Learning Generative AI #9: How Generative AI Can Provide Tutoring for Everyone

The dream for education has been a tutor for every student and now generative AI creates that possibility.

I recently published a new book: “How Generative AI Can Transform Education.” It is well-received by educators who are already beginning to apply its lessons to improve their productivity and reach.

The book is currently available from Amazon in printed, Kindle, and Audible formats below:

How Generative AI Can Transform Education

In this article, I provide an excerpt from this new book, a chapter that discusses how generative AI provides universal tutoring for all. I focus on Khanmigo, as the leading system for generative AI tutoring.

If you want to read my articles on emerging technologies, click this link to access an extensive and interesting library.

AI as Direct Tutoring Support for Students

Generative AI can serve as a direct tutoring resource for students who lack access to personalized academic support. Although it may not perfectly replace human tutoring, AI-based tutoring systems offer unprecedented opportunities for large-scale learning assistance.

The most significant advantage of AI tutoring is its accessibility. Unlike human tutors, who are constrained by geography, scheduling, and capacity limitations, AI tutoring systems can be available to students anywhere, anytime, without requiring appointments or transportation barriers. This "always on" availability means students can receive support precisely when needed, while working on homework after school, preparing for a test on the weekend, or reviewing complex concepts before class.

For students in resource-constrained environments, this accessibility represents a transformative opportunity. A student without educated family members to assist with homework, living in a community without tutoring centers, or unable to afford professional tutoring services can now access personalized academic support.

Modern generative AI systems demonstrate several capabilities that make them particularly valuable for tutoring applications.

• First, they can engage in natural language dialogue that approximates human tutorial conversations, allowing students to ask questions, receive explanations, and participate in Socratic exchanges without rigid interfaces or predefined response limitations.

• Second, these systems can provide multiple representations and explanations of challenging concepts, adapting to different learning preferences and prior knowledge. If a student doesn't understand an algebraic explanation of a concept, the AI can offer a visual representation, a real-world application, or an analogy to more familiar ideas.

• Third, advanced generative AI can implement scaffolding approaches similar to those used by human tutors, providing varying levels of support based on student responses. Rather than simply solving problems or providing answers, these systems can offer hints, ask guiding questions, highlight errors for correction, or model expert thinking processes.

• Fourth, AI tutoring systems can maintain detailed records of student interactions, identifying patterns in understanding and tailoring future support accordingly.

The limitations of current AI tutoring systems must also be acknowledged. These systems lack the intuitive understanding of student emotional states, the ability to notice subtle cues of confusion or disengagement, and the motivational impact of a caring human relationship. Despite these limitations, the accessibility and scalability of AI tutoring present unprecedented opportunities to provide personalized academic support to students who would otherwise receive none.

Khan Academy's Khanmigo AI Tutoring

Among the most promising implementations of AI tutoring is Khan Academy's Khanmigo, an AI-powered educational assistant designed to provide personalized tutoring across various subjects and grade levels. Launched as generative AI capabilities matured, Khanmigo represents a significant evolution from Khan Academy's original model of video lessons and practice problems to interactive, adaptive tutoring experiences.

Khanmigo builds upon Khan Academy's established educational framework while adding the dynamic responsiveness of AI tutoring. The system combines several key elements: a conversational interface that enables natural dialogue between the student and tutor, integration with Khan Academy's existing content library and knowledge map, and the application of research-based tutoring approaches, such as Socratic questioning and graduated hints.

The fundamental design philosophy behind Khanmigo is rooted in key principles from effective human tutoring practices. Rather than simply providing answers, the system engages students in guided discovery, helping them develop deeper conceptual understanding and metacognitive strategies. When a student asks for help with a math problem, for instance, Khanmigo typically responds with questions about the student's current knowledge, offers strategic hints that point toward solution pathways without revealing answers, and provides encouragement that normalizes productive struggle as part of the learning process.

Integrating Khan Academy's existing platform creates a seamless learning ecosystem that allows students to transition between structured lessons, practice problems, and conversational tutoring as needed. This integrated approach enables Khanmigo to reference specific videos or practice activities that might help a student understand a challenging concept, creating a more comprehensive learning experience than standalone tutoring.

For teachers and parents, Khanmigo provides visibility into student interactions through summary reports highlighting topics covered, concepts requiring extensive support, and patterns in student questions or misconceptions. This transparency helps human educators understand where students might need additional attention and prevent AI tutoring from becoming a "black box" separate from other educational supports.

Khan Academy has positioned Khanmigo as a complement to human teaching, rather than a replacement, emphasizing how the system can provide additional practice, reinforcement, and individualized support while teachers focus on complex educational aspects that require human judgment and relationship-building. This complementary positioning acknowledges both the capabilities and limitations of current AI tutoring systems.

Early research on Khanmigo suggests promising results for student learning and engagement. Students using the system demonstrate increased persistence in working through challenging problems, higher completion rates for assigned work, and improved performance on subsequent assessments. Particularly encouraging are reports of positive impacts for students previously reluctant to seek help from human teachers or tutors, suggesting that the judgment-free nature of AI interaction may create psychological safety for some learners.

Khanmigo Examples

We can examine specific examples across different grade levels and subject areas to understand the practical applications and impact of AI tutoring. These examples illustrate how Khanmigo and similar systems adapt their approach based on student age, subject matter, and specific learning needs.

Example #1: Elementary mathematics problem-solving

Ten-year-old Aaliyah is working on a challenging word problem for her fourth-grade math homework: "A farmer has 48 feet of fencing to create a rectangular animal pen. What dimensions would give the largest possible area for the pen?"

Unsure how to begin, Aaliyah turns to Khanmigo for help. Rather than simply providing the answer, the AI tutor engages her in a guided exploration:

Khanmigo: "This is an interesting problem about maximizing area! Before we dive in, do you know what we're trying to find here?"

Aaliyah: "We need to find the length and width that give the biggest area."

Khanmigo: "That's right! And we have a constraint—a limitation—of 48 feet of fencing. Let's think about what we know about rectangles. If we call the length L and the width W, how would we calculate the perimeter of the rectangle?"

Aaliyah: "Perimeter is L + W + L + W."

Khanmigo: "Good! And we can simplify that to 2L + 2W, right? Since we have 48 feet of fencing, what equation can we write about the perimeter?"

Aaliyah: "2L + 2W = 48"

Khanmigo: "Perfect! Now, how do we calculate the area of a rectangle?"

Aaliyah: "Area is length times width, so A = L × W."

Khanmigo: "Exactly! So, we want to find values for L and W that make L × W as large as possible while keeping 2L + 2W = 48. Let's solve for one variable. If 2L + 2W = 48, then L + W = 24, which means L = 24 - W. Can you substitute this into the area formula?"

Aaliyah: "A = L × W, and L = 24 - W, so A = (24 - W) × W."

Khanmigo: "Great work! Now we have the area in terms of just W. Let's try some values to see which gives the largest area. What if W = 6?"

Through this guided exploration, Khanmigo helps Aaliyah discover that a square with sides of 12 feet produces the maximum area of 144 square feet. Rather than giving her the answer, the AI tutor supports her through each step of the problem-solving process, using questioning techniques that help her apply existing knowledge to this new challenge. This approach develops both her specific understanding of the relationship between perimeter and area and her broader mathematical problem-solving strategies.

Example #2: Middle school science concept clarification

Thirteen-year-old Miguel is studying cell biology in his seventh-grade science class. While reviewing his notes on cell structure and function, he realizes he's confused about the differences between plant and animal cells. He turns to Khanmigo for clarification:

Miguel: "I'm getting confused about plant and animal cells. Can you help me understand the differences?"

Khanmigo: "I'd be happy to help you with that, Miguel! Let's start with what you already know. Could you tell me one or two differences between plant and animal cells that you remember from class?"

Miguel: "Well, I know plant cells have cell walls and animal cells don't. And I think plant cells have chloroplasts?"

Khanmigo: "That's a great start! You're absolutely right about both of those key differences. The cell wall gives plant cells their rigid structure, and chloroplasts are where photosynthesis happens, allowing plants to make their own food using sunlight. Let me ask you this: have you learned about vacuoles in both cell types?"

Miguel: "Yes, but I don't remember exactly how they're different."

Khanmigo: "Good question! Both plant and animal cells have vacuoles, but they differ in size and function. Plant cells typically have one large central vacuole that can take up to 90% of the cell's volume. It stores water, nutrients, and waste products, and helps maintain the cell's rigidity through turgor pressure.

Animal cells, on the other hand, have multiple smaller vacuoles that primarily store waste products, ions, and water.

Would it help to see a side-by-side comparison of the major differences?"

Miguel: "Yes, that would be great."

Khanmigo then provides a structured comparison of key differences, including cell walls, chloroplasts, vacuoles, shape, size, and energy storage. Throughout the conversation, the AI tutor checks Miguel's understanding, builds on his existing knowledge, and introduces new information in manageable chunks. When Miguel asks follow-up questions about specific organelles, Khanmigo provides more detailed explanations while maintaining an appropriate level of complexity for seventh-grade biology.

This tutoring approach helps Miguel develop a clearer conceptual understanding of cellular structures while also modeling how to compare and contrast related biological concepts—a key science literacy skill he can apply to other topics.

Example #3: High school English literature analysis

Seventeen-year-old Sophia is working on an essay analyzing symbolism in F. Scott Fitzgerald's "The Great Gatsby" for her 11th-grade English class. She understands the basic concept of symbolism but struggles with developing a deeper analysis. She turns to Khanmigo for guidance:

Sophia: "I need to write about symbolism in 'The Great Gatsby,' but I'm not sure how to go beyond just identifying symbols. Can you help?"

Khanmigo: "Absolutely, Sophia. Moving from identification to analysis is an important skill in literary criticism. Let's start with a symbol you've identified in the novel. What's one symbol you're planning to discuss in your essay?"

Sophia: "I'm writing about the green light at the end of Daisy's dock."

Khanmigo: "The green light is a central symbol in the novel. You've successfully identified it, which is the first step. Now, to deepen your analysis, let's explore several dimensions of this symbol:

1. Context: When does the symbol appear in the novel, and how does its significance change over time?

2. Multiple meanings: What different things might the green light represent? For Gatsby specifically and in the broader themes of the novel?

— Connection to character development: How does this symbol reflect or influence Gatsby's character?

— Connection to themes: How does this symbol connect to more prominent themes like the American Dream, wealth, or illusion versus reality?

Which of these dimensions would you like to explore first?"

Sophia: "I think the multiple meanings. I know it represents Gatsby's hope of being with Daisy, but I'm unsure what else it could symbolize."

Khanmigo guides Sophia through a deeper exploration of the green light symbol, helping her understand how it represents not just Gatsby's desire for Daisy but also the broader American Dream, the illusion of progress, and the tension between past and future. The AI tutor suggests specific passages where Fitzgerald develops this symbolism and models how to connect textual evidence to interpretive claims. (I recently reread The Great Gatsby and learned from this example.)

Throughout the conversation, Khanmigo uses a Socratic approach, asking questions that prompt Sophia to develop her own insights rather than simply providing analysis for her to repeat. This process helps her develop the critical thinking skills essential for sophisticated literary analysis while still providing enough structure to keep her from feeling overwhelmed by the complexity of the task.

Example #4: Homework help across multiple subjects

Fourteen-year-old Jamal is working through a night of homework assignments across several subjects. His parents both work evening shifts, so he doesn't have anyone at home to ask for help when he gets stuck. Khanmigo serves as his homework support across multiple subjects:

For his algebra homework, Jamal encounters a problem involving systems of equations that he can't quite solve. Khanmigo walks him through a step-by-step approach, first confirming which method (substitution, elimination, or graphing) he's trying to use, then offering targeted hints at each stage of the solution process rather than simply providing the answer.

When he moves to his geography assignment on climate zones, Jamal needs to understand the factors that create different climate regions. Khanmigo helps him understand the relationships between latitude, ocean currents, elevation, and prevailing winds, using analogies and simplified explanations that make these complex interactions more accessible.

For his English assignment, Jamal needs to revise a paragraph that his teacher marked as having "unclear organization." Rather than rewriting it for him, Khanmigo asks questions about his central idea, supporting points, and the logical connection between sentences. Through this guided reflection, Jamal identifies where his paragraph structure breaks down and learns strategies for creating more coherent writing.

Throughout the evening, Khanmigo adapts its approach based on the subject matter and Jamal's specific needs in each area. The AI tutor maintains a balance between providing necessary support and promoting independent learning, offering more substantial scaffolding when Jamal is genuinely stuck while encouraging him to work through problems he can solve with minimal guidance.

For Jamal, Khanmigo provides not just academic support but also the confidence that comes from knowing help is available when needed. This security allows him to approach challenging assignments with less anxiety and more willingness to persist through difficulties, a crucial aspect of developing academic resilience.

Example #5: Supporting English language learners through multilingual tutoring

Twelve-year-old Lin moved to the United States from China six months ago and is still developing English proficiency while trying to keep up with seventh-grade academic content. Her family speaks Mandarin at home, and while the school provides some ESL support, she often struggles with homework assignments.

Khanmigo's multilingual capabilities allow it to provide tutoring support that bridges Lin's languages while developing her English academic vocabulary. When Lin encounters difficulty understanding a science assignment about the water cycle, she can begin the interaction in Mandarin:

Lin: "我不明白水循环中的'condensation'是什么意思。" (I don't understand what 'condensation' means in the water cycle.)

Khanmigo: "凝结 (condensation) 是指当水蒸气冷却并变回液态水时发生的过程。例如，当温暖的水蒸气接触到冷窗户时，它会凝结成水滴。在水循环中，凝结发生在云形成时。你想要用英语或中文继续讨论这个概念吗?" (Condensation is the process that occurs when water vapor cools and turns back into liquid water. For example, when warm water vapor contacts a cold window, it condenses into water droplets. In the water cycle, condensation occurs when clouds form. Would you like to continue discussing this concept in English or Chinese?)

Lin chooses to switch to English to develop her academic vocabulary, and Khanmigo adjusts its language level accordingly, using simpler sentence structures and providing more visual explanations while still covering the key scientific concepts. Throughout the tutoring session, Lin can switch between languages when needed, gradually building her confidence in discussing scientific concepts in English.

This multilingual support ensures that Lin's content learning doesn't stall while she develops English proficiency, maintaining her academic progress across subjects during the language acquisition process. For the millions of students in similar situations, this linguistically responsive tutoring provides crucial academic support that might otherwise be unavailable.

Benefits of AI Tutoring:
Toward a "Tutor for All"

The examples above illustrate the transformative potential of AI tutoring to address longstanding educational inequities. By analyzing these applications, we can identify several key benefits that contribute to the vision of a "tutor for all"—personalized academic support available to every student regardless of geographical, financial, or social circumstances.

Universal Accessibility and Scale

The most significant benefit is unprecedented accessibility on a large scale. Unlike human tutoring, which is constrained by logistics, location, scheduling, and cost, AI tutoring can be available to millions of students simultaneously, 24 hours a day, without requiring appointments or transportation barriers. This accessibility has profound equity implications, providing personalized academic support to students in rural areas without tutoring centers, low-income communities where private tutoring is financially out of reach, and home environments where family members may lack the educational background to provide homework help.

Cost-Effectiveness for School Systems

For schools and districts with limited resources, AI tutoring offers a cost-effective approach to providing personalized support. While high-quality human tutoring programs typically cost $1,500 to $4,000 per student annually, AI tutoring systems can operate at a fraction of this cost when implemented on a large scale. This economic reality means schools can potentially provide personalized tutoring support to all students rather than only those identified as most at-risk, an approach that prevents the development of learning gaps rather than attempting to remediate them after they occur.

Khanmigo’s pricing is currently:

• The cost for a Khanmigo District Partnership is $15 per student per year, with a minimum of 250 licenses, and is only applicable to U.S.-based districts.

• For Individual Learners and Families, the cost is $4 per month or $44 per year for parents and learners.

• For Teachers, Khanmigo is currently free.

Psychological Safety and Reduced Academic Anxiety

The judgment-free nature of AI tutoring creates psychological safety that benefits many learners, particularly those who have experienced academic anxiety or stereotype threat in traditional educational settings. Students can ask "basic" questions without embarrassment, make multiple attempts without fear of trying an adult's patience, and work at their own pace without social comparison. This emotional safety increases willingness to seek help, a crucial academic behavior that predicts long-term educational success but is unevenly distributed among students.

Personalization

The data-informed personalization of advanced AI tutoring systems creates increasingly tailored learning experiences over time. As students interact with the system, it builds comprehensive models of their knowledge, misconceptions, learning preferences, and growth patterns. This continuous assessment enables increasingly precise targeting of explanations, examples, and practice opportunities to each student's specific needs, a level of personalization difficult to achieve even in high-quality human tutoring without extensive documentation and analysis.

Benefits to Teachers: Complementary Support and Time Optimization

For teachers, AI tutoring provides valuable complementary support without replacing their essential role. When students have access to AI tutoring for foundational skills practice, procedural fluency development, and basic concept clarification, teachers can focus more classroom time on complex thinking, collaborative learning, and the social-emotional aspects of education that benefit most from human facilitation. This complementary relationship enhances overall educational effectiveness while acknowledging the irreplaceable value of human teachers.

The practical impact on teacher workload and effectiveness is substantial. When AI systems handle routine homework help and basic explanations, teachers can redirect their limited time and energy toward higher-leverage instructional activities, such as designing rich learning experiences, facilitating meaningful discussions, providing nuanced feedback on complex work, and building relationships with students. The AI system serves as a force multiplier for teacher impact rather than a replacement, allowing educators to focus on the aspects of teaching that most benefit from human creativity, judgment, and empathy.

Differentiated Support Based on Student Ability

AI tutoring systems excel at providing appropriately differentiated support based on individual student abilities, learning speeds, and background knowledge. For struggling students, these systems can break concepts into smaller steps, provide additional scaffolding, offer multiple explanations of challenging ideas, and create extensive practice opportunities with immediate feedback. This intensive support helps prevent the cumulative knowledge gaps that often develop when students fail to master foundational concepts before moving to more advanced material.

• For on-level students, AI tutors can provide balanced support that reinforces classroom learning, helps clarify minor confusions before they become significant obstacles, and offers appropriate challenge to maintain engagement.

• For advanced learners, AI tutoring can provide acceleration and enrichment often unavailable in standard classroom settings.

Supporting English Language Learners

AI tutoring systems offer compelling benefits for English Language Learners (ELLs), who often face compounded challenges in educational settings. These students must simultaneously develop English proficiency while mastering academic content, a dual cognitive load that can be overwhelming without appropriate support. AI tutors with multilingual capabilities provide a uniquely effective resource for these learners in several ways.

AI tutoring systems can provide academic support in both the student's native language and English, creating a bridge that maintains content learning while language acquisition progresses. This capability addresses a critical challenge in ELL education: the learning gaps that develop when students cannot access content instruction due to language barriers. A middle school ELL student who understands algebra conceptually but lacks the English vocabulary to follow classroom instruction can work with an AI tutor in their native language, ensuring mathematical development continues uninterrupted while English proficiency develops.

The cumulative impact of these benefits can be transformative for English Language Learner (ELL) students, who represent one of the fastest-growing and most academically vulnerable populations in U.S. schools. By providing linguistically responsive academic support that maintains high expectations for content learning, AI tutoring systems help ensure that language status doesn't become destiny in determining educational outcomes. This application represents one of the most powerful equity implications of AI tutoring technology, addressing a persistent challenge that schools have struggled to solve through traditional means.

Conclusion: A Tutor for All

The emergence of generative AI as a tutoring partner represents a potentially transformative development in addressing one of education's most persistent challenges: providing personalized learning support to all students, not just those with privileged access to human tutors. While not equivalent to high-quality human tutoring in all dimensions, AI tutoring systems like Khanmigo offer unprecedented opportunities to scale personalized academic support while complementing rather than replacing the essential work of human educators.

If you are interested in my previous article on generative AI for education, you can use this link:

If you are not currently subscribed to this newsletter and want to follow generative AI and other technologies, please subscribe here. It’s a free subscription and will remain so.

You can also share this article with someone who you think might be interested.

Thank you,

Michael