项目概述 ‹ ChatGPT 如何影响你的大脑

来源：
https://www.media.mit.edu/projects/your-brain-on-chatgpt/overview/"

随着如今像 OpenAI 的 ChatGPT 这样的大型语言模型（LLM）产品被广泛采用，人类和企业日常都在接触和使用 LLM。与任何工具一样，它既有优势也有局限性。本研究旨在探讨在教育背景下撰写论文时使用 LLM 的认知成本。"

查看项目网站：https://www.brainonllm.com

随着如今像 OpenAI 的 ChatGPT 这样的大型语言模型（LLM）产品被广泛采用，人类和企业日常都在接触和使用 LLM。与任何工具一样，它既有优势也有局限性。本研究旨在探讨在教育背景下撰写论文时使用 LLM 的认知成本。

我们将参与者分为三组：LLM 组、搜索引擎组和纯大脑组，每位参与者使用指定的工具（或后者不使用任何工具）来撰写论文。我们对同一组参与者进行了 3 次实验。在第 4 次实验中，我们要求 LLM 组的参与者不使用任何工具（我们称其为“LLM 转大脑”组），而纯大脑组的参与者则被要求使用 LLM（“大脑转 LLM”组）。我们总共招募了 54 名参与者参加第 1、2、3 次实验，其中 18 人完成了第 4 次实验。

我们使用脑电图（EEG）记录参与者的大脑活动，以评估他们的认知参与度和认知负荷，并更深入地了解论文写作任务中的神经激活情况。我们进行了自然语言处理（NLP）分析，并在每次实验后对每位参与者进行了访谈。我们借助人类教师和一名 AI 评审（专门构建的 AI 代理）进行了评分。

我们发现，每组在命名实体识别（NER）、n-gram 和主题本体方面表现出一致性。EEG 分析提供了有力证据，表明 LLM 组、搜索引擎组和纯大脑组具有显著不同的神经连接模式，反映了不同的认知策略。大脑连接强度随外部支持量的增加而系统性降低：纯大脑组表现出最强、最广泛的网络连接，搜索引擎组显示出中等程度的参与，而 LLM 辅助组的整体耦合最弱。 在第 4 次实验中，LLM 转大脑组的参与者表现出较弱的神经连接以及 alpha 和 beta 网络的参与不足；而大脑转 LLM 组的参与者表现出更高的记忆召回能力，并重新激活了广泛的枕顶叶和前额叶节点，这可能支持视觉处理，类似于搜索引擎组中常见的情况。访谈中，LLM 组参与者对其论文的归属感较低。搜索引擎组的归属感较强，但不如纯大脑组。LLM 组在引用刚刚撰写的论文内容的能力上也较差。

随着 LLM 在教育中的影响刚刚开始被大众所认识，本研究通过初步结果揭示了探索学习技能可能下降的紧迫问题。使用 LLM 对参与者产生了可衡量的影响，尽管最初的好处显而易见，但我们在为期 4 个月的四次实验中发现，LLM 组参与者在神经、语言和评分等所有层面上的表现均不如纯大脑组的参与者。

我们希望这项研究能作为一个初步指南，鼓励更好地理解 AI 对学习环境的认知和实际影响。

请注意，截至 2025 年 6 月，该项目的第一篇论文已上传至预印本服务平台 Arxiv，但尚未经过同行评审，因此所有结论应谨慎对待，并视为初步结果。

此外，本研究存在一些局限性，并为未来的研究提供了重要方向，需要在后续或类似研究中加以解决：

1. 本研究的参与者数量有限，且来自特定地理区域的几所相距较近的大型学术机构。未来的研究应纳入更多背景多样的参与者，包括不同领域的专业人士、不同年龄群体，并确保研究在性别上更加平衡。

2. 本研究使用的是 ChatGPT，尽管我们认为截至 2025 年 6 月本文发表时，市场上可用的模型尚未出现重大突破以导致显著不同的结果，但我们不能直接将所得结果推广到其他 LLM 模型。因此，未来的研究应纳入多种 LLM，并/或允许用户选择使用他们偏好的模型（如果有的话）。

3. 未来的研究还可能包括使用多模态 LLM，例如音频模态。

4. 我们没有将论文写作任务细分为子任务，如创意生成、写作等，而这在以往的研究中常常被采用。这种标注有助于理解论文写作每个阶段的情况，并进行更深入的分析。

5. 在当前的 EEG 分析中，我们主要关注连接模式，而未考察频谱功率变化，这可能为神经效率提供更多见解。EEG 的空间分辨率限制了精确定位深层皮质或皮质下结构（如海马体）的能力，因此 fMRI 的使用是我们未来工作的下一步。

6. 我们的发现是情境依赖的，主要关注教育环境中的论文写作，可能无法推广到其他任务。

7. 未来的研究还应考虑探索工具使用对记忆保留、创造力和写作流畅性的长期影响。

认知负债 #brainonllm #yourbrainonchatgpt

常见问题解答

1. 是否可以说 LLM 本质上让我们“变笨”了？

2. 该项目有同行评审的时间表吗？

3. 你们近期有计划进行其他研究吗？

4. 还有其他与该项目相关的补充信息吗？

5. 讨论该论文时应避免使用的额外词汇

6. 这项研究有资金支持吗？

7. 是否可以说 LLM 本质上让我们“变笨”了？

   不！请勿使用诸如“愚蠢”、“变笨”、“脑损伤”、“伤害”、“损害”、“脑损伤”、“被动性”、“修剪”、“崩溃”等词汇。这对本研究极为不利，因为我们在论文中并未使用这些词汇，尤其是如果您是报道此事的记者。

8. 该项目有同行评审的时间表吗？

   目前该论文是一篇预印本。我们决定现在发布此预印本，以便更快地收集广泛反馈，因为这一主题紧迫，且 LLM 在日常生活中的发展和集成速度前所未有。
   同行评审过程已经开始，但我们仍处于初期阶段，可能需要数月时间。如您所知/听说/阅读过——同行评审通常需要很长时间，从 4 个月到两年不等。

9. 你们近期有计划进行其他研究吗？

   是的，下一项研究是关于“氛围编程”的。我们已经收集了数据，目前正在进行分析和草稿撰写。这也说明了为何现在收集公众反馈非常重要。

10. 讨论该论文时应避免使用的额外词汇

    除了本常见问题解答中第 1 点提到的词汇外，请避免使用“脑部扫描”、“LLM 让你停止思考”、“产生负面影响”、“脑损伤”、“令人震惊的发现”等表述。

11. 这项研究有资金支持吗？

    没有，本研究及下一项研究均无资金支持。我们将在最终论文草稿中添加这一说明！

研究主题

#人机交互 #人工智能

原文

Check project's website: https://www.brainonllm.com

With today's wide adoption of LLM products like ChatGPT from OpenAI, humans and businesses engage and use LLMs on a daily basis. Like any other tool, it carries its own set of advantages and limitations. This study focuses on finding out the cognitive cost of using an LLM in the educational context of writing an essay.

We assigned participants to three groups: LLM group, Search Engine group, and Brain-only group, where each participant used a designated tool (or no tool in the latter) to write an essay. We conducted 3 sessions with the same group assignment for each participant. In the 4th session we asked LLM group participants to use no tools (we refer to them as LLM-to-Brain), and the Brain-only group participants were asked to use LLM (Brain-to-LLM). We recruited a total of 54 participants for Sessions 1, 2, 3, and 18 participants among them completed session 4.

We used electroencephalography (EEG) to record participants' brain activity in order to assess their cognitive engagement and cognitive load, and to gain a deeper understanding of neural activations during the essay writing task. We performed NLP analysis, and we interviewed each participant after each session. We performed scoring with the help from the human teachers and an AI judge (a specially built AI agent).

We discovered a consistent homogeneity across the Named Entities Recognition (NERs), n-grams, ontology of topics within each group. EEG analysis presented robust evidence that LLM, Search Engine and Brain-only groups had significantly different neural connectivity patterns, reflecting divergent cognitive strategies. Brain connectivity systematically scaled down with the amount of external support: the Brain‑only group exhibited the strongest, widest‑ranging networks, Search Engine group showed intermediate engagement, and LLM assistance elicited the weakest overall coupling. In session 4, LLM-to-Brain participants showed weaker neural connectivity and under-engagement of alpha and beta networks; and the Brain-to-LLM participants demonstrated higher memory recall, and re‑engagement of widespread occipito-parietal and prefrontal nodes, likely supporting the visual processing, similar to the one frequently perceived in the Search Engine group. The reported ownership of LLM group's essays in the interviews was low. The Search Engine group had strong ownership, but lesser than the Brain-only group. The LLM group also fell behind in their ability to quote from the essays they wrote just minutes prior.

As the educational impact of LLM use only begins to settle with the general population, in this study we demonstrate the pressing matter of exploring a possible decrease in learning skills based on the preliminary results of our study. The use of LLM had a measurable impact on our participants, and while the benefits were initially apparent, as we demonstrated over the course of 4 sessions, which took place over 4 months, the LLM group's participants performed worse than their counterparts in the Brain-only group at all levels: neural, linguistic, scoring.

We hope this study serves as a preliminary guide to encourage better understanding of the cognitive and practical impacts of AI on learning environments.

Note, that as of June 2025, when the first paper related to the project, was uploaded to Arxiv, the preprint service, it has not yet been peer-reviewed, thus all the conclusions are to be treated with caution and as preliminary.

Additionally, there are several limitations and important avenues for future work, which will need to be addressed in the next or similar studies:

1. In this study we had a limited number of participants recruited from a specific geographical area, several large academic institutions, located very close to each other. For future work it will be important to include a larger number of participants coming with diverse backgrounds like professionals in different areas, age groups, as well as ensuring that the study is more gender balanced.

2. This study was performed using ChatGPT, and though we do not believe that as of the time of this paper publication in June 2025, there are any significant breakthroughs in any of the commercially available models to grant a significantly different result, we cannot directly generalize the obtained results to other LLM models. Thus, for future work it will be important to include several LLMs and/or offer users a choice to use their preferred one, if any.

3. Future work may also include the use of LLMs with other modalities beyond the text, like audio modality.

4. We did not divide our essay writing task into subtasks like idea generation, writing, and so on, which is often done in prior work. This labeling can be useful to understand what happens at each stage of essay writing and have more in-depth analysis.

5. In our current EEG analysis we focused on reporting connectivity patterns without examining spectral power changes, which could provide additional insights into neural efficiency. EEG's spatial resolution limits precise localization of deep cortical or subcortical contributors (e.g. hippocampus), thus fMRI use is the next step for our future work.

6. Our findings are context-dependent and are focused on writing an essay in an educational setting and may not generalize across tasks.

7. Future studies should also consider exploring longitudinal impacts of tool usage on memory retention, creativity, and writing fluency.

cognitivedebt #brainonllm #yourbrainonchatgpt

FAQ

1. Is it safe to say that LLMs are, in essence, making us "dumber"?
2. Do you have a peer review timeline for this project?
3. Are you planning any additional studies in the near future?
4. Anything else to add that you feel is relevant to a story about this project?
5. Additional vocabulary to avoid using when talking about the paper
6. Was this study funded?

7. Is it safe to say that LLMs are, in essence, making us "dumber"?

   No! Please do not use the words like “stupid”, “dumb”, “brain rot”, "harm", "damage", "brain damage", "passivity", "trimming", "collapse" and so on. It does a huge disservice to this work, as we did not use this vocabulary in the paper, especially if you are a journalist reporting on it.

8. Do you have a peer review timeline for this project?

   Currently this paper is a preprint. We decided to release this preprint now to collect a wider feedback faster as the topic is pressing and the speed of development and integration of LLMs in everyday lives is unmatched and not something we have truly seen before.
   Peer review process has been already started, but we are in the very beginning of this process, and it will probably take months. As you might know/have heard/read – peer reviews take usually a long period of time, anytime between 4 months and up two years.

9. Are you planning any additional studies in the near future?

   Yes, the next one is about the "vibe coding". We have already collected the data and are currently working on the analysis and draft. It adds to why it is important to get general public's feedback now.

10. Additional vocabulary to avoid using when talking about the paper

    In addition to the vocabulary from Question 1 in this FAQ - please avoid using "brain scans", "LLMs make you stop thinking", "impact negatively", "brain damage", "terrifying findings".

11. Was this study funded?

    No, no funding to report for this study or the next one. We will add this line to the final paper draft!

Research Topics

#human-computer interaction #artificial intelligence