Introducing study mode in ChatGPT

今天，我们在 ChatGPT 中推出了“学习模式”——这是一种帮助你逐步解决问题的学习体验，而不仅仅是直接给出答案。从今天起，登录用户（包括免费版、Plus、Pro、团队版）均可使用该功能，ChatGPT Edu 版本将在接下来的几周内上线。

ChatGPT 正成为全球最广泛使用的学习工具之一。学生们借助它解决难题、备考和探索新概念。但在教育领域的应用也带来了一个重要问题：如何确保它真正支持学习，而不仅仅是提供答案而不帮助学生理解？

我们设计了学习模式来回应这个问题。当学生使用学习模式时，系统会通过引导性问题，根据他们的目标和技能水平调整回答，帮助他们建立更深层次的理解。学习模式旨在互动且引人入胜，帮助学生真正学到东西，而不仅仅是完成任务。

学习模式背后是我们与教师、科学家和教育学专家合作编写的定制系统指令，体现了支持深度学习的核心行为，包括：

• 鼓励主动参与
• 管理认知负荷
• 主动培养元认知和自我反思
• 激发好奇心
• 提供可操作且支持性的反馈

这些行为基于长期的学习科学研究，塑造了学习模式对学生的响应方式。

Common Sense Media AI项目高级总监 Robbie Torney 说：“学习模式不是替学生完成作业，而是鼓励他们批判性地思考学习内容。这类功能是有效利用 AI 促进学习的积极一步。即使在 AI 时代，最好的学习仍然发生在学生对课程内容充满热情并积极参与时。”

• 互动提示：结合苏格拉底式提问、提示和自我反思，引导理解，促进主动学习，而非直接给出答案。
• 分层回答：信息分成易于理解的部分，突出主题间的关键联系，提供恰当的上下文，减少复杂话题的负担。
• 个性化支持：根据用户的技能水平和之前对话中的记忆，调整课程难度。
• 知识检测：通过测验和开放式问题，配合个性化反馈，帮助跟踪进度，促进知识保持和应用。
• 灵活切换：可在对话中随时开启或关闭学习模式，灵活适应不同学习目标。

学习模式主要面向大学生。早期测试中学生反馈：

• “我觉得它就像一个全天候、无所不知的‘答疑时间’。”——Noah Campbell，大学生
• “学习模式很好地将复杂内容拆解成清晰、节奏适中的讲解。”——Caleb Masi，大学生
• “我用学习模式辅导自己学习正弦位置编码这个多次尝试未果的概念。它就像一个永远不会厌烦我提问的导师。经过3小时的学习，我终于理解得足够好，充满信心。”——Maggie Wang，大学生

用户希望学习博弈论的广泛内容及其日常生活中的应用，学习模式会制定分阶段的高层次学习计划，从基础概念到实际应用，逐步引导学习：

• 第一阶段：核心基础（博弈论定义、游戏类型、基本概念）
• 第二阶段：经典游戏与概念（囚徒困境、混合策略均衡等）
• 第三阶段：模型深化（重复博弈、扩展式博弈、贝叶斯博弈）
• 第四阶段：现实生活相关性（机制设计、行为博弈论、日常应用）
• 第五阶段：高级与选修内容（进化博弈论、算法博弈论、个人相关）

学习过程中穿插引导示例、思维实验、现实情境映射和快速检测，确保理解。

学习模式特别适合作业辅导、考试准备和新知识学习。只需在 ChatGPT 工具中选择“学习与学习”模式，输入问题即可开始。

学习模式是我们提升 ChatGPT 学习体验的第一步。目前它基于定制系统指令，便于快速根据学生反馈迭代改进，尽管可能导致对话中表现不一。未来，我们计划将此行为直接训练进主模型。

我们还在探索更多功能，如：

• 更清晰的复杂概念可视化
• 跨对话的目标设定与进度跟踪
• 更深度的个性化，针对学生技能和目标

教育领域关于 AI 最佳应用的研究仍在发展中。OpenAI 通过 NextGenAI 计划和与斯坦福大学 SCALE Initiative 合作，支持相关研究，研究 AI 工具对学习效果的影响。

我们将持续进行长期研究，发布关于模型设计与认知关系的深入分析，基于洞见优化产品体验，并与教育生态系统合作，确保 AI 造福全球学习者。

Today we’re introducing study mode in ChatGPT—a learning experience that helps you work through problems step by step instead of just getting an answer. Starting today, it’s available to logged in users on Free, Plus, Pro, Team, with availability in ChatGPT Edu coming in the next few weeks.

ChatGPT is becoming one of the most widely used learning tools in the world. Students turn to it to work through challenging homework problems, prepare for exams, and explore new concepts. But its use in education has also raised an important question: how do we ensure it is used to support real learning, and doesn’t just offer solutions without helping students make sense of them?

We’ve built study mode to help answer this question. When students engage with study mode, they’re met with guiding questions that calibrate responses to their objective and skill level to help them build deeper understanding. Study mode is designed to be engaging and interactive, and to help students learn something—not just finish something.  

How study mode was built

Under the hood, study mode is powered by custom system instructions we’ve written in collaboration with teachers, scientists, and pedagogy experts to reflect a core set of behaviors that support deeper learning including: encouraging active participation, managing cognitive load, proactively developing metacognition and self reflection, fostering curiosity, and providing actionable and supportive feedback. These behaviors are based on longstanding research in learning science and shape how study mode responds to students.

“Instead of doing the work for them, study mode encourages students to think critically about their learning. Features like these are a positive step toward effective AI use for learning. Even in the AI era, the best learning still happens when students are excited about and actively engaging with the lesson material.” —Robbie Torney, Senior Director of AI Programs at Common Sense Media.

Key features

Interactive prompts: Combines Socratic questioning, hints, and self-reflection prompts to guide understanding and promote active learning, instead of providing answers outright.

Scaffolded responses: Information is organized into easy-to-follow sections that highlight the key connections between topics, keeping information engaging with just the right amount of context and reducing overwhelm for complex topics.

Personalized support: Lessons are tailored to the right level for the user, based on questions that assess skill level and memory from previous chats.

Knowledge checks: Quizzes and open-ended questions, along with personalized feedback to track progress, support knowledge retention and the ability to apply that knowledge in new contexts.

Flexibility: Easily toggle study mode on and off during a conversation, giving you the flexibility to adapt to your learning goals in each conversation.

How students use study mode

Study mode was built with college students in mind. Some feedback we heard from students during early testing:

“The best way I’d describe it is a live, 24/7, all-knowing “office hours.” —Noah Campbell, college student 

“Study mode did a great job breaking down dense material into clear, well-paced explanations.” —Caleb Masi, college student 

“I put study mode to the test to tutor me on a concept I have attempted to learn many times before: sinusoidal positional encodings. It was like a tutor who doesn’t get tired of my questions. After a 3-hour working session, I finally understood it well enough to feel confident.” —Maggie Wang, college student

Ways to use study mode inspired by college students from the ChatGPT lab

I want to learn about Game Theory, specifically the broad spectrum that the field entails, then also the ways in which you think it'd be useful for me to understand in my daily life. I want you to of course follow my curiousity, but mostly you will be teaching me about it, and keeping a high level plan to iterate through so I can cover the full scope here. I will ask questions when I am curious, but be deadset on quickly educating me on this.

Perfect. Here’s the high-level plan I’ll drive us through—think of it as a layered roadmap from foundational ideas to practical intuition:

🔹 Phase 1: Core Foundations

1. What is Game Theory, really?

• Strategic interaction, rational agents, assumptions.

2. Types of Games

• Cooperative vs non-cooperative
• Simultaneous vs sequential
• Zero-sum vs non-zero-sum

3. Basic Concepts

• Players, strategies, payoffs, information
• Dominant strategies
• Nash equilibrium (core to nearly everything)

🔸 Phase 2: Classic Games & Concepts

4. Famous Games

• Prisoner’s Dilemma
• Chicken
• Battle of the Sexes
• Stag Hunt
• Matching Pennies

5. Mixed Strategy Equilibria

• When no pure strategy works
• Randomization and expected payoffs

🔹 Phase 3: Deepening the Model

6. Repeated Games & Folk Theorems

• Cooperation emerging over time
• Tit for Tat, Grim Trigger, etc.

7. Extensive-Form Games

• Game trees, backward induction
• Subgame Perfect Equilibria

8. Bayesian Games

• Games with incomplete information
• Beliefs, types, and expected utility

🔸 Phase 4: Real-Life Relevance

9. Mechanism Design / Reverse Game Theory

• Voting systems, auctions, contracts

10. Behavioral Game Theory

• Bounded rationality, fairness, heuristics

11. Daily Life Applications

• Negotiation, dating, politics, AI systems, etc.

🔹 Phase 5: Advanced & Optional

12. Evolutionary Game Theory

• Strategies that reproduce over time

13. Algorithmic Game Theory

• Network effects, internet auctions

14. Your world: How it matters to Parker

• Systems design, incentives, power structures

Throughout this, I’ll mix in:

• 🔄 Guided examples
• 🧠 Thought experiments
• 🎭 Real-life situations to map theory onto
• 🧩 Quick checks to make sure ideas are sticking

Let’s kick off Phase 1.
First quick one: In your own words, what do you think Game Theory is about? No need to be polished—just get it out.

I think Game Theory is sort of this meta aspect framing of modern society in life where we look at people in situations and complex systems and we try to analyze how people in them make decisions, sometimes why they make it, and how this affects the situation/system

Nice—you're already circling the core.

Let’s sharpen it a bit together. Game Theory is fundamentally about strategic decision-making—situations where the outcome for each person depends not just on what they do, but on what others do too.

So real quick:

If you’re making a decision and it depends entirely on nature or chance—like a dice roll—is that Game Theory?

And then:

If you’re making a decision and it depends on what someone else does—like whether they cooperate or betray you—is that Game Theory?

Walk me through your thinking on both.

No random chance isn't game theory, though agents in game theory may end up having things depend on random chance. But if youre decision matters then of course it is. Also please remember you should be teaching this to me, via paragraphs of information and moving through a high level plan. If I want to do interactions and questions, I will ask. Continue

Got it—I'll stay in teaching mode and drive forward through the roadmap unless you tap in.

📍Phase 1: Core Foundations — What Game Theory Is

At its heart, Game Theory studies strategic interactions: any situation where the outcome for each agent (person, company, nation, etc.) depends on the actions of others. It’s not about games in the traditional sense, but rather any “game” where decisions must be made in the presence of others who are also making decisions.

Key Assumptions (Classical Game Theory):

1. Rational agents – Everyone acts to maximize their own utility (whatever that means to them).

2. Strategic thinking – Agents consider what others might do before acting.

3. Common knowledge – Everyone knows the rules, the payoffs, and that everyone is rational (though later variants relax this).

🧩 Core Building Blocks

Here are the elements every formal game model needs:

• Players – The decision-makers.
• Strategies – The complete set of actions each player could take.
• Payoffs – The outcomes (usually represented numerically) each player receives depending on the combination of strategies chosen.
• Rules of play – Whether players move simultaneously or in turns, whether they see each other’s moves, whether they play once or repeatedly.

These get framed into a “game form”, and from there, you look for solutions—most famously, Nash equilibrium (more on that soon).

🎲 Quick Clarification: Game Theory vs Probability

You're exactly right—pure chance (like dice) isn’t Game Theory. That’s probability theory. But if you are choosing a strategy knowing that others may randomize their own moves? That’s Game Theory.

For example: in rock-paper-scissors, the “optimal” strategy is not deterministic—it’s a mixed strategy, which involves randomizing your choices. But you’re doing so strategically, in response to an opponent.

🧠 Where We’re Going Next

Next up, I’ll walk through types of games (Phase 1 continued), then move into dominant strategies and Nash equilibrium to set up the first actual game analysis.

Let me know if you want to pause, go deeper, or shift gears—but otherwise I’ll keep pressing forward.

I like the rock paper scissors example. Keep going, connecting to my life where possible. want to internalize this to be a better person/decision maker

Getting started 

Study mode is especially useful for homework help, test prep, and learning new topics. To try it, select “Study and learn” from tools in ChatGPT and ask a question. 

Limitations and what’s next

This is a first step in a longer journey to improve learning in ChatGPT.  Today, study mode is powered by custom system instructions. We chose this approach because it lets us quickly learn from real student feedback and improve the experience—even if it results in some inconsistent behavior and mistakes across conversations. We plan on training this behavior directly into our main models once we’ve learned what works best through iteration and student feedback. 

We’re also exploring functionality to make study mode more engaging and helpful for students including:

• Clearer visualizations for complex or text-heavy concepts
• Goal setting and progress tracking across conversations
• Deeper personalization tailored to each student’s skill level and goals

The research landscape is still taking shape on the best ways to apply AI in education. OpenAI is enabling further research on learning and AI through various partners in its NextGenAI⁠ initiative, and working with experts from the SCALE Initiative at Stanford University’s Accelerator for Learning⁠ to study and share how AI tools influence learning outcomes.

As we run longer-term studies on how students learn best with AI, we intend to publish a deeper analysis of what we’ve learned about the links between model design and cognition, shape future product experiences based on these insights, and work side by side with the broader education ecosystem to ensure AI benefits learners worldwide.

Generated by RSStT. The copyright belongs to the original author.

Source