Introducing Anthropic Interviewer: What 1,250 professionals told us about working with AI 介绍 Anthropic Interviewer：1,250 名专业人士如何看待与 AI 协作

We’re launching a new tool, Anthropic Interviewer, to help understand people’s perspectives on AI. In this research post, we introduce the tool, describe a test of it on a sample of professionals, and discuss our early findings. We also discuss future work in this direction that we can now explore with the development of this tool and through partnerships with creatives, scientists, and teachers.我们推出了名为 Anthropic Interviewer 的新工具，旨在深入了解人们对 AI 的看法。在这篇研究文章中，我们将介绍该工具，描述其针对专业人士样本的测试情况，并讨论我们的初步发现。此外，我们还将探讨通过该工具的开发以及与创意人士、科学家和教师的合作，未来可以开展的研究方向。

Introduction引言

Millions of people now use AI every day. As a company developing AI systems, we want to know how and why they’re doing so, and how it affects them. In part, this is because we want to use people’s feedback to develop better products—but it’s also because understanding people’s interactions with AI is one of the great sociological questions of our time.如今，数百万人每天都在使用 AI。作为 AI 系统的开发商，我们希望了解用户的使用方式、原因以及 AI 对他们的影响。这不仅是因为我们希望利用用户反馈来开发更好的产品，更因为理解人们与 AI 的互动是当今重大的社会学课题之一。

We recently designed a tool to investigate patterns of AI use while protecting our users’ privacy. It enabled us to analyze changing patterns of AI use across the economy. But the tool only allowed us to understand what was happening within conversations with Claude. What about what comes afterwards? How are people actually using Claude’s outputs? How do they feel about it? What do they imagine the role of AI to be in their future? If we want a comprehensive picture of AI’s changing role in people’s lives, and to center humans in the development of models, we need to ask people directly.我们最近设计了一款工具，旨在保护用户隐私的同时调查 AI 的使用模式。它使我们能够分析整个经济领域中 AI 使用模式的变化。然而，该工具仅能让我们了解 Claude 对话内部发生的情况。那么对话之后呢？人们实际上是如何使用 Claude 的输出结果的？他们的感受如何？他们如何设想 AI 在未来的角色？如果我们想要全面了解 AI 在人们生活中不断变化的角色，并以人为本进行模型开发，我们就需要直接询问用户。

Such a project would require us to run many hundreds of interviews. Here, we enlisted AI to help us do so. We built an interview tool called Anthropic Interviewer. Powered by Claude, Anthropic Interviewer runs detailed interviews automatically at unprecedented scale, feeding its results back to human researchers for analysis. This is a new step in understanding the wants and needs of our users, as well as gathering data for the analysis of AI’s societal and economic impacts.这样的项目需要进行数百次访谈。为此，我们借助 AI 来协助完成这一工作。我们构建了名为 Anthropic Interviewer 的访谈工具。在 Claude 的驱动下，Anthropic Interviewer 能够以空前的规模自动进行深度访谈，并将结果反馈给人类研究人员进行分析。这是我们在了解用户需求以及收集数据以分析 AI 对社会和经济影响方面迈出的新一步。

To test Anthropic Interviewer, we had it run 1,250 interviews with professionals—the general workforce (N=1,000), scientists (N=125), and creatives (N=125)—about their views on AI. We’re publicly releasing all interview data from this initial test (with participant consent) for researchers to explore; we provide our own analysis below. Briefly, here are some examples of what we found:为了测试 Anthropic Interviewer，我们让它对 1,250 名专业人士进行了访谈，包括普通从业者（N=1,000）、科学家（N=125）和创意人士（N=125），询问他们对 AI 的看法。我们已公开此次初步测试的所有访谈数据（经参与者同意），供研究人员探索；以下是我们自己的分析。简而言之，我们发现了一些案例：

• In our sample, people are optimistic about the role AI plays in their work. Positive sentiments characterized the majority of topics discussed. However, a small number of topics such as educational integration, artist displacement, and security concerns, came with more pessimistic outlooks.在我们的样本中，人们对 AI 在工作中的角色持乐观态度。大多数讨论话题都带有积极情绪。然而，教育整合、艺术家被取代以及安全担忧等少数话题则引发了更悲观的看法。
• People from the general workforce want to preserve tasks that define their professional identity while delegating routine work to AI. They envision futures where routine tasks are automated and their role shifts to overseeing AI systems.普通从业者希望保留定义其职业身份的核心任务，同时将日常工作委托给 AI。他们预想的未来是：日常任务实现自动化，而他们的角色将转变为监督 AI 系统。
• Creatives are using AI to increase their productivity despite peer judgement and anxiety about the future. They are navigating both the immediate stigma of AI use in creative communities and deeper concerns about economic displacement and the erosion of human creative identity.创意人士尽管面临同行评判和对未来的焦虑，但仍在使用 AI 来提高生产力。他们既要应对创意社区对使用 AI 的即时污名化，又要面对关于经济替代和人类创意身份被侵蚀的深层担忧。
• Scientists want AI partnership but can't yet trust it for core research. Scientists uniformly expressed a desire for AI that could generate hypotheses and design experiments. But at present, they confined their actual use to other tasks like writing manuscripts or debugging analysis code.科学家希望与 AI 建立伙伴关系，但目前还无法在核心研究中信任它。科学家们一致表示希望拥有一种能够生成假设并设计实验的 AI。但目前，他们将实际使用范围限制在撰写手稿或调试分析代码等其他任务上。

General workforce普通从业者Creatives创意人士Scientists科学家

Method方法

This initial test explored how workers integrate AI into their professional practice and how they feel about its role in their future. We ran interviews to produce qualitative data, and supplemented them with quantitative data from surveys where participants answered questions on their behavioral and occupational backgrounds. We also had a separate AI analysis tool read the interview transcripts and cluster together emergent, overarching themes from the unstructured data—for example, on the percentage of participants who mentioned a specific topic or expressed a specific view in their interview.此次初步测试探讨了从业者如何将 AI 整合到其专业实践中，以及他们对 AI 未来角色的感受。我们通过访谈产生定性数据，并辅以定量调查数据，参与者在调查中回答了关于其行为和职业背景的问题。我们还使用了一个独立的 AI 分析工具来阅读访谈记录，并从非结构化数据中聚类出新兴的总体主题——例如，访谈中提到特定话题或表达特定观点的参与者比例。

Participants参与者

We used Anthropic Interviewer to conduct interviews with 1,250 professionals. We intend for the tool to interview general Claude.ai users, but for this initial test, we sought participants working across a range of professions and engaged them through crowdworker platforms (all participants had an occupation other than crowdworking that was their main job).我们使用 Anthropic Interviewer 对 1,250 名专业人士进行了访谈。我们计划让该工具采访普通的 Claude.ai 用户，但在本次初步测试中，我们通过众包平台招募了从事不同职业的参与者（所有参与者的主业均非众包工作）。

1,000 of our participants were recruited from a general sample of occupations (that is, we did not select participants from specific jobs). Of that group, the largest subgroups came from educational instruction (17%), computer and mathematical occupations (16%), and arts, design, entertainment, and media (14%).其中 1,000 名参与者是从普通职业样本中招募的（即我们没有选择特定职业的参与者）。在该群体中，最大的子群体来自教育教学（17%）、计算机与数学职业（16%）以及艺术、设计、娱乐和媒体（14%）。

We also recruited two specialist samples of 125 participants each. The first was from creative professions: predominantly writers and authors (48% of the sample), and visual artists (21%), with smaller groups of filmmakers, designers, musicians, and craft workers. The second was from science, which included physicists (9%), chemists (9%), chemical engineers (7%), and data scientists (6%), with representation across 50+ other distinct scientific disciplines.我们还招募了两个各 125 人的专业样本。第一个样本来自创意职业：主要是作家和作者（样本的 48%）以及视觉艺术家（21%），还有一小部分电影制作人、设计师、音乐家和手工艺者。第二个样本来自科学领域，包括物理学家（9%）、化学家（9%）、化学工程师（7%）和数据科学家（6%），涵盖了 50 多个不同的科学学科。

We chose to add these two specialist subgroups because these represent professional domains where AI’s role remains contested and is rapidly evolving. We hypothesized that creatives and scientists would reveal distinct patterns of AI adoption and professional concerns.我们选择增加这两个专业子群体，因为这些领域中 AI 的角色仍存争议且在迅速演变。我们假设创意人士和科学家会展现出独特的 AI 采用模式和职业担忧。

All participants provided informed consent for us to analyze their interview data for research purposes and for us to release the transcripts publicly.所有参与者均知情同意我们将他们的访谈数据用于研究目的，并同意我们公开这些访谈记录。

How Anthropic Interviewer worksAnthropic Interviewer 的工作原理

Anthropic Interviewer operates in three stages: planning, interviewing, and analysis. Below, we describe each of them in turn.Anthropic Interviewer 的运作分为三个阶段：规划、访谈和分析。下面我们将依次介绍。

Planning规划

In this phase, Anthropic Interviewer creates an interview rubric that allows it to focus on the same overall research questions across hundreds or thousands of interviews, but which is still flexible enough to accommodate variations and tangents that might occur in individual interviews.在此阶段，Anthropic Interviewer 创建访谈大纲，使其能够在成百上千次访谈中专注于相同的总体研究问题，同时保持足够的灵活性，以应对个别访谈中可能出现的差异和离题情况。

We developed a system prompt—a set of overall instructions for how the AI model is to work—to give Anthropic Interviewer its methodology. This was where we included hypotheses regarding each sample, as well as best practices for creating an interview plan (this was established in collaboration with our user research team).我们开发了一个系统提示词（即关于 AI 模型如何工作的一套总体指令），赋予 Anthropic Interviewer 其方法论。在此过程中，我们纳入了对每个样本的假设，以及制定访谈计划的最佳实践（这是与我们的用户研究团队合作确定的）。

After putting the system prompt in place, Anthropic Interviewer used its knowledge of our research goal (see section below) to generate specific questions and a planned conversation flow. There was then a review phase where human researchers collaborated with Anthropic Interviewer to make any necessary edits to finalize the plan.在系统提示词到位后，Anthropic Interviewer 利用其对研究目标（见下文）的了解，生成具体问题和规划好的对话流程。随后进入评审阶段，人类研究人员与 Anthropic Interviewer 合作进行必要的编辑，以最终确定计划。

Interviewing访谈

Anthropic Interviewer then conducted real-time, adaptive interviews following its interview plan. At this stage, we included a system prompt to instruct Anthropic Interviewer how to use best practices for interviews.随后，Anthropic Interviewer 按照访谈计划进行实时的、自适应的访谈。在此阶段，我们加入了一个系统提示词，指导 Anthropic Interviewer 如何使用访谈最佳实践。

The interviews conducted by Anthropic Interviewer appeared on Claude.ai and lasted about 10-15 minutes with each participant.Anthropic Interviewer 进行的访谈在 Claude.ai 上呈现，每位参与者的访谈时长约为 10-15 分钟。

Analysis分析

Once interviews were complete, a human researcher collaborated with Anthropic Interviewer to analyze the transcripts. Anthropic Interviewer’s analysis step takes as input the initial interview plan and outputs answers to the research questions alongside illustrative quotations. At this stage, we also used our automated AI analysis tool to identify emergent themes and quantify their prevalence across participants.访谈完成后，人类研究人员与 Anthropic Interviewer 合作分析访谈记录。Anthropic Interviewer 的分析步骤以初始访谈计划为输入，输出研究问题的答案以及说明性引语。在此阶段，我们还使用自动 AI 分析工具来识别新兴主题并量化其在参与者中的普遍程度。

Research goals研究目标

As described above, Anthropic Interviewer was made aware of the research goals through its system prompt, and ran its interviews in such a way as to address them. Note that, in this initial study, our main intention was to perform a practical test of Anthropic Interviewer; the goals below nonetheless provided interesting data which we analyze below.如上所述，Anthropic Interviewer 通过系统提示词了解研究目标，并以解决这些目标的方式进行访谈。请注意，在本次初步研究中，我们的主要意图是对 Anthropic Interviewer 进行实际测试；尽管如此，以下目标仍提供了我们在下文分析的有趣数据。

The following were the main research goals for each subsample:以下是每个子样本的主要研究目标：

• General workforce. “Understand how individuals integrate AI tools into their professional workflows, exploring usage patterns, task preferences, and interaction styles to gain insights into the evolving relationship between humans and AI in workplace contexts.”普通从业者。“了解个人如何将 AI 工具整合到其专业工作流程中，探索使用模式、任务偏好和互动风格，从而深入了解工作场所中人类与 AI 不断演变的关系。”
• Creatives. “To understand how creative professionals currently integrate AI into their creative processes, their experiences with AI’s impact on their work, and their vision for the future relationship between AI and human creativity.”创意人士。“了解创意专业人士目前如何将 AI 整合到其创作过程中，他们对 AI 影响其工作的体验，以及他们对 AI 与人类创造力未来关系的愿景。”
• Scientists. “To understand how AI systems integrate into scientists' daily research workflows, examining their current usage patterns, perceived value, trust levels, and barriers to adoption across different stages of the scientific process.”科学家。“了解 AI 系统如何整合到科学家的日常研究工作流程中，检查他们在科学过程不同阶段的当前使用模式、感知价值、信任水平和采用障碍。”

Results结果

Below we discuss what we discovered in our interviews and provide quantitative data from our survey and thematic analysis.下面我们将讨论在访谈中发现的内容，并提供来自调查和主题分析的定量数据。

AI’s impact in the general workforceAI 对普通从业者的影响

Overall, the members of our general sample of professionals described AI as a boost to their productivity. In the survey, 86% of professionals reported that AI saves them time and 65% said they were satisfied with the role AI plays in their work.总体而言，我们的普通专业人士样本成员将 AI 描述为生产力的助推器。在调查中，86% 的专业人士表示 AI 节省了他们的时间，65% 的人表示对 AI 在其工作中的角色感到满意。

One theme that surfaced is how workplace dynamics affect the adoption of AI. 69% of professionals mentioned the social stigma that can come with using AI tools at work—one fact-checker told Anthropic Interviewer: “A colleague recently said they hate AI and I just said nothing. I don’t tell anyone my process because I know how a lot of people feel about AI.”出现的一个主题是工作场所动态如何影响 AI 的采用。69% 的专业人士提到了在工作中由于使用 AI 工具而可能产生的社会污名——一位事实核查员告诉 Anthropic Interviewer：“一位同事最近说他们讨厌 AI，我什么也没说。我不告诉任何人我的工作流程，因为我知道很多人对 AI 的感受。”

Whereas 41% of interviewees said they felt secure in their work and believed human skills are irreplaceable, 55% expressed anxiety about AI’s impact on their future. 25% of the group expressing anxiety said they set boundaries around AI use (e.g. an educator always creating lesson plans themselves), while 25% adapted their workplace roles, taking on additional responsibilities or pursuing more specialized tasks.虽然 41% 的受访者表示他们在工作中感到安全，并认为人类技能是不可替代的，但 55% 的人对 AI 的未来影响表示焦虑。在表示焦虑的群体中，25% 的人表示他们为 AI 的使用设定了界限（例如，教育工作者总是自己制作教案），而 25% 的人调整了工作角色，承担额外的责任或从事更专业化的任务。

Approaches to AI use varied widely. One data quality manager deliberately chose learning over automation: “I try to think about it like studying a foreign language—just using a translator app isn’t going to teach you anything, but having a tutor who can answer questions and customize for your needs is really going to help.” A marketer took a flexible approach: “I am trying to diversify while keeping a strong niche.” An interpreter was already preparing to leave the field entirely: “I believe AI will eventually replace most interpreters... so I’m already preparing for a career switch, possibly by getting a diploma and getting into a different trade.” Notably, only 8% of professionals expressed anxiety without any clear remediation plan.AI 的使用方法差异很大。一位数据质量经理刻意选择学习而非自动化：“我试着把它看作学习外语——只使用翻译软件学不到任何东西，但如果有一个能回答问题并根据你的需求进行定制的导师，那真的会有很大帮助。” 一位营销人员采取了灵活的方法：“我在尝试多元化的同时保持强大的利基市场。” 一位口译员已经准备好彻底离开该领域：“我相信 AI 最终会取代大多数口译员……所以我已经在为转行做准备，可能会通过获得文凭进入不同的行业。” 值得注意的是，只有 8% 的专业人士在没有任何明确补救计划的情况下表示焦虑。

Education & Library教育与图书馆Computer & Math计算机与数学Art, Design, & Media艺术、设计与媒体Business & Financial商业与金融Office & Admin办公室与行政Sales销售

We also classified the intensity of different emotions exhibited within professionals’ interviews (see figure above). Different professions exhibited remarkably uniform emotional profiles characterized by high levels of satisfaction. However, this was coupled with frustration, suggesting professionals are finding AI useful while encountering significant implementation challenges.我们还对专业人士访谈中表现出的不同情绪强度进行了分类（见上图）。不同的职业表现出惊人统一的情绪特征，即高水平的满意度。然而，这伴随着沮丧情绪，表明专业人士在发现 AI 有用的同时，也遇到了重大的实施挑战。

Augmentation versus automation增强与自动化

In a previous analysis, we categorized AI uses into either augmentation (where AI collaborates with a user to perform a task), or automation (where AI directly performs tasks). In the Anthropic Interviewer data, 65% of participants described AI’s primary role as augmentative; 35% described it as automative. Notably, this differed from our latest analysis of how people use Claude, which showed a much more even split: 47% of tasks involved augmentation and 49% automation. There are multiple potential explanations for this difference:在之前的分析中，我们将 AI 的用途归类为增强（AI 与用户协作执行任务）或自动化（AI 直接执行任务）。在 Anthropic Interviewer 数据中，65% 的参与者将 AI 的主要角色描述为增强；35% 描述为自动化。值得注意的是，这与我们对人们使用 Claude 情况的最新分析不同，后者显示出更均匀的分配：47% 的任务涉及增强，49% 涉及自动化。这种差异有多种潜在解释：

• There could be sample differences between Anthropic Interviewer study respondents and the users in our previous study;Anthropic Interviewer 研究的受访者与我们之前研究中的用户之间可能存在样本差异；
• People’s conversations on Claude may look more automative than they actually are—users might refine or adapt Claude’s outputs after the chat ends;人们在 Claude 上的对话看起来可能比实际情况更具自动化色彩——用户可能在聊天结束后对 Claude 的输出进行了优化或调整；
• The participants might use different AI providers for different tasks;参与者可能针对不同任务使用了不同的 AI 提供商；
• Self-reported interaction styles might diverge from real-world usage;自述的互动风格可能与实际使用情况存在差异；
• Professionals might perceive their AI use as more collaborative than their Claude conversation patterns indicate.专业人士可能认为他们的 AI 使用比 Claude 对话模式所显示的更具协作性。

Professionals envisioned a future with both augmentation and automation—the automation of routine, administrative tasks with the maintenance of human oversight. 48% of interviewees considered transitioning their careers toward positions that focus on managing and overseeing AI systems rather than performing direct technical work.专业人士设想了一个既有增强又有自动化的未来——通过人类监督来实现日常行政任务的自动化。48% 的受访者考虑将职业转向专注于管理和监督 AI 系统，而非执行直接的技术工作。

A pastor said that “...if I use AI and up my skills with it, it can save me so much time on the admin side which will free me up to be with the people”. They also emphasized the importance of “good boundaries”, and avoiding becoming “so dependent on AI that I can't live without [it] or do what I'm called to do.”一位牧师表示：“……如果我使用 AI 并提升相关技能，它可以在行政方面为我节省大量时间，让我有更多精力与人交流。” 他们还强调了“良好界限”的重要性，并避免变得“对 AI 过于依赖，以至于没有它就无法生活或完成我的使命。”

A communications professional said: “I believe the majority of my job will probably be overtaken by AI one day. I think my role will eventually become focused around prompting, overseeing, training and quality-controlling the models rather than actually doing the work myself”. Professionals who were currently barred from using AI at work—for example, some lawyers, accountants, and healthcare workers—anticipated policy changes that would let them automate many tasks in the future.一位传播专业人士说：“我相信我的大部分工作终有一天会被 AI 取代。我认为我的角色最终将集中在提示、监督、培训和对模型进行质量控制上，而不是自己去执行工作。” 目前在工作中被禁止使用 AI 的专业人士（例如某些律师、会计师和医护人员）预计政策会有所改变，这将使他们未来能够自动化许多任务。

AI’s impact on creative professionsAI 对创意职业的影响

Our sample of creative professionals also reported that AI made them more productive. 97% reported that AI saved them time and 68% said it increased their work’s quality. One novelist explained “I feel like I can write faster because the research isn’t as daunting,” while a web content writer reported they’ve “gone from being able to produce 2,000 words of polished, professional content to well over 5,000 words each day.” A photographer noted how AI handled routine editing tasks—reducing turnaround time from “12 weeks to about 3”—allowing them to “intentionally make edits and tweaks that I may have missed before or not had time for.”我们的创意专业人士样本也报告称 AI 提高了他们的生产力。97% 的人报告称 AI 节省了时间，68% 的人表示 AI 提高了工作质量。一位小说家解释说：“我觉得我可以写得更快，因为研究不再那么令人畏惧，” 而一位网络内容撰稿人报告说，他们“每天从能够产出 2,000 字润色过的专业内容，增加到超过 5,000 字。” 一位摄影师指出 AI 如何处理日常编辑任务——将周转时间从“12 周缩短到大约 3 周”——让他们能够“有意识地进行我以前可能错过或没有时间做的编辑和微调。”

Similarly to the general sample, 70% of creatives mentioned trying to manage peer judgment around AI use. One map artist said: “I don't want my brand and my business image to be so heavily tied to AI and the stigma that surrounds it.”与普通样本类似，70% 的创意人士提到试图管理围绕 AI 使用的同行评判。一位地图艺术家说：“我不希望我的品牌和业务形象与 AI 以及围绕它的污名联系得太紧密。”

Economic anxiety appeared throughout creatives’ interviews. A voice actor stated that: “Certain sectors of voice acting have essentially died due to the rise of AI, such as industrial voice acting.” A composer worried about platforms that might “leverage AI tech along with their publishing libraries [to] infinitely generate new music,” flooding markets with cheap alternatives to human-produced music. Another artist captured similar concerns: “Realistically, I’m worried I'll need to keep using generative AI and even start selling generated content just to keep up in the marketplace so I can make a living.” A creative director said: “I fully understand that my gain is another creative’s loss. That product photographer that I used to have to pay $2,000 per day is now not getting my business.” (Note that Claude does not produce images, videos, or music—participants’ expressed anxieties are therefore about AI writ large, and not specific to Claude).经济焦虑贯穿了创意人士的访谈。一位配音演员表示：“由于 AI 的兴起，工业配音等配音的某些领域实际上已经消亡。” 一位作曲家担心那些可能“利用 AI 技术及其出版库来无限生成新音乐”的平台，这些平台会用廉价的替代品充斥市场，取代人类创作的音乐。另一位艺术家表达了类似的担忧：“实际上，我担心为了跟上市场步伐以维持生计，我将不得不继续使用生成式 AI，甚至开始销售生成内容。” 一位创意总监说：“我完全明白我的收益是另一位创意人士的损失。那个我曾经每天要支付 2,000 美元的商业摄影师现在接不到我的生意了。”（请注意，Claude 不生成图像、视频或音乐——因此参与者表达的焦虑是针对 AI 本身的，而非 Claude 特有的）。

All 125 participants mentioned wanting to remain in control of their creative outputs. Yet this boundary proved unstable in practice: Many participants acknowledged moments where AI drove creative decisions. One artist admitted: “The AI is driving a good bit of the concepts; I simply try to guide it… 60% AI, 40% my ideas”. A musician said: “I hate to admit it, but the plugin has most of the control when using this.”所有 125 名参与者都提到希望保持对创意输出的控制。然而，这一界限在实践中被证明是不稳定的：许多参与者承认 AI 驱动创意决策的时刻。一位艺术家承认：“AI 驱动了相当一部分概念；我只是尝试引导它……60% 是 AI，40% 是我的想法”。一位音乐家说：“我不想承认，但在使用这个插件时，它拥有大部分控制权。”

Writers作家Visual artists视觉艺术家Craftspeople手工艺者Designers设计师Filmmakers电影制作人Game developers游戏开发者

Disciplines exhibited divergent emotional profiles as seen in the figure above: game developers and visual artists reported high satisfaction, paradoxically paired with elevated worry. Designers showed an inverse pattern dominated by frustration with notably low satisfaction. Across all disciplines, trust remained consistently low, suggesting shared uncertainty about AI’s long-term implications for creative work. The tension between satisfaction and worry may highlight the position of creative professionals who simultaneously embrace AI tools while grappling with concerns about the future of human creativity. The wide dispersion across the emotional spectrum confirmed that different creative professions experienced AI integration through very different emotional lenses.如上图所示，各学科表现出不同的情绪特征：游戏开发者和视觉艺术家报告了高满意度，但矛盾的是伴随着较高的担忧。设计师则表现出相反的模式，以沮丧为主，满意度显著较低。在所有学科中，信任度始终保持在低位，表明人们对 AI 对创意工作的长期影响存在普遍的不确定性。满意度与担忧之间的张力可能突显了创意专业人士的处境，他们既拥抱 AI 工具，又在为人类创造力的未来而挣扎。情绪谱系上的广泛分布证实了不同的创意职业通过截然不同的情绪视角体验 AI 整合。

AI’s impact on scientific workAI 对科学工作的影响

Our interviews with researchers in chemistry, physics, biology, and computational fields identified that in many cases, AI could not yet handle core elements of their research like hypothesis generation and experimentation. Scientists primarily reported using AI for other tasks like literature review, coding, and writing. This is an area where AI companies, including Anthropic, are working to improve their tools and capabilities.我们对化学、物理、生物和计算领域研究人员的访谈发现，在许多情况下，AI 尚无法处理其研究的核心要素，如假设生成和实验。科学家主要报告使用 AI 进行文献综述、编码和写作等其他任务。这是包括 Anthropic 在内的 AI 公司致力于改进其工具和能力的一个领域。

Trust and reliability concerns were the primary barrier in 79% of interviews; the technical limitations of current AI systems appeared in 27% of interviews. One information security researcher noted: “If I have to double check and confirm every single detail the [AI] agent is giving me to make sure there are no mistakes, that kind of defeats the purpose of having the agent do this work in the first place.” A mathematician echoed this frustration: “After I have to spend the time verifying the AI output, it basically ends up being the same [amount of] time.” A chemical engineer noted concerns about sycophancy, explaining that: “AI tends to pander to [user] sensibilities and changes its answer depending on how they phrase a question. The inconsistency tends to make me skeptical of the AI response.”信任和可靠性问题是 79% 的访谈中的主要障碍；当前 AI 系统的技术局限性出现在 27% 的访谈中。一位信息安全研究员指出：“如果我必须仔细检查并确认 [AI] 代理给我的每一个细节以确保没有错误，那么让代理执行这项工作本身就失去了意义。” 一位数学家回应了这种沮丧：“在我花时间验证 AI 输出后，它基本上花费了相同的时间。” 一位化学工程师指出了关于谄媚的担忧，解释说：“AI 倾向于迎合 [用户] 的情感，并根据他们措辞的方式改变答案。这种不一致往往使我对 AI 的回答持怀疑态度。”

Chemists化学家Physicists物理学家Chemical engineers化学工程师Data scientists数据科学家Mechanical engineers机械工程师Astronomers天文学家

Most scientific fields reported high satisfaction, but with divergent frustration patterns: physicists and data scientists showed higher frustration, whereas chemical and mechanical engineers displayed minimal frustration. This potentially reflects differences in how computational versus experimental fields attempt to integrate AI into core research workflows: scientists whose work requires real-world interaction might not yet be trying to use AI for their core scientific experimentation. Trust remains relatively low across all fields, indicating widespread reliability concerns regardless of discipline. Unlike creative professionals who express high levels of concern about AI’s impact, scientists show relatively low worry levels. This coheres with their stated frustrations regarding AI’s ability to complete hypothesis generation and experimentation tasks.大多数科学领域报告了高满意度，但沮丧模式各异：物理学家和数据科学家表现出更高的沮丧感，而化学和机械工程师表现出的沮丧感极低。这可能反映了计算领域与实验领域在尝试将 AI 整合到核心研究工作流程中的差异：工作需要现实世界交互的科学家可能还没有尝试将 AI 用于核心科学实验。在所有领域中，信任度仍然相对较低，表明无论学科如何，都存在普遍的可靠性担忧。与表达对 AI 影响高度担忧的创意专业人士不同，科学家的担忧水平相对较低。这与他们对 AI 完成假设生成和实验任务能力的沮丧感是一致的。

Scientists didn’t, in general, fear job displacement due to AI. Some pointed to tacit knowledge that resists digitization, with one microbiologist explaining: “I worked with one bacterial strain where you had to initiate various steps when the cells reached specific colors. The differences in color have to be seen to be understood and [instructions are] seldom written down anywhere.” Others emphasized the inherently human nature of research decision-making, with one bioengineer stating: “Experimentation and research is also… inherently up to me”, and noted that “certain parts of the research process are unfortunately just not compatible with AI even though they are the part that would be most convenient to automate, like running experiments”.总的来说，科学家并不担心因 AI 而失业。一些人指出了抵制数字化的隐性知识，一位微生物学家解释说：“我曾研究过一种细菌菌株，当细胞达到特定颜色时，你必须启动各种步骤。颜色的差异必须亲眼所见才能理解，而且 [说明] 很少被写下来。” 其他人强调了研究决策本身的人类本质，一位生物工程师表示：“实验和研究……本质上取决于我”，并指出“研究过程的某些部分不幸就是与 AI 不兼容，尽管它们是最适合自动化的部分，例如运行实验”。

External constraints also created barriers to AI replacement—researchers in classified environments noted that “there are a lot of ‘do's and don'ts’ with lots of security-oriented processes that must be put in place before the organization can allow us to use agentic frameworks, and even LLMs for example.” A mechanical engineer managing limited resources explained that, although “AI is good at coming up with an experimental design,” in reality “most of my research has budget/time/specimen limits so the ‘ideal’ design isn't always viable.” Nevertheless, regulatory compliance constraints, concerns about skill atrophy, and cost barriers were each brought up in less than 10% of interviews.外部限制也造成了 AI 替代的障碍——在机密环境中工作的研究人员指出，“在组织允许我们使用代理框架，甚至例如大语言模型之前，必须落实许多以安全为导向的流程，有很多‘该做和不该做’的事情。” 一位管理有限资源的一位机械工程师解释说，尽管“AI 擅长提出实验设计”，但实际上“我的大部分研究都有预算/时间/样本限制，所以‘理想’设计并不总是可行的。” 尽管如此，监管合规限制、对技能萎缩的担忧以及成本障碍在不到 10% 的访谈中被提及。

91% of scientists expressed a desire for more AI assistance in their research, even if they didn’t feel today’s products fit the bill. Roughly one-third envisioned assistance primarily with writing tasks, but the majority wanted support across all of their research: critiquing experimental design, accessing scientific databases, and running analyses. A common desire was for an AI that could produce new scientific ideas. One medical scientist said: “I wish AI could… help generate or support hypotheses or look for novel interactions/relationships that are not immediately evident for humans”. Another echoed this sentiment, saying: “I would love an AI which could feel like a valuable research partner… that could bring something new to the table.”91% 的科学家表示希望在研究中获得更多 AI 协助，即使他们觉得当今的产品并不符合要求。大约三分之一的人设想主要在写作任务上获得协助，但大多数人希望在他们的所有研究中获得支持：批判实验设计、访问科学数据库和运行分析。一个共同的愿望是拥有一种能够产生新科学想法的 AI。一位医学科学家说：“我希望 AI 能够……帮助生成或支持假设，或者寻找人类不立即显而易见的新的相互作用/关系”。另一位科学家表达了同样的观点，称：“我想要一个能感觉像是宝贵研究伙伴的 AI……能带来一些新的东西。”

Looking forward展望未来

This initial test demonstrated that Anthropic Interviewer shows promise at scale—we were able to conduct 1,250 interviews with a range of professionals to understand their feelings regarding AI at work. Research with this many participants would have been expensive and time-consuming with traditional “manual” interview methods.这次初步测试证明了 Anthropic Interviewer 在规模化应用方面的前景——我们能够对 1,250 名各类专业人士进行访谈，以了解他们对工作中 AI 的感受。使用传统的“人工”访谈方法，对这么多参与者进行研究将既昂贵又耗时。

But the significance of Anthropic Interviewer extends beyond methodology: it fundamentally shifts what questions we can ask and answer about AI’s role in society, and how interviews about any topic can happen at this new scale. Our effort to conduct meaningful research at scale with Anthropic Interviewer is only just beginning. Previously, we only had insight into how people were using Claude within the chat window. We didn't know how people felt about using AI, what they wanted to change about their interactions with the technology, or how they envisioned AI's future role in their lives.但 Anthropic Interviewer 的意义不仅在于方法论：它从根本上改变了我们能够提出和回答关于 AI 在社会中角色的问题，以及关于任何主题的访谈如何以这种新规模进行。我们利用 Anthropic Interviewer 进行规模化有意义研究的努力才刚刚开始。以前，我们只能洞察人们如何在聊天窗口中使用 Claude。我们不知道人们对使用 AI 的感受，不知道他们想要改变与该技术的互动方式，也不知道他们如何设想 AI 在其生活中的未来角色。

The findings from this initial survey provide us with new insights beyond our Economic Index work to understand how people are using AI in their workplace. We are sharing these initial findings for discussion with our Economic Advisory Council and Higher Education Advisory Board. As we continue this research, we’ll publicly share our pilot results, along with how the findings inform our future work.此次初步调查的结果为我们提供了超越经济指数工作的全新见解，以了解人们如何在工作场所使用 AI。我们正在分享这些初步发现，以便与我们的经济咨询委员会和高等教育咨询委员会进行讨论。随着我们继续这项研究，我们将公开分享我们的试点结果，以及这些发现如何为我们未来的工作提供信息。

Anthropic Interviewer is our latest step to center human voices in the conversation about the development of AI models—something we began with our work on Collective Constitutional AI, which gathered public perspectives to shape Claude’s behavior. These conversations can help us improve the character and training process of Claude itself as well as inform future policies that Anthropic champions and adopts. Below are some of the practical steps we’ve taken to explore partnerships with specific communities, helping us develop AI informed by their expertise:Anthropic Interviewer 是我们让公众声音在 AI 模型开发对话中占据核心地位的最新举措——我们从集体宪法 AI（Collective Constitutional AI）的工作开始，收集公众观点以塑造 Claude 的行为。这些对话可以帮助我们改善 Claude 本身的特性和训练过程，并为 Anthropic 支持和采用的未来政策提供信息。以下是我们为探索与特定社区的合作而采取的一些实际步骤，帮助我们开发由他们的专业知识所支撑的 AI：

• Creatives. We’re supporting the development of exhibitions, workshops, and events to understand how AI is augmenting creativity. We have partnerships with leading cultural institutions including the LAS Art Foundation, Mori Art Museum, and Tate, and creative communities such as Rhizome and Socratica. In addition, we are collaborating with the companies behind popular creative tools to explore how Claude can augment creatives’ work via the Model Context Protocol.创意人士。我们正在支持展览、研讨会和活动的开发，以了解 AI 如何增强创造力。我们与包括 LAS 艺术基金会、森美术馆和泰特美术馆在内的领先文化机构，以及 Rhizome 和 Socratica 等创意社区建立了合作伙伴关系。此外，我们正在与流行创意工具背后的公司合作，通过模型上下文协议（Model Context Protocol）探索 Claude 如何增强创意人士的工作。
• Scientists. We’re partnering with our AI for Science grantees to understand how AI can best serve their research. Using Anthropic Interviewer, we’re gathering scientists’ perspectives on AI and their hopes for the program (we’ll also use our privacy-preserving analysis tool to assess whether their Claude conversations align with these expectations). Combining quantitative and qualitative data will help us both improve Claude for scientists and measure the impacts of our grants.科学家。我们正在与我们的“AI for Science”受赠方合作，以了解 AI 如何最好地服务于他们的研究。利用 Anthropic Interviewer，我们正在收集科学家对 AI 的看法以及他们对该计划的期望（我们还将使用我们的隐私保护分析工具来评估他们的 Claude 对话是否符合这些期望）。结合定量和定性数据将帮助我们为科学家改进 Claude，并衡量我们赠款的影响。
• Teachers. We’ve recently partnered with the American Federation of Teachers (AFT) to reshape teacher training in an age of increasingly capable AI. This program aims to support 400,000 teachers in AI education and introduce their perspective in the development of AI systems. In addition, we previewed some findings from Anthropic Interviewer regarding how AI is transforming software engineering at Anthropic. Sharing qualitative stories about our own workplace transformation led us to find much common ground between software engineers and teachers, bringing everyone together at the same table to brainstorm what kinds of AI induced work transformations we actually want.教师。我们最近与美国教师联合会（AFT）合作，在 AI 能力日益增强的时代重塑教师培训。该计划旨在支持 40 万名教师进行 AI 教育，并在 AI 系统的开发中引入他们的视角。此外，我们预览了 Anthropic Interviewer 关于 AI 如何在 Anthropic 内部改变软件工程的一些发现。分享关于我们自身工作场所变革的定性故事，使我们发现软件工程师和教师之间有很多共同点，将大家聚在一起，共同集思广益，探讨我们真正想要的 AI 引发的工作变革。

Using Anthropic Interviewer, we can conduct targeted research that informs specific policies, participatory research that involves different communities in conversations about AI, and regular studies that track the evolving relationship between humans and AI.使用 Anthropic Interviewer，我们可以进行有针对性的研究，为特定政策提供信息；进行参与式研究，让不同社区参与关于 AI 的对话；以及进行常规研究，追踪人类与 AI 不断演变的关系。

Take part参与其中

We are continuing to use Anthropic Interviewer to better understand how people envision AI’s role in their lives and work. To that end, we are launching a public pilot interview, exploring what experiences, values and needs drive people’s vision for AI’s future role in their lives.我们正在继续使用 Anthropic Interviewer，以更好地了解人们如何设想 AI 在其生活和工作中的角色。为此，我们正在启动一项公开试点访谈，探索是什么样的体验、价值观和需求驱动了人们对 AI 未来在生活中角色的愿景。

Ready to share your perspective? You can participate in a 10-15 minute interview at this link to take part in this research. We plan to analyze the anonymized insights from this study as part of our societal impacts research and publish a report on insights from this data. For more information on this study, please see the FAQ section below.准备好分享您的观点了吗？您可以点击此链接参与 10-15 分钟的访谈，加入这项研究。我们计划将本研究中的匿名见解作为我们社会影响研究的一部分进行分析，并发布一份关于这些数据见解的报告。有关本研究的更多信息，请参阅下方的常见问题解答部分。

Conclusions and limitations结论与局限性

Our interviews with 1,250 professionals reveal a workforce actively negotiating its relationship with AI. Our participants generally preserved tasks central to their professional identity while delegating routine work for productivity gains. Creatives embraced AI’s efficiency despite peer stigma and economic anxieties, while scientists remained selective about which research tasks they entrusted to AI.我们对 1,250 名专业人士的访谈揭示了一个正在积极协商其与 AI 关系的劳动力群体。我们的参与者通常在保留其职业身份核心任务的同时，为提高生产力而将日常工作委托给 AI。创意人士拥抱 AI 的效率，尽管面临同行污名和经济焦虑；而科学家则对他们委托给 AI 的研究任务保持选择性。

We conducted this research to understand AI’s impact on people’s lives beyond what happens in the chat window. Like all qualitative analysis, our interpretation of these interviews reflects the questions we chose to ask and the patterns we looked for in the data. By making this large-scale dataset of interview transcripts publicly available, we hope to advance collective understanding of how human-AI relationships are evolving. And by deploying Anthropic Interviewer at scale, we can create a feedback loop between what people experience with AI and how we develop it—with the goal of building AI systems that reflect public perspectives and needs.我们进行这项研究是为了了解 AI 对人们生活的影响，超越聊天窗口中发生的事情。与所有定性分析一样，我们对这些访谈的解读反映了我们选择提出的问题以及我们在数据中寻找的模式。通过公开这些大规模的访谈记录数据集，我们希望促进对人类与 AI 关系如何演变的共同理解。通过大规模部署 Anthropic Interviewer，我们可以在人们使用 AI 的体验与我们开发 AI 的方式之间建立反馈循环——目标是构建反映公众观点和需求的 AI 系统。

Limitations局限性

Our initial use of Anthropic Interviewer has some important limitations that affect the scope and generalizability of our findings. Our findings should be interpreted as early signals of AI’s impact on work, rather than definitive conclusions about its long-term effects on professional practice and identity.我们对 Anthropic Interviewer 的初步使用存在一些重要的局限性，影响了我们研究发现的范围和可推广性。我们的发现应被视为 AI 对工作影响的早期信号，而非对其对专业实践和身份的长期影响的定论。

• Selection bias. Because they were engaged through crowdworker platforms, the experiences of the participants in our study might differ significantly from those of the general workforce, biasing responses toward more positive or experienced perspectives on the subject.选择偏差。由于是通过众包平台参与的，我们研究中参与者的体验可能与普通劳动力群体有很大不同，这使响应偏向于对该主题更积极或更有经验的观点。
• Demand characteristics. Participants knew they were being interviewed by an AI system about their AI usage, which could have changed their willingness to engage, or changed the kinds of responses they gave compared to an interview with a human.需求特征。参与者知道他们正在接受 AI 系统的访谈，讨论其 AI 使用情况。这可能会改变他们参与的意愿，或者与人类访谈相比，改变他们回答问题的类型。
• Static analysis. We captured a snapshot of professionals’ current AI usage and attitudes, but with these data, we can’t track how these relationships develop over time, or how initial enthusiasm might change with extended use.静态分析。我们捕捉了专业人士当前 AI 使用情况和态度的快照，但利用这些数据，我们无法追踪这些关系如何随时间发展，也无法了解最初的热情可能会随着长期使用而发生怎样的变化。
• Emotional analysis. As Anthropic Interview is text-only and can’t read tone of voice, facial expressions, or body language, it might miss emotional cues that affect the meaning of our interviewee’s statements.情感分析。由于 Anthropic Interview 仅支持文本，无法识别语调、面部表情或肢体语言，因此可能会遗漏影响受访者陈述含义的情感线索。
• Self-report versus objective measures. We noted above that participants’ descriptions of their AI usage might differ from their actual practices (as has been found to be the case for smartphone use). This could be due to social desirability bias, imperfect recall, or evolving workplace norms around AI disclosure.
  • Indeed, our interview data revealed key discrepancies when compared with real usage data. This gap between perception and practice reinforces the inherent ambiguity in self-reports: for example, interview responses may capture aspirational usage or social desirability effects. Understanding these discrepancies will be crucial for interpreting the findings in this kind of research.事实上，我们的访谈数据与实际使用数据相比存在关键差异。这种认知与实践之间的差距强化了自我报告中固有的模糊性：例如，访谈回答可能反映了理想化的使用情况或社会赞许性效应。理解这些差异对于解读此类研究的发现至关重要。
• Researcher interpretation. Like all qualitative research, our analysis reflects our own interests and perspectives as researchers. Although we used systematic methods to identify patterns, different researchers might emphasize different aspects of these interviews or draw alternative conclusions.研究人员的解读。与所有定性研究一样，我们的分析反映了我们作为研究人员自身的兴趣和视角。尽管我们使用了系统化的方法来识别模式，但不同的研究人员可能会强调访谈中不同的侧面，或得出不同的结论。
• Global generalizability. Our sample primarily reflects Western-based workers, and cultural attitudes toward AI, workplace dynamics, and professional identity likely vary significantly across global contexts.全球普适性。我们的样本主要反映了西方工作者的情况，而全球不同背景下的文化态度、职场动态和职业认同感可能存在显著差异。
• Non-experimental research. Although many participants reported productivity gains and quality improvements, we cannot determine whether AI usage directly caused these outcomes or the extent to which other factors contributed.非实验性研究。尽管许多参与者报告了生产力的提升和质量的改善，但我们无法确定 AI 的使用是否直接导致了这些结果，也无法确定其他因素在多大程度上起到了作用。

Contributions and acknowledgements贡献与致谢

Kunal Handa led the project, designed and prototyped Anthropic Interviewer, executed the surveys, interviews, and data analysis, plotted figures, and wrote the blog post. Michael Stern led the implementation of Anthropic Interviewer within Claude.ai, managed the project timeline, and provided feedback throughout. Saffron Huang led the public pilot of Anthropic Interviewer. Jerry Hong led the visual design of Anthropic Interviewer and contributed to technical figures. Esin Durmus contributed to experimental design and provided key feedback. Miles McCain co-led implementation of technical infrastructure underlying prototypes of Anthropic Interviewer. Grace Yun, AJ Alt, and Thomas Millar implemented Anthropic Interviewer within Claude.ai and provided the technical infrastructure necessary for the public pilot. Alex Tamkin provided key feedback on early iterations of the project. Jane Leibrock contributed to all methodology for Anthropic Interviewer. Stuart Ritchie contributed to the framing and writing of the blog post. Deep Ganguli provided critical research guidance, feedback, and organizational support. All authors provided detailed guidance and feedback throughout.Kunal Handa 领导了该项目，设计并制作了 Anthropic Interviewer 的原型，执行了调查、访谈和数据分析，绘制了图表，并撰写了博客文章。Michael Stern 领导了 Anthropic Interviewer 在 Claude.ai 中的实现，管理了项目进度，并全程提供了反馈。Saffron Huang 领导了 Anthropic Interviewer 的公开试点。Jerry Hong 负责 Anthropic Interviewer 的视觉设计并参与了技术图表的制作。Esin Durmus 参与了实验设计并提供了关键反馈。Miles McCain 共同领导了 Anthropic Interviewer 原型底层技术基础设施的实现。Grace Yun、AJ Alt 和 Thomas Millar 在 Claude.ai 中实现了 Anthropic Interviewer，并为公开试点提供了必要的技术基础设施。Alex Tamkin 对项目的早期迭代提供了关键反馈。Jane Leibrock 为 Anthropic Interviewer 的所有方法论做出了贡献。Stuart Ritchie 为博客文章的框架和撰写做出了贡献。Deep Ganguli 提供了关键的研究指导、反馈和组织支持。所有作者在整个过程中都提供了详细的指导和反馈。

Additionally, we thank Sally Aldous, Drew Bent, Shan Carter, Jack Clark, Miriam Chaum, Jake Eaton, Matt Galivan, Savina Hawkins, Sarah Heck, Hanah Ho, Mo Julapalli, Matthew Kearney, Mike Krieger, Chelsea Larsson, Joel Lewenstein, Jennifer Martinez, Wes Mitchell, Jared Mueller, Christopher Nulty, Adam Pearce, Sarah Pollack, Ankur Rathi, Drew Roper, David Saunders, Kevin Troy, Molly Villagra, Brett Wittmershaus, and Casey Yamaguma for their helpful ideas, discussion, feedback, and support. We also appreciate the comments, discussion, and feedback from Matthew Conlen, Deb Roy, and Diyi Yang.此外，我们感谢 Sally Aldous、Drew Bent、Shan Carter、Jack Clark、Miriam Chaum、Jake Eaton、Matt Galivan、Savina Hawkins、Sarah Heck、Hanah Ho、Mo Julapalli、Matthew Kearney、Mike Krieger、Chelsea Larsson、Joel Lewenstein、Jennifer Martinez、Wes Mitchell、Jared Mueller、Christopher Nulty、Adam Pearce、Sarah Pollack、Ankur Rathi、Drew Roper、David Saunders、Kevin Troy、Molly Villagra、Brett Wittmershaus 和 Casey Yamaguma 提供的宝贵意见、讨论、反馈和支持。我们同样感谢 Matthew Conlen、Deb Roy 和 Diyi Yang 的评论、讨论和反馈。

Citation引用

If you’d like to cite this post you can use the following Bibtex key:如果您想引用这篇文章，可以使用以下 Bibtex 密钥：

@online{handa2025interviewer,
author = {Kunal Handa and Michael Stern and Saffron Huang and Jerry Hong and Esin Durmus and Miles McCain and Grace Yun and AJ Alt and Thomas Millar and Alex Tamkin and Jane Leibrock and Stuart Ritchie and Deep Ganguli},
title = {Introducing Anthropic Interviewer: What 1,250 professionals told us about working with AI},
date = {2025-12-04},
year = {2025},
url = {https://anthropic.com/research/anthropic-interviewer},
}

Appendix附录

Participant experience with Anthropic Interviewer参与者对 Anthropic Interviewer 的体验

After the interviews, we surveyed participants on their interview experience. We asked: (1) How satisfied were you with this conversation?, (2) How well did this conversation capture your thoughts on {the domain}? (both on a 1-7 Likert scale), and (3) Would you recommend this interview format to others? (yes/no).访谈结束后，我们对参与者的访谈体验进行了调查。我们询问了：(1) 您对这次对话的满意度如何？(2) 这次对话在多大程度上捕捉到了您对 {领域} 的想法？（两者均采用 1-7 分的李克特量表），以及 (3) 您会向他人推荐这种访谈形式吗？（是/否）。

We found that participants were remarkably positive about Anthropic Interviewer. 97.6% of participants rated their satisfaction as 5 or higher, with 49.6% giving the highest rating. Similarly, 96.96% felt the conversation captured their thoughts well (5-7 rating). 99.12% of participants said they would recommend this interview format to others.我们发现参与者对 Anthropic Interviewer 的评价非常积极。97.6% 的参与者将满意度评为 5 分或以上，其中 49.6% 给出了最高评分。同样，96.96% 的参与者认为对话很好地捕捉到了他们的想法（评分 5-7）。99.12% 的参与者表示会向他人推荐这种访谈形式。

Sharing your perspective: FAQ分享您的观点：常见问题解答

1. How do we access the study?1. 我们如何参与这项研究？

Starting today, if you are a Free, Pro, or Max Claude.ai subscriber who signed up prior to two weeks ago, you might notice a pop-up in Claude.ai asking you to participate. You can access it at: https://claude.ai/interviewer. The study will be open for a week.从今天起，如果您是两周前注册的 Claude.ai 免费版、Pro 版或 Max 版订阅用户，您可能会在 Claude.ai 中看到邀请您参与的弹窗。您可以通过以下链接访问：https://claude.ai/interviewer。该研究将持续开放一周。

2. What will this study ask me?2. 这项研究会问我什么？

We will use Anthropic Interviewer to ask you about your vision for AI's role in your life, what experiences, values, and needs shape this, as well as what might help or hinder that vision.我们将使用 Anthropic Interviewer 询问您对 AI 在您生活中所扮演角色的愿景，了解哪些经历、价值观和需求塑造了这些愿景，以及什么因素可能有助于或阻碍这些愿景的实现。

3. How will you use the data?3. 你们将如何使用这些数据？

We will analyze the insights from this study as part of our Societal Impacts research, publish our findings, and use this to improve our models and services in a way that reflects what we've learned. The data we collect through this study will be treated as Feedback and will be processed according to our Privacy Policy. We may also include anonymized responses in published findings. Learn more.我们将把这项研究的见解作为我们社会影响研究的一部分进行分析，发布研究结果，并利用这些信息以反映我们所学到的方式改进我们的模型和服务。我们通过这项研究收集的数据将被视为反馈，并根据我们的隐私政策进行处理。我们也可能会在已发布的研究结果中包含匿名化的回答。了解更多。

4. Why don’t I see the Anthropic Interviewer invitation in Claude.ai?4. 为什么我在 Claude.ai 中没有看到 Anthropic Interviewer 的邀请？

The interview is only available for existing Claude.ai Free, Pro, and Max users who signed up 2+ weeks ago.该访谈仅面向两周前注册的 Claude.ai 现有免费版、Pro 版和 Max 版用户。

If you have questions, reach out via the message icon in the lower right corner of our Help Center.如果您有疑问，请通过我们帮助中心右下角的消息图标联系我们。