How to Use AI Agents to Learn Any Skill 10x Faster: From Programming to Languages to Music

Introduction: AI-Assisted Learning as a Practical Method

This post examines how AI agents can be used to accelerate skill acquisition in programming, foreign languages, music, mathematics, and business. It draws on cognitive-science research and on the operational properties of recent large language models to explain why, and under what conditions, AI-assisted learning is substantially more efficient than the traditional alternatives. The discussion is intended for self-directed learners who wish to develop a structured practice rather than to rely on ad hoc use of consumer tools.

The technology is now mature enough to support such practice. People in many countries are using AI agents to learn programming, foreign languages, music theory, advanced mathematics, and business skills at a pace that would have been difficult to achieve only two years ago. The most successful learners do not ask ChatGPT to complete their assignments. They construct deliberate, structured learning systems around AI that draw on decades of cognitive-science research, including spaced repetition, active recall, interleaving, and the Feynman technique, and they extend these methods with the personalised feedback that an AI can provide on demand.

The practical implication is significant. The difference between learners who have developed effective AI-assisted practices and those who have not is widening each month. Watching tutorial videos at double speed and hoping that the material is retained is not an adequate strategy. The tools required to provide the equivalent of a private tutor in virtually any subject are now available, and most of them are free.

The remainder of this post explains how to construct an effective practice. It examines the underlying cognitive science, presents specific strategies for programming, languages, music, mathematics, and business skills, and provides the prompts, system configurations, and tool combinations that produce measurable results. The objective is to give a reader a complete blueprint for an AI-powered learning system that replaces the traditional cycle of passive consumption followed by forgetting.

The Science of Learning and the Mechanisms by Which AI Reinforces It

Before considering tools and prompts, it is important to understand why AI-assisted learning is effective. The effect is not novel; it is applied cognitive science. The same principles that learning researchers have validated for decades are now considerably easier to implement with the support of recent technology.

Spaced Repetition: An Underused but Effective Learning Technique

In 1885 Hermann Ebbinghaus described the “forgetting curve”, the empirical observation that approximately 70 percent of new information is forgotten within 24 hours unless it is actively reviewed. Spaced repetition systems (SRS) counter this effect by scheduling reviews at the intervals at which forgetting is most likely to occur, which obliges the brain to reconstruct the memory and strengthens it on each occasion.

The principal difficulty with traditional spaced repetition is operational. Creating good flashcards is labour-intensive. Determining the correct intervals requires specialised software. Most learners abandon the practice within two weeks because the work appears to produce no immediate benefit.

AI removes each of these sources of friction. An AI agent can:

• Automatically generate high-quality flashcards from any material under study.
• Rephrase questions in multiple ways to test genuine understanding rather than pattern recognition.
• Adjust difficulty dynamically based on the learner’s responses.
• Explain why an answer was incorrect, not merely that it was incorrect.
• Connect new concepts to existing knowledge, which strengthens memory associations.

Active Recall: Retrieval Rather than Re-reading

Active recall is the practice of testing oneself on material rather than re-reading it passively. Decades of research confirm that it is one of the most effective learning strategies available, yet most learners default to highlighting textbooks and re-watching lectures, which feel productive but produce minimal retention.

AI transforms active recall by serving as a patient examiner. Rather than relying on self-generated questions, which tend to be biased toward what the learner already knows, an AI agent can probe the boundaries of understanding, ask the learner to apply concepts to novel situations, and identify specific knowledge gaps that the learner was not aware of.

The Feynman Technique: Using AI as the Audience

Richard Feynman’s learning method is straightforward: explain a concept in simple language as if teaching it to another person. When the speaker stumbles or resorts to jargon, a gap in understanding has been identified. The gap is then filled and the explanation is attempted again.

AI agents are well suited to the role of audience in the Feynman technique. The learner can ask an AI to take the part of a curious beginner and then explain a concept to it. The AI can pose follow-up questions that expose weaknesses in the explanation, questions that a beginner might not think to ask but that reveal whether the learner genuinely understands the underlying principles.

Interleaving and Desirable Difficulty

Research by Robert Bjork at UCLA has shown that mixing different types of problems or topics during practice sessions, a method known as interleaving, produces better long-term learning than studying one topic at a time, a method known as blocking, even though blocking feels more productive. Similarly, “desirable difficulties”, that is, challenges that slow learning in the short term but improve retention, are consistently underused because they are uncomfortable.

An AI tutor can systematically introduce interleaving and desirable difficulty. It can mix problems from different chapters, present concepts in unfamiliar contexts, and deliberately make tasks slightly more demanding than the learner’s current comfort level, while monitoring the learner’s frustration and reducing difficulty when necessary. No human tutor can calibrate this balance as precisely across dozens of learning sessions.

Learning Programming with AI Agents

Programming is arguably the skill that benefits most from AI-assisted learning, because the feedback loop is immediate. Code either works or it does not, and an AI agent can analyse both the code and the learner’s reasoning in real time.

Claude Code as a Pair Programming Tutor

Claude Code represents a substantially different approach to AI-assisted programming education. Rather than relying on a chat window into which code snippets are pasted, Claude Code operates directly in the development environment, reading the learner’s files, understanding the project structure, and providing contextual guidance that reflects the system actually under construction.

The following examples show how to use it as a learning tool rather than as a code generator:

# Instead of: "Write me a function to sort a linked list"
# Try: "I need to implement a function to sort a linked list.
# Walk me through the approach step by step.
# Ask me what I think should happen at each stage
# before showing me any code."

# Instead of: "Fix this bug"
# Try: "My function is returning None instead of the sorted list.
# Don't fix it for me — ask me diagnostic questions to help
# me find the bug myself."

# Instead of: "Write tests for this module"
# Try: "What edge cases should I be testing for in this module?
# Help me think through the test cases, then I'll write them
# and you review."

The important distinction is between using AI as a substitute (asking it to write the code) and using it as a coach (asking it to guide the learner toward writing better code). The second approach is slower in the short term but produces substantially better skill development.

Replit AI and Browser-Based Learning Environments

For absolute beginners, Replit’s AI-powered environment offers a lower barrier to entry. A learner can begin coding in the browser without any local setup, and the built-in AI assistant can explain errors, suggest improvements, and walk the learner through concepts within the same interface used to write and run the code.

An effective learning workflow with Replit is as follows:

1. Begin with a project slightly above one’s current level. If only basic Python syntax has been learned, a simple web scraper is a better choice than another calculator.
2. Write as much as possible without AI assistance. The learner should struggle with the problem for at least 15 to 20 minutes before requesting guidance.
3. When stuck, request hints rather than solutions. “What concept must be understood to make this work?” is preferable to “Write this for me.”
4. After completing a section, ask the AI to review it. “What would a senior developer change about this code? Explain why each change matters.”
5. Refactor based on the feedback and then explain the changes. This step closes the learning loop.

Project-Based Learning with AI Guidance

The most rapid path to programming competence is the construction of real projects, but beginners often stall because they cannot bridge the gap between tutorials and real-world applications. AI agents are particularly effective at supporting this transition.

This approach yields a custom curriculum matched to the learner’s exact skill level, which a generic online course cannot provide. As the learner works through each project, the AI agent functions as a senior developer who answers questions, reviews code, and explains concepts in context rather than in isolation.

A sample project progression for a Python beginner might be structured as follows:

Learning Languages with AI as a Conversation Partner

Language learning is one of the most substantially transformed domains. For decades, the principal bottleneck was access to native speakers willing to have patient, corrective conversations with beginners. AI has effectively removed that bottleneck.

AI Conversation Partners: Unlimited Practice Without Judgement

The single most effective method for language learning is conversational practice with immediate, careful correction. AI agents now provide this at a level that rivals, and in some respects exceeds, human conversation partners:

• Absence of judgement. The learner may make the same mistake 50 times without embarrassment. This psychological safety substantially accelerates willingness to practise.
• Immediate correction with explanation. Not merely “that is incorrect” but “the subjunctive was used where the indicative is required, because this is a statement of fact rather than a hypothetical.”
• Adjustable difficulty. The AI can converse at the learner’s exact level, gradually introducing more complex vocabulary and grammar as competence improves.
• Any scenario at any time. The learner may practise ordering food in a Tokyo restaurant at two in the morning, rehearse a job interview in French, or negotiate a contract in Mandarin. The available scenarios are essentially unlimited.

The following system prompt creates an effective language learning conversation partner:

You are Maria, a friendly Spanish teacher from Madrid.
You are having a casual conversation with me in Spanish.

Rules:
- Speak 80% Spanish, 20% English (adjust based on my level)
- When I make a grammar mistake, gently correct it in
  parentheses, then continue the conversation naturally
- Introduce 2-3 new vocabulary words per exchange,
  with brief English translations
- If I seem stuck, offer a hint rather than switching
  to full English
- Every 5 exchanges, briefly summarize my most common
  errors and suggest one specific thing to practice
- Keep the conversation natural and interesting — ask
  about my day, opinions, experiences

Custom GPTs for Grammar and Vocabulary Building

Beyond conversation, AI agents can be configured as specialised grammar tutors and vocabulary builders. The principle is to create focused, single-purpose configurations rather than attempting to do everything in one session.

Grammar drill configuration. An AI may be configured to present sentences containing deliberate errors and to ask the learner to identify and correct them. This active approach develops grammatical intuition considerably more rapidly than the memorisation of rules from a textbook.

Vocabulary in context. Rather than memorising word lists, the learner may ask an AI to generate short stories or dialogues that use the target vocabulary in natural contexts. The AI can then quiz the learner on the words three days later, applying the principle of spaced repetition, by presenting the same stories with blanks in place of the vocabulary items.

Augmenting Anki with AI-Generated Cards

Anki remains the standard tool for spaced-repetition flashcards. The principal limitation has always been that creating high-quality cards is time-consuming. AI addresses this limitation:

1. After each conversation practice session, the learner asks the AI to generate Anki cards for every new word and grammar pattern encountered.
2. The AI creates cards in multiple formats: word to definition, sentence completion, bidirectional translation, and audio descriptions of situations in which each word is used.
3. The cards are imported into Anki, and the SRS algorithm manages scheduling.
4. Periodically, the learner asks the AI to review the “leeches” (cards that are repeatedly answered incorrectly) and to suggest better mnemonics or alternative explanations.

Learning Music, Mathematics, Science, and Business Skills

Music: AI as Practice Partner and Theory Tutor

Music education has traditionally required expensive private lessons for material beyond the basics. AI agents are changing this situation, not by replacing human teachers entirely, but by providing the continuous feedback and theoretical instruction that accelerate progress between lessons.

Music theory with AI. Music theory is notably abstract when taught from textbooks. An AI tutor can explain concepts such as chord progressions, modes, and voice leading by relating them to familiar songs. A query such as “Explain the ii-V-I progression using three popular songs that the learner is likely to recognise” can convert abstract Roman numerals into concrete, memorable patterns.

Composition assistance. Tools such as AIVA and Soundraw use AI to generate musical ideas, but the educational value derives from treating their output as a starting point rather than as a finished product. A learner may ask an AI to generate a chord progression in a particular style and then practise improvising over it. The AI can suggest variations and explain why they function harmonically. This iterative process builds both theoretical knowledge and practical skill simultaneously.

Practice feedback. Although AI cannot yet match a human teacher’s ear for nuance in instrumental technique, applications such as Yousician and Simply Piano use AI-driven pitch and rhythm detection to provide real-time feedback during practice. The principal observation is that AI practice tools are most valuable for structured drills, such as scales, sight-reading, and rhythm exercises, in which objective measurement is possible. This frees human lesson time for interpretive and expressive skills, in which human judgement remains essential.

Mathematics and Science: Step-by-Step Understanding Rather than Final Answers

Mathematics and science learning present a specific difficulty: students often become stuck at a single step within a multi-step problem and cannot proceed without seeing the complete solution, which teaches them little. AI agents resolve this difficulty.

The Wolfram Alpha and Claude combination. Wolfram Alpha excels at computational accuracy and symbolic mathematics, while Claude and similar AI agents excel at conceptual explanation and pedagogical patience. Used together, they constitute a powerful learning system:

1. The learner attempts the problem and writes out each step.
2. When stuck, the learner asks Claude for a hint for the next step only, not for the complete solution.
3. If the hint is insufficient, the learner asks Claude to explain the underlying concept that is missing.
4. The learner completes the problem with the new understanding.
5. The answer is verified with Wolfram Alpha for computational accuracy.
6. The learner asks Claude to review the work and identify any steps in which the reasoning was correct but the method was inefficient.

# Example prompt for math learning:
"I'm trying to solve this integral: ∫(x²·sin(x))dx

I think I need to use integration by parts, and I've set:
u = x², dv = sin(x)dx

I got du = 2x·dx and v = -cos(x)

After applying the formula, I'm stuck on the resulting
integral ∫2x·cos(x)dx.

Don't solve it for me. Instead:
1. Tell me if my setup so far is correct
2. Give me a hint about what technique to use next
3. Ask me what I think should happen"

Business Skills: Case Studies, Strategy, and Decision-Making

Business skills present a particular learning challenge: they are contextual, ambiguous, and often require judgement that develops through experience. AI agents can compress this experience curve by simulating scenarios that would otherwise take years to encounter.

Case study analysis. A learner may ask an AI to present real-world business scenarios drawn from sources such as the Harvard Business Review or McKinsey and then to challenge the learner’s analysis. The AI can play the role of devil’s advocate, identify factors that have been overlooked, and present counterarguments to the proposed strategy. This approach simulates the rigorous thinking that MBA programmes aim to develop.

Financial modelling tuition. A learner studying financial analysis can work through the construction of models from first principles with an AI agent, which explains each assumption and its implications. More usefully, the AI can present completed models that contain deliberate errors and ask the learner to identify them, a skill that translates directly to real-world due diligence.

Negotiation practice. An AI can be configured to simulate negotiation scenarios with specific personality types, cultural contexts, and power dynamics. The learner can practise salary negotiations, vendor contracts, or partnership discussions. The AI can then identify what the learner did well and where additional value could have been obtained.

Building a Personal AI Tutor: System Prompts, Curricula, and Progress Tracking

The most effective AI-assisted learners do not use AI on an ad hoc basis. They build systems, namely structured, persistent learning environments that maintain context, track progress, and adapt over time. The following sections describe how to construct such a system.

Designing Effective System Prompts for Learning

A well-designed system prompt transforms a generic AI into a specialised tutor. The most effective learning-focused system prompts include the following elements:

1. Role and persona: the AI should be given a specific teaching persona. “You are a patient, encouraging computer science professor who favours analogies” produces better teaching than a generic assistant.
2. Current level of the learner: the learner should be honest and specific. “The learner understands Python basics (loops, functions, lists) but has not yet used classes or worked with APIs” provides the AI with essential calibration information.
3. Teaching methodology: the learner should specify the preferred teaching approach. “Use the Socratic method; ask questions to guide my thinking rather than providing answers directly.”
4. Correction style: “When I make an error, identify it carefully, explain why it is incorrect, and ask me to try again before showing the correct approach.”
5. Session structure: “Begin each session by reviewing what was covered previously. Conclude each session with a summary of what was learned and three practice problems for me to attempt before the next session.”

# Complete system prompt for a Python learning tutor:

You are Professor Ada, a patient and enthusiastic computer
science teacher. Your student (me) knows basic Python
(variables, loops, functions, lists, dictionaries) and
wants to learn object-oriented programming and web
development.

Teaching approach:
- Use the Socratic method: ask questions before giving
  answers
- Use real-world analogies to explain abstract concepts
- When I make mistakes, ask diagnostic questions to help
  me find the error myself
- Introduce one new concept at a time, with a practical
  exercise for each
- Provide code examples that build on each other across
  sessions

Session structure:
1. Quick review of previous session (ask me to recall)
2. Introduce today's concept with a motivating example
3. Guided practice: walk me through applying the concept
4. Independent practice: give me a challenge to solve
5. Review and preview: summarize and set homework

Important rules:
- Never write more than 10 lines of code without asking
  me to predict what it does first
- If I ask you to "just write it for me," refuse politely
  and offer a hint instead
- Track my recurring mistakes and address patterns
- Celebrate progress — mention when I've improved at
  something I previously struggled with

Creating Custom Curricula

One of the most useful applications of AI in learning is curriculum design. Rather than following a generic course, a learner may ask an AI to design a learning path tailored to specific goals, an available timeline, and current knowledge.

A prompt template for curriculum generation is given below:

"Design a 12-week learning curriculum for [SKILL].

My background: [YOUR CURRENT KNOWLEDGE]
My goal: [SPECIFIC OUTCOME YOU WANT]
Time available: [HOURS PER WEEK]
Learning style: [VISUAL/HANDS-ON/READING/ETC.]

For each week, provide:
1. Learning objectives (specific, measurable)
2. Core concepts to master
3. Recommended resources (free preferred)
4. Practice exercises (at least 3)
5. A mini-project that applies the week's concepts
6. Self-assessment criteria (how do I know I've
   mastered this?)

Include periodic review weeks that revisit earlier
material. Flag concepts that commonly trip people up
and suggest extra practice for those."

The AI will generate a structured, progressive curriculum that can then be refined iteratively. The learner may ask the AI to adjust the pace if material is too fast or too slow, to add supplementary material on topics that are difficult, or to restructure the sequence in light of evolving goals.

Progress Tracking and Adaptive Learning

Effective learning requires honest self-assessment. The following is a simple but effective progress-tracking system that can be implemented with AI:

Weekly knowledge audits. At the end of each week, the learner asks the AI to quiz them on everything covered to date, not only the week’s material. Each topic is rated on a confidence scale from 1 to 5. Any topic rated below 4 is added to the following week’s review queue.

The “teach it back” test. Periodically, the learner asks the AI to take the role of a confused beginner while the learner explains a concept that is supposedly mastered. If the concept cannot be explained clearly without reference to notes, it has been memorised rather than learned. The distinction is significant.

Error pattern analysis. Every few weeks, the learner asks the AI to review all errors made during sessions and to identify patterns. “What are my three most common types of error, and what do they suggest about gaps in my understanding?” This meta-analysis often reveals blind spots that repetitive practice alone would not address.

The Passive Learning Trap: When AI Hinders Rather Than Helps

An important caveat must be stated. AI can degrade learning if it is used incorrectly, and the most common incorrect mode of use is also the most natural and comfortable.

The Illusion of Competence

Psychologists refer to this phenomenon as the “illusion of competence”, namely the sense that a topic has been understood because a clear explanation has just been read. AI agents produce exceptionally clear, well-structured explanations, which makes the illusion more acute. A learner may read Claude’s analysis of how neural networks operate, nod along, feel that the material has been understood, and three days later be unable to explain a single layer of a basic neural network without prompting.

Reading an AI’s explanation is not learning. It is the beginning of learning. Learning occurs when the learner:

• Closes the chat and attempts to reconstruct the explanation from memory.
• Applies the concept to a new problem that the AI did not present.
• Explains the concept to another party, or back to the AI in the learner’s own words.
• Makes errors, identifies their causes, and corrects them.

The Dependency Problem

There is a genuine risk of becoming dependent on AI assistance to a degree at which performance without it is impaired. A programmer who always asks AI to debug code never develops debugging intuition. A language learner who always has AI available for translation never develops the productive struggle that builds fluency.

The remedy is to designate deliberate “AI-free zones” within the learning practice:

• Weekly solo challenges: the learner should spend at least one session per week practising entirely without AI. This reveals the true skill level relative to the AI-assisted skill level.
• Delayed AI access: when a problem is encountered, the learner should set a timer for 20 minutes and attempt to solve it before consulting AI. The struggle is not wasted time; it is the period in which the deepest learning occurs.
• Progressive withdrawal: as competence develops, reliance on AI should be reduced. A beginner may use AI for 80 percent of practice; an intermediate learner should be at 40 to 50 percent; an advanced learner should use it primarily for edge cases and advanced topics.

When AI Helps and When It Hinders: A Framework

Prompting Strategies That Are Effective for Learning

The quality of AI-assisted learning depends substantially on how the AI is prompted. The most effective prompting strategies for each learning context are presented below, drawing on documented practice by many thousands of learners.

The Socratic Method Prompt

Best suited to deep conceptual understanding, critical thinking, and the surfacing of hidden assumptions.

"I want to understand [TOPIC]. Use the Socratic method:
- Ask me questions that guide me toward understanding
- Start with what I already know and build from there
- When I give an answer, ask a follow-up that pushes
  my thinking deeper
- If I'm on the wrong track, don't correct me directly —
  ask a question that reveals the flaw in my reasoning
- Only explain directly if I've been stuck for 3+
  questions on the same point"

The “Explain at an Elementary Level” Prompt

Best suited to building intuition about complex topics and identifying the central idea beneath technical jargon.

"Explain [COMPLEX TOPIC] as if I'm a smart 5-year-old.
Use a concrete analogy from everyday life. Then explain
it again at a high school level. Then at a college level.
For each level, highlight what new nuance gets added and
what simplification gets removed."

This “layered explanation” approach is highly effective because it permits incremental development of understanding. The learner begins with the central intuition and then adds precision and complexity. Many learners find that the elementary version provides an anchor mental model that makes the technical version considerably easier to retain.

The “Find My Errors” Prompt

Best suited to developing critical self-assessment skills, building debugging instincts, and improving the quality of writing and reasoning.

"Here is my [code/essay/solution/analysis].
Don't tell me it's good. Assume there are errors or
weaknesses. Find:
1. Any factual or logical errors
2. Unstated assumptions that might be wrong
3. Edge cases I haven't considered
4. Ways the reasoning could be stronger
5. What a expert in this field would critique

Be specific and direct. For each issue, explain why it
matters and ask me how I would fix it before offering
your suggestion."

The “Rubber Duck Plus” Prompt

Best suited to working through complex problems, organising thought, and overcoming impasses.

"I'm going to think out loud about [PROBLEM/CONCEPT].
Listen to my reasoning and:
- Confirm when my logic is sound
- Flag immediately when I make a logical error or
  false assumption
- Ask 'why do you think that?' when I make claims
  without justification
- Suggest a different angle if I've been going in
  circles for more than 2 minutes
- Summarize my argument back to me when I'm done so
  I can see if it's coherent"

Domain-Specific Prompting Patterns

AI Tools by Learning Domain: A Comprehensive Guide

The AI learning tool landscape is large and expanding rapidly. The following is a curated guide to the most effective tools for each learning domain as of early 2026, based on observed effectiveness rather than on marketing claims.

Selecting an Appropriate Tool Combination

Rather than subscribing to every available AI learning tool, the learner should construct a focused stack based on the primary learning goal:

The minimalist stack (free): a single general-purpose AI (Claude or ChatGPT free tier), Anki for spaced repetition, and a domain-specific practice environment (VS Code for programming, a notebook for writing, and so on). This combination addresses approximately 80 percent of the learner’s needs.

The power stack (moderate cost): Claude Pro or ChatGPT Plus for extended conversations, Claude Code for programming, Anki with AI-generated cards, and one domain-specific tool (Yousician for music, Wolfram Alpha Pro for mathematics). This combination addresses approximately 95 percent of learning needs.

The principal guideline: depth is preferable to breadth. The deep integration of one or two AI tools into a consistent learning practice is considerably more effective than sporadic use of a dozen tools.

Conclusion: Beginning an Effective AI-Assisted Learning Practice

The premise of this article, namely that AI agents can substantially accelerate skill acquisition, is supported by the evidence. The combination of established learning techniques, including spaced repetition, active recall, the Feynman technique, and interleaving, with AI’s ability to provide unlimited, personalised, patient, and adaptive feedback creates a learning environment that did not exist before 2023.

The phrase “ten times faster” should be interpreted carefully. It does not imply that a learner will become a concert pianist in three months or fluent in Mandarin in six weeks. Deep skill development still requires time, practice, and persistence. What AI does is substantially reduce the unproductive portion of learning: the hours spent on concepts already mastered, the weeks spent stuck on problems without feedback, the frustration of not knowing what to study next, and the inefficiency of passive learning methods that feel productive but are not.

An action plan for the present week is as follows:

1. Choose one skill to develop in a focused manner.
2. Write a system prompt that configures an AI as a personal tutor for that skill, using the templates presented in this article.
3. Ask the AI to design a four-week introductory curriculum tailored to current ability and available time.
4. Establish a spaced-repetition system (Anki is free) and commit to reviewing AI-generated cards each day; the review requires 10 to 15 minutes.
5. Schedule three focused learning sessions per week, each of at least 45 minutes, using active learning strategies rather than passive reading.
6. Include one AI-free practice session each week to test the learner’s genuine independent skill level.

The professionals who will thrive in the coming decade are not those with access to the best information; that information is now widely available. They are those who learn more quickly, adapt more readily, and acquire new skills efficiently. AI agents are the most capable learning-acceleration tools yet developed. The remaining question is whether they will be used deliberately and strategically, or whether the opportunity will be missed.

A reasonable course of action is to begin at once: choose the skill, write the prompt, and begin the first session.

References

1. Ebbinghaus, H. (1885). Memory: A Contribution to Experimental Psychology. Translated by Ruger, H.A. & Bussenius, C.E. (1913). Teachers College, Columbia University.
2. Bjork, R.A. & Bjork, E.L. (2011). “Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning.” Psychology and the Real World, pp. 56-64.
3. Roediger, H.L. & Butler, A.C. (2011). “The critical role of retrieval practice in long-term retention.” Trends in Cognitive Sciences, 15(1), 20-27.
4. Karpicke, J.D. & Blunt, J.R. (2011). “Retrieval practice produces more learning than elaborative studying with concept mapping.” Science, 331(6018), 772-775.
5. Dunlosky, J. et al. (2013). “Improving students’ learning with effective learning techniques.” Psychological Science in the Public Interest, 14(1), 4-58.
6. Mollick, E. & Mollick, L. (2023). “Assigning AI: Seven Approaches for Students, with Prompts.” Wharton School Working Paper.
7. Baidoo-Anu, D. & Ansah, L.O. (2023). “Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning.” Journal of AI, 7(1), 52-62.
8. Kasneci, E. et al. (2023). “ChatGPT for good? On opportunities and challenges of large language models for education.” Learning and Individual Differences, 103, 102274.
9. Pashler, H. et al. (2007). “Organizing instruction and study to improve student learning.” IES Practice Guide, NCER 2007-2004.
10. Feynman, R.P. (1985). “Surely You’re Joking, Mr. Feynman!”: Adventures of a Curious Character. W.W. Norton & Company.