above, the inventors believe that in educational learning situations, it is important to avoid giving incorrect answers to students even if this means a slightly slower response speed. Although the response speed is slower, it is still compared to Ai Sensei, and compared to a real teacher responding to a student's question from the beginning, the response speed is considerably faster (probably less than 1 / 10 of the work time).

[0016] This is related to the concept of the present invention, so it will be explained in more detail using figures. Figure 7 is a schematic diagram of the initial model of Ai Sensei. This initial model of Teacher Ai was expected to be a success because it could immediately respond to questions from the client (student) X. However, her responses were in robotic (mechanical, inorganic) language, and as can be seen from the references in the active learning section above, she lacked any sense of humanity, and her responses were met with mixed reviews.

[0017] Figure 8 is a schematic diagram of the latest model of Ai Sensei. As mentioned above, this latest model of Ai Sensei has enhanced processing capabilities to the point where it can respond in human-like language (e.g. ChatGPT). Figuratively speaking, the latest model, Ai Sensei, can be said to be human-like only in terms of her face (the round part).

[0018] 9 and 10 are schematic diagrams comparing the latest model of Ai Sensei (hereinafter simply referred to as Ai Sensei) with a real human teacher. As can be seen from these diagrams, the answers (Answer Ai, Answer A) of Mr. Ai and the real Mr. Ai are not the same. The reason is that Mr. Ai is not a living being and does not have emotions like human joy, anger, sorrow and happiness (in other words, a "heart"). The difference between the two is particularly pronounced when it comes to a language like Japanese, which is rich in emotional expressions such as "compassion," "consideration," "gratitude," and "hospitality." The present invention does not deny all AI, but limits the objects of information exchange (including text, voice, and images) to living things such as humans. In other words, the present invention excludes language interactions between robots, such as those in industrial machinery. For this reason, the adaptive learning system of the present invention is limited to subjects that are living things (including animals and plants) that have a "mind" like a human being.

[0019] As shown in Figure 10, the difference between Ai Sensei and Real Sensei is the presence or absence of "heart" in their answers. However, the presence or absence of this "heart" is a characteristic of each individual's personality.

[0020] In other words, Ms. Ai, who does not have a "heart," does not feel stressed about answering questions. In addition, because Ai Sensei's questions are two- or three-dimensional and not influenced by emotions, he is able to give stable answers. The explanation of dimensions is the understanding of the inventor, who is an expert on the mind, and details will be omitted in this specification, which is a technical document; however, simply put, if we consider flat objects to be two-dimensional and three-dimensional objects to be three-dimensional, then the "mind" means something that is four or more dimensions. Since Ai Sensei's answers are based on predetermined instructions (textbooks, educational guidelines, etc.), she can give consistent answers that are not influenced by her skills or experience as an educator like real teachers. Needless to say, Ai Sensei's answers can even be automatically processed to generate sentences at a speed dozens of times faster than humans. These are what we can call Mr. Ai's main strengths.

[0021] On the other hand, some of Ms. Ai's weaknesses were noted at the time of this application, including occasional errors in language translation, a tendency to make errors in answering calculation questions, and an inability to provide specific answers to current information, such as events occurring right in front of her. However, we hope that these shortcomings will be improved in the future as AI technology advances.

[0022] On the other hand, the strengths and weaknesses of Real Sensei, who has a "heart" inherent in him, are almost the exact opposite of Ai Sensei. In other words, because they have a "heart," they feel stress, are easily swayed by emotions, and find it difficult to give consistent answers, which can be a disadvantage. On the other hand, having a "heart" is an advantage in that it allows people to express loving emotions such as "compassion," "consideration," "gratitude," and "hospitality." This point will be discussed later.

[0023] Incidentally, the client (student) X in the adaptive learning system of the present invention is a human being with a "mind" of four or more dimensions. Therefore, no matter how excellent Mr. Ai is, when he directly talks with client (student) X, problems will arise because the two are on different levels. This point will be explained with an example to make it easier to understand.

[0024] Imagine that there is a picture of an apple drawn in pencil on a blank piece of paper. Anyone who sees this picture of an apple will recognize it as "a picture of an apple drawn on a flat surface." However, suppose that this picture of an apple is colored with shading, making it so that it is mistaken for a three-dimensional apple. When someone sees this picture of an apple, they will not think that it is a two-dimensional drawing, but will be misled into thinking that it is a "real apple." These are commonly known as "trick paintings" or "optical illusions." In other words, a two-dimensional drawing cannot become three-dimensional, no matter how much you try to resemble it in three dimensions. Therefore, no matter how excellent Mr. Ai is, even if he directly talks with the client (student) X, who is a human being with a "mind" of more than four dimensions, problems will arise in the form of discomfort, distrust, and a breakdown in the relationship of trust, because the two people are on different dimensions. To reiterate, in this invention, robots Ai in which information exchange takes place within the same dimension are excluded, and the adaptive learning system is limited to learning subjects that are humans (living things) with "minds" for the reasons stated above.

[0025] On the other hand, let's think about real teachers. As mentioned above, what could be called the strengths and weaknesses of Real Sensei are almost the exact opposite of Ai Sensei. So, what is the significance of Professor Real's answers? They are answers that have heart. Here, a "heartfelt answer" is, in other words, a "loving answer." The answer is to think carefully about what kind of response would be best to give to the person seeking advice (student) X, after learning about the current situation of that person (student) X. The answer may be sometimes harsh words toward the person (student) X, or it may be kind and sympathetic. However, no matter what the words are, at the core they are "words of love" toward the person (student) X.

[0026] Therefore, in the configuration of the present invention, Mr. Ai's answer is provided to the client (student) through the supervision of a real human teacher. According to the present invention, a "thoughtful answer" for the person seeking advice (student) is provided through "co-creation" between the "Ai teacher" and the "real teacher." For example, in addition to Ai-sensei's answers, the real teacher might say things like, "You did well, Mr. / Ms. XX," "That's really good," "That's close, think about it again," or "Keep up the good work on the next question." Simply adding a few words that correspond to the students' learning level can increase their motivation to learn.

[0027] At this point, those who lack literacy about the "mind" may be thinking, "Why not just set the prompts in advance and let the AI ​​teacher learn them?", so I'll go ahead and explain. Recall the story of the apple on the paper mentioned above. It should be obvious that a realistic drawing on paper, no matter how three-dimensional it may appear, is in fact two-dimensional (a drawing drawn on a flat surface), and so defects will occur. However, when a person looks at this picture, they misunderstand it and perceive it in three dimensions, and actually feel emotional emotions such as "it looks delicious" or "I want to eat it." This is the characteristic of a human being with a heart.

[0028] In other words, even if Ms. Ai gives an incorrect answer (not just a simple mistake, but also an answer that is not in line with the student's learning level or the curriculum guidelines), the student will still feel emotionally that "it's too difficult and I don't understand" or "it's impossible." Because this student's emotions are directly linked to a loss of motivation to learn and causing them to drop out of classes, it is far too dangerous to rely solely on Ai's answers. In particular, beginners learning new knowledge are at high risk of not even questioning Ai Sensei's answers and instead simply accepting them at face value. Of course, this risk applies to real teachers as well. Therefore, by "co-creating" between the Ai teacher and the real teacher, making the most of each other's characteristics, the real teacher can reduce the amount of time he or she spends working, and at the same time be able to consistently provide thoughtful answers to students.

[0029] In the "setting of answer conditions" of the present invention, at least the name of the client (student) is clearly indicated. When humans (living things) are called by a name, nickname, or other name that identifies them, they unconsciously pay attention. At first glance, this may seem like a natural thing, but in reality, even real teachers often forget to call students by their names politely. On the other hand, it would be unnatural to call each student by their name every time they answer. Therefore, by calling the student's name at the appropriate time during the response, it is possible to draw the student's attention and increase their motivation to learn, and it is also possible for the student to feel as if they are receiving a private tutor's response. In addition, motivation to learn can be further increased by including answers that incorporate technical techniques such as "empathy," "conformity," "praise," and "echoing" that are backed by psychology, human behavior, and neuroscience. Effect of the Invention

[0030] The present invention has the following advantages. AI and humans can work together to provide an adaptive learning system and learning programs that will increase students' motivation to learn, making it as if each student had a specialized private tutor who was tailored to them. [Brief description of the drawings]

[0031] [Figure 1] FIG. 1 is a schematic diagram showing an example in which the present invention is applied to E-learning (self-study system). [Diagram 2] FIG. 2 is a schematic diagram showing an example of an operation state carried out behind the scenes of the E-learning (self-study system) shown in FIG. [Diagram 3] 2 is a chart showing an example of a learning program S used in the present invention. [Figure 4] 1 is a schematic diagram illustrating an example of a prompt P associated with a learning program S. FIG. [Diagram 5] FIG. 3 is a schematic diagram showing an embodiment in which there are a plurality of clients X in FIG. 2. [Figure 6] FIG. 6 is a schematic diagram showing an embodiment of an adaptive learning system S in which the real teacher shown in FIGS. 1 to 5 is the real teacher himself. [Figure 7] This is a schematic diagram of the early model of Ai Sensei. [Figure 8] This is a schematic diagram of the latest model of Ai Sensei. [Figure 9] This is a schematic diagram comparing the latest model of Ai Sensei with a real human Sensei. [Figure 10] This is a schematic diagram comparing the latest model of Ai Sensei with a real human Sensei. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0032] Hereinafter, an embodiment of the present invention (hereinafter also referred to as "the present embodiment") will be described with reference to the drawings. FIG. 1 is a schematic diagram showing an example in which the present invention is applied to E-learning (self-study system). Reference numeral 100 denotes a screen of a PC, tablet, or the like, and in the upper portion of the screen 100, a display surface 110 for E-learning (self-study system) is displayed. By following the instructions on this display surface 110 and selecting and writing in answer columns (omitted) shown on the same display surface 110, students etc. progress through the learning content step by step and acquire new knowledge etc. The display surface 110 may be in a stepwise manner in which learning tasks are provided step by step according to the learning content (subject), or in a non-stepwise manner in which the learning task can be started from any learning task. Moreover, on the display surface 110, not only text display but also a combination of photographs, audio, video, illustrations, etc. is used. Reference numeral 120 denotes a chat screen where questions and answers are exchanged between students and consultants regarding any doubts that arise during their studies. Questions from the client (student) are basically entered as text, but can also be entered by voice or selected from a preset pull-down menu. In this embodiment, a case where a general text sentence is input is shown as an example. To the side of the chat screen 120, a snap screen 130 is displayed in which a face photo, video, etc. of a teacher (for example, a homeroom teacher) is displayed. Reference numeral 140 denotes a time display unit for measuring the study time. The time display section 140 not only has a timer function for specifying the time until an answer is given, but also a stopwatch function for displaying the cumulative learning time.

[0033] Figure 2 is a schematic diagram showing an example of the operating state taking place behind the scenes of the E-learning (self-study system) shown in Figure 1, Figure 3 is a schematic diagram showing an example of the learning program S, and Figure 4 is a schematic diagram showing an example of a prompt P linked to the learning program S. In addition, parts common to those in FIG. 1 are given the same numbers and their explanations are omitted. A consultant (student) X proceeds with his / her study while looking at the display surface 110, but when a question arises about a problem along the way, he / she writes the question in text form on the chat screen 120 and sends it (1). The received question is automatically answered by the background teacher Ai, and a specific instruction is set in the prompt P so that the answer Ai▲2▼ is to be in the language that a real teacher (e.g., the homeroom teacher) would use. This is to reduce, even just a little, the amount of work that Real Sensei has to do to correct texts. In addition, the answer Ai▲2▼ is not immediately returned to the client (student) X, but is first displayed in the screening section K and is supervised by the real teacher. After the real teacher judges that the answer Ai▲3▼ is free of any problems and checks it for the final time, the answer Ai▲3▼ is sent directly to the chat screen 120 of the client X as answer A. On the other hand, if the real teacher judges that an answer Ai▲3▼' is problematic, the real teacher corrects the answer Ai▲3▼' and supervises it again. Only the answer A that has been finally checked is sent back to the chat screen 120 of the client X. According to the present invention, the existence of Mr. Ai, who has been "set as a supporting actor," is not known to the client (student) X. In addition, Answer A, which is returned after the supervision of Real Sensei, is a heartfelt Answer A filled with love from Real Sensei, and so it resonates with the heart of the person seeking advice (student) X, increasing the person seeking advice (student) X's motivation to learn. On the other hand, the answer Ai▲2▼ before supervision was instantly completed as a text sentence by Mr. Ai based on the “answer condition settings.” Moreover, since Ai-sensei is not influenced by emotions, she is able to give consistent answers. Therefore, the real teacher does not feel any stress until the answer Ai▲2▼ is given. The real teacher only has to supervise answer Ai▲2▼, so with minimal work he can provide a thoughtful answer A to the client (student) X. In other words, a real teacher with a heart can communicate with a client who also has a heart, as people who share the same emotional dimension. This human communication gradually builds trust between the students, and the consultant, like a trusted private tutor, prevents the student from abandoning their studies, brings out the student's learning ability and makes them realize that "studying is fun," thereby increasing their motivation to learn. For this reason, Ms. Ai must strictly adhere to her role as a supporting actor and must not let her client (student) X know of her existence. It should be noted that with future advances in AI technology, it is expected that the answers given by Ai Sensei will become even closer to the answers given by the real teacher. However, this will simply mean that the real teacher will have less correction work to do, and will not eliminate the need for supervision by the real teacher, who is a human.

[0034] FIG. 5 is a schematic diagram showing an embodiment in which the number of clients X in FIG. 2 is multiple. 2 are given the same reference numbers, and their explanation is omitted. Only the features of FIG. 5 will be explained. Questions are sent via text to each of the people seeking advice (X1, X2, X3...) through an account that is identified for each person ▲1▼. The supporting role of Mr. Ai is provided with a control unit (not shown) and is connected to each of the Ai teachers (X1, X2, X3, ...) to which "predetermined instruction content" corresponding to each client (X1, X2, X3, ...) is set. Therefore, even if the client (X1, X2, X3...) is in a class with a large number of students, Answer A that is given after Real Sensei's supervision is a heartfelt Answer A filled with love from Real Sensei, so Answer A that resonates with the hearts of each student (X1, X2, X3...) can be given, thereby increasing the learning motivation of each client (student) (X1, X2, X3...).

[0035] FIG. 6 is a schematic diagram showing an embodiment of the adaptive learning system S in which the real teacher shown in FIGS. 1 to 5 is the real teacher himself. 6. Since the numbers in the figure are the same as those in FIGS. 1 to 5, their explanation will be omitted and only the features of FIG. 6 will be explained. The embodiment of FIG. 6 shows a case where a user directly chats with a real teacher. For example, in an adaptive learning system that uses online meetings such as Zoom, it may be necessary for the real teacher to meet with client (student) X. The question ▲1▼ received from the client X is automatically answered by the supporting role Ai-sensei, and the answer Ai▲2▼ is displayed on the screening section K of the real teacher. Therefore, after watching and supervising the screening section K, the real teacher can respond to the client X with a heartfelt answer A not only in text form but also verbally depending on the situation. ▲4▼ In the embodiment of Figs. 1 to 6, the real teacher responds with a text sentence, but it is of course possible for the real teacher to respond by voice while looking at the text sentence or by converting it into voice. [Explanation of symbols]

[0036] ▲1▼…(Consultant X) sends ▲2▼…(Ai Sensei)'s answer Ai ▲3▼…(Before supervision) Answer Ai ▲4▼…(After supervision) Answer A 100…PC screen 110…Display surface 120...Chat screen 130…Snapshot screen 140…Time display section S…Adaptive learning system K: Screening section

Claims

1. An adaptive learning system that provides learning content optimized for each individual learner, enabling more efficient and effective learning. In a dialogue-based AI system S equipped with a large-scale language model that automatically responds with appropriate answers to questions from learners, This is an adaptive learning system in which a prompt P with specified instruction content is linked to the Ai teacher installed in this interactive Ai system S.

2. 2. The adaptive learning system according to claim 1, wherein the instructions of the prompt P clearly indicate "setting of role," "setting of answer," and "setting of answer condition" of the teacher Ai.

3. 3. The adaptive learning system according to claim 1, wherein the AI ​​teacher's automatic responses to questions from the learner are provided to the learner through supervision by a real human teacher.

4. An adaptive learning system as described in any one of claims 1 to 3, wherein questions from a learner are posed to a real human teacher, and answers from the Ai teacher are provided to the learner through supervision by the real human teacher.

5. The adaptive learning system according to any one of claims 1 to 4, wherein the "answer condition setting" indicates at least the learner's name.

6. A learning program for use in the adaptive learning system according to any one of claims 1 to 5.

Citation Information

Patent Citations

  • Learning system, learning method, and learning program

    JP2022007770A

  • Proactively Incorporating Unsolicited Content into Human-Computer Dialogue

    JP2023063444A

  • Machine learning video curation for selection and display

    JP2023513095A

Cited By

  • system

    JP2026155552A

  • system

    JP7900551B1

  • Communication Learning Support System

    JP7944637B1