Has AI Teaching Reached Parity with Human Teachers?—How "8 Minutes of Pre-Class Dialogue" Can Transform Learning

Since generative AI began entering educational settings, the debate over whether "AI will replace teachers" has been repeatedly discussed. AI can provide individualized explanations, engage with struggling students repeatedly, and answer questions even late at night. While these advantages are often highlighted, there is a strong counterargument that a teacher's role is not merely about transmitting knowledge. Reading children's expressions, motivating them, managing groups, and supporting them after failures—can AI be entrusted with such human interactions? This question remains significant among educators.

The research from the Hong Kong University of Science and Technology has introduced new empirical evidence into this debate. The research team demonstrated that just 8 to 10 minutes of one-on-one conversation with a teacher before an online lecture can improve students' brain activity synchronization and learning outcomes. Notably, whether the conversation partner was a human teacher or an AI teacher, there was no significant difference in learning outcomes.

However, it is premature to simply interpret these results as "AI teachers can completely replace human teachers." The research does not show that AI and humans achieved the same results in the same way. Rather, human teachers and AI teachers reached similar learning outcomes through different pathways. This is where the intrigue of this research lies.

What the Research Entailed

The experiment involved 57 university students divided into three groups.

The first group received video lectures without any prior conversation.
The second group had an 8 to 10-minute conversation with a human teacher before the lecture.
The third group conversed for the same duration with an AI teacher that resembled a human in appearance and voice.

GPT-4 was used for the AI teacher, integrating voice recognition, content generation, voice synthesis, and real-time talking head display. Students were aware they were conversing with an AI, meaning it wasn't an experiment to make them believe the AI was human.

Afterward, all students watched the same 14-minute video lecture. Notably, the study measured not only post-lecture test results but also brain activity and gaze movement during the lecture. The research team used MRI and eye-tracking to examine how students directed their attention and processed information in the brain during the lecture.

A Few Minutes of Conversation Changes Learning Readiness

The results were intriguing. Students who spoke with a human teacher and those who spoke with an AI teacher both showed better learning outcomes than those without prior dialogue. The effect was particularly noted in challenging comprehension questions.

In terms of memory, understanding, and application of knowledge, there was no significant difference between the human teacher group and the AI teacher group. This means that under the conditions of this experiment, a short dialogue with an AI teacher boosted learning just as effectively as a short dialogue with a human teacher.

It is important to note that the AI did not conduct the lecture itself. The AI merely engaged in a brief conversation with students before the lecture. Yet, this still resulted in differences in learning outcomes. This suggests that short pre-class dialogues play a role in preparing students to learn.

In educational theory, the support provided by a teacher to help learners reach new knowledge is called "scaffolding." This study suggests that such scaffolding can be provided to a certain extent by AI, not just humans.

Human Teachers and AI Teachers Are Not the Same

However, the fact that learning outcomes were similar does not mean that human teachers and AI teachers have become the same. The research clearly shows the differences between the two.

Students who conversed with human teachers had more synchronized gaze movements during the lecture. Their gazes, whether among students or between students and teachers, tended to align, focusing attention on the same materials. This suggests that conversations with teachers socially organized "where to look" and "what to focus on."

Moreover, interactions with human teachers were perceived as having stronger social closeness and emotional connections. The presence of a human teacher made students feel "observed" and "attended to," which could influence concentration and persistence in learning.

On the other hand, students who conversed with AI teachers did not exhibit as strong a sense of social closeness or gaze alignment as those with human teachers. Yet, their learning outcomes were similar to the human teacher group. This indicates that AI can support learning through different mechanisms without forming the exact same social relationships as humans.

AI teachers likely prepared students cognitively through organizing learning content, posing questions, directing attention, and providing individualized responses. In other words, human teachers might have used social and emotional routes more strongly, while AI teachers might have focused on cognitive support.

It's Not About "AI Replacing Teachers"

Some might view this study as evidence that "AI can indeed replace teachers." However, caution is needed.

This experiment was conducted under relatively limited conditions with 57 university students. The subjects were university students, not elementary or middle school students. The lecture was a 14-minute video, and AI did not support learners over the long term. It was not a study comparing the broader roles of teachers, such as classroom management, career counseling, or emotional care.

Therefore, concluding from these results that "AI teachers make human teachers unnecessary" would be an overreach. Instead, this study should be seen as material for considering where AI can supplement teachers' work, rather than whether AI can replace them.

In particular, online learning often leads to learner isolation. Simply playing video materials does not sustain concentration. Learners may not know what to focus on, and they may not be mentally prepared to learn. These issues have long been pointed out in MOOCs and remote classes.

If AI can engage in brief pre-lecture conversations, organize the day's learning content, address learners' questions, and draw their interest, it might compensate for the weaknesses of online education. The value of AI teachers emerges not as complete replacements but as assistants at the gateway to learning.

Expectations and Concerns Seen on Social Media

 

This topic is also prone to debate on social media. While large-scale reactions to this particular article are still limited, public discussions about AI teachers and AI education generally fall into three categories.

The first is the expectation for AI.
AI can tailor explanations to each learner, never getting upset no matter how many questions are asked. It can explain difficult content using different metaphors. For regions lacking teachers or children who cannot receive sufficient learning support at home, AI might be a significant help. Such opinions are often seen in communities positive about AI.

Indeed, among users sharing their experiences of learning math or languages with AI, there are reactions like "I finally understood content that I couldn't grasp at school thanks to AI's explanation." These voices appreciate AI's ability to adapt explanations to the learner's pace.

The second is a cautious stance emphasizing the role of teachers.
On educational social media, many responses emphasize that "a teacher's job is not just to pass on information." There are students who cannot concentrate in class, those who are not self-motivated to learn, and those who have personal struggles. Teachers support the foundational aspects of learning by observing each student's condition.

On social media, opinions such as "School is not just a place to learn knowledge," "Teachers teach human relationships and social skills," and "AI alone cannot even keep children seated and focused" are seen. This aligns with the social and emotional strengths of human teachers highlighted in this study.

The third is the perspective of considering collaboration rather than replacement.
From this standpoint, AI is viewed not as something that erases teachers but as a tool for teachers. For example, AI excels in areas like lesson preparation, grading assistance, material creation, rubric development, and initial responses to individual questions. Teachers can then spend more time on dialogue with students, confirming deep understanding, and classroom building.

This study also aligns well with this collaborative model. AI handles brief pre-learning dialogues, while human teachers take on more complex support. Alternatively, AI could organize students' understanding levels and difficulties beforehand, allowing teachers to adjust their lessons accordingly. In such uses, AI becomes not a competitor to teachers but a line of support to enhance educational quality.

Distrust in Educational Business

Meanwhile, there is also a strong sense of distrust towards AI education on social media. Particularly when companies claim "AI can make education more efficient," many people become wary. Is AI being used not to enhance educational quality but to reduce labor costs? Is the work of teachers being undervalued? Are students being confined to corporate services? These are the concerns.

In discussions about language apps and learning apps, there are voices questioning whether the introduction of AI is for the benefit of learners or for monetization. While the phrase "individual optimization of learning" is appealing, when tied to advertising, monetization, and data collection, trust in education is undermined.

Therefore, when introducing AI teachers, it is necessary to ask not just "Is it effective?" but "For whose benefit is it being used?" Is it to deepen learners' understanding? To lighten teachers' burdens? Or to cut costs? Depending on the purpose, the meaning of the same AI technology in educational settings can change significantly.

Is "Human-likeness" Necessary for AI Teachers?

What is particularly intriguing about this study is that AI teachers achieved results without completely mimicking human teachers. This holds important implications for designing AI education.

Many AI education services strive to enhance human-likeness—natural voices, expressions, nods, and friendly characters. Of course, these elements might contribute to learners' sense of security. However, appearing human-like may not be the sole key to educational effectiveness.

What is required of AI is not a complete copy of human teachers but the educational design of AI's unique strengths. For instance, changing questions according to students' understanding, estimating stumbling points from past answers, providing explanations in multiple formats, confirming learning goals before lectures, and identifying gaps in understanding after lectures. These are areas where AI excels.

Human teachers have the ability to build relationships that only humans can. AI has flexibility and scalability unique to it. By combining these two, online learning might become stronger.

How Can It Be Used in Japanese Educational Settings?

In Japan, online classes, video materials, learning apps, and AI drills are already widespread. However, in many cases, the focus is on distributing materials or solving problems. This study highlighted the importance of "conversation" that occurs before the materials.

For example, before starting a video lesson, AI might say to students:
"What do you know about today's theme?"
"Are there any areas from the previous content that you're unsure about?"
"In this lecture, let's focus on this point first."
"After watching, try to answer this question."

Even such short dialogues might change learners' attention. Instead of simply playing videos, entering the lecture with a learning mindset. This concept can be applied not only to school education but also to corporate training, qualification learning, reskilling, and university on-demand classes.

Particularly in Japan, the overwork of teachers is a significant issue. If AI handles pre-dialogues and understanding checks, teachers can focus more on providing human support. If AI is used not to reduce teachers but to allow them to return to their core work, it is more likely to be accepted in the field.

Conclusion: The Value of AI Teachers Lies Not in "Replacing Teachers" but in "Being the Gateway to Learning"

This study deepens the discussion surrounding AI teachers.
AI has the potential to enhance learning outcomes similarly to human teachers under certain conditions. However, this does not mean human teachers become unnecessary. Human teachers and AI teachers have different strengths.

Human teachers support learners through gaze, expression, atmosphere, and trust relationships.
AI teachers support learners through individualization, repetition, immediate response, and cognitive organization.

The important question is not which will prevail, but in which situations each strength will be utilized.

The challenges of online education are not just about a lack of materials. They lie in learners being isolated, losing concentration, and not being prepared to learn. This study presented a simple solution to this problem: "short pre-class dialogues." It also showed that AI has potential as a facilitator of these dialogues.

AI teachers might become a new presence that organizes the gateway to learning, not a substitute for human teachers.
The future of education will be determined not by "human or AI," but by "how to combine humans and AI."

---

Source URL

Source for confirming the research overview, experimental design, main results, and researcher comments.
https://phys.org/news/2026-05-ai-human-teachers-pre-chat.html

Press release by the Hong Kong University of Science and Technology. Source for confirming experimental conditions, paper information, and educational significance.
https://www.eurekalert.org/news-releases/1127038

Publication information of related papers. For confirming paper title, journal, and DOI.
https://www.sciencedirect.com/science/article/pii/S0896627326002746

Reference for SNS reactions. Discussion on "Can AI replace teachers?" in the teacher community.
https://www.reddit.com/r/teaching/comments/1lmudnv/can_ai_replace_teachers/

Reference for SNS reactions. Discussion on "Can AI be a better teacher than a human?" in the AI community.
https://www.reddit.com/r/artificial/comments/1nv7qia/can_ai_be_a_better_teacher_than_a_human_my/

Reference for SNS reactions. Discussion on "Will ChatGPT ever replace teachers?" by general users.
https://www.reddit.com/r/ChatGPT/comments/1rjjnzv/do_you_think_chatgpt_will_ever_replace_teachers/

Reference for SNS reactions. Confirmation source for opposition and caution towards AI education and educational app companies.
https://www.reddit.com/r/antiwork/comments/1krroti/duolingo_ceo_says_ai_is_a_better_teacher_than/