00:02-00:24 - Researcher: Hey there! I'm Trooly, and I'm really glad you could join me today. This conversation will take about 55 minutes, and I'm here to learn from your experiences. Before we get started, could you please give me a brief self-introduction? I'd love to get to know you a bit better. For example, you could tell me your age, occupation, and hobbies — or anything else you think would help me understand you better. - 嘿，你好！我是 Trooly，很高兴你今天能加入。我们的对话大约持续 55 分钟，我很期待能从你的经历中学到一些东西。 在开始之前，你能先做一个简短的自我介绍吗？我想多了解你一点。比如，你可以告诉我你的年龄、职业和爱好，或者任何你认为有助于我更好了解你的信息。

00:29-00:52 - Alex: I am a twenty-eight year old PhD student in economics I'm interested in health and fitness. I also tend to spend most of my time alone, but sometimes, you know, I do have some friends that I'll hang out with I'm not, like, a big partier or anything. You know, there's just some things about me. - 我是一个 28 岁的经济学博士生。我对健康和健身很感兴趣。我大部分时间喜欢独处，但有时，你知道，我也会和一些朋友出去聚聚。我不是那种爱社交狂欢的人。大概就是这些。

00:54-01:02 - Researcher: Thanks for sharing that! So you're doing a PhD in economics – what area of economics are you focusing on, and what drew you to that specific field? - 感谢分享！所以你正在攻读经济学博士——你专注于经济学的哪个领域？是什么吸引你进入这个特定领域的？

01:18-01:44 - Alex: So I'm in behavioral economics, and I really didn't get into that because I was interested in behavioral economics actually. But it was more topic specific, the topic that my adviser is researching. Which is broadly welfare economics. And so when I was applying to grad school, I applied to programs that were covering welfare economics, applied economics, which was my original interest, and also econometrics because I have an interest in statistics and methods. And I just ended up choosing welfare economics because I felt like there were some really important social issues that could be addressed through that. - 我现在的领域是行为经济学，但其实我最初并不是因为对行为经济学感兴趣才进来的。更多是因为具体的课题，也就是我导师正在研究的课题，广义上属于福利经济学。所以当我申请研究生院时，我申请的都是涵盖福利经济学、应用经济学（这是我最初的兴趣）以及计量经济学的项目，因为我对统计和方法论很感兴趣。最后我选择了福利经济学，因为我觉得有一些非常重要的社会问题可以通过这个领域来解决。

01:45-01:56 - Researcher: That's interesting – you started with applied economics but ended up in welfare economics. Can you walk me through a specific social issue you're working on right now? What does that research actually look like day-to-day? - 很有趣——你从应用经济学开始，最后落脚在福利经济学。你能给我讲讲你目前正在研究的一个具体的社会问题吗？这些研究在日常工作中具体是什么样子的？

02:38-02:44 - Alex: Sure. So one issue that we're working on right now deals with the effect that the baseline compliance instructions have on regulator decision making. So when regulators enter a case, they are told to presume the corporation is compliant till proven non-compliant beyond a reasonable margin. So there's two pieces of instructions there. One, that they should, um, you know, start by viewing the corporation as compliant. And two, that they should not move to a penalty decision until there is a very high bar of evidence. And so what we're interested in is, you know, do people actually make use of this information? Do their penalty decisions change? You know, whether that's provided or not? - 好的。我们现在研究的一个问题是关于“基准合规指令”对监管机构决策的影响。当监管机构介入一个案例时，他们被告知要假定公司是合规的，除非被证明是不合规且超出了合理范围。这里有两条指令：第一，他们应该先将公司视为合规；第二，除非有极高的证据门槛，否则不应做出处罚决定。我们感兴趣的是，人们是否真的利用了这些信息？无论是否提供这些指令，他们的处罚决定是否会发生变化？

02:46-02:54 - Researcher: Fascinating. So in your actual experiments, how do you test whether people are really using that baseline compliance? What's a typical study setup? - 太迷人了。在你的实际实验中，你如何测试人们是否真的使用了那个基准合规性？典型的研究设置是怎样的？

03:03-03:30 - Alex: They're just, you know, usually kinda survey experimental studies. So one group gets the instructions, one group doesn't. They're provided with a fictitious, you know, vignette case of a violation that occurred. And they're asked to, you know, issue or or violation that someone's been investigated for, given the evidence for the case, and then asked to, you know, make a decision we look at, you know, both, you know, binary non-compliant or compliant penalty decision and like, a probability of non-compliance, um, to see, you know, kinda how people match up to that. - 通常就是那种问卷实验研究。一组人拿到了指令，另一组人没拿到。我们会给他们提供一个虚构的违规案例情境。在给定案例证据的情况下，要求他们针对正在调查的违规行为做出决定。我们会观察两种结果：一种是二元决策（合规或不合规的处罚决定），另一种是不合规的概率评估，以此来看人们的表现。

03:33-03:38 - Researcher: Got it. And what have you been finding – do the instructions actually make a difference in how people judge non-compliance? - 明白了。那你们发现了什么——这些指令真的会对人们判断非合规性产生影响吗？

03:41-05:20 - Alex: It's kinda TBD because our origin the way we analyze the data seemed like no. But I I think when I reanalyzed the data kinda looking at a different way, it does seem like there's some effects. I think part of it might have been the fact that the way we kind of validate our stimuli. So really, for this question, it's kind of almost like the instructions are telling you that if you think there's, a fifty one percent chance the person's non-compliant, then you probably shouldn't penalize them. And the law never really you know, dictates to the regulators how they should actually approach that or the court never dictates it. They keep it very vague, but that's kind of, I think, to most people what it means that, like, if you think it's just, like, a little more likely than not, that's probably not enough for a penalty. You have to be pretty sure. That means that the stimuli that we use, you know, these vignette cases where a corporation might be non-compliant, might be not, should probably kinda hover around some, like, you know, sixty five percent threshold or something where it's like, okay. It's probably non-compliant, but it's not a slam dunk case. And I think our vignettes might have just been too low in probability. They were kinda more, like, fifty fifty. In which case, you know, for half of people, you know, the instructions wouldn't make a difference because they'd say compliant anyway. So but but kinda looking at kind of that subset, it does seem like there's sort of a little bump for people who are around kind of this, like, elevated but not you know, really high level of you know, probability of non-compliance. So it's not entirely clear yet. - 结果还有待确定。因为按照最初分析数据的方式，似乎没影响。但我认为当我换一种方式重新分析数据时，似乎确实存在一些影响。我认为部分原因在于我们验证刺激物（案例素材）的方式。实际上，对于这个问题，这些指令几乎是在告诉你：如果你认为这个人有 51% 的可能性不合规，那你可能不应该处罚他。法律从未真正规定监管机构或法院应该如何处理这种情况，表达非常模糊，但我认为对大多数人来说，这意味着：如果你觉得只是“可能性稍微大一点”，那可能不足以做出处罚，你必须相当确定。这意味着我们使用的素材（即公司可能违规也可能不违规的情境案例）应该徘徊在 65% 左右的阈值，也就是“可能违规，但不是铁证如山”。我觉得我们的情境案例模拟的概率可能太低了，更像是五五开。在这种情况下，对一半的人来说指令没区别，因为他们反正会判定为合规。但看那一小部分特定数据，对于那些处于“概率较高但并非极高”区间的人来说，指令似乎确实产生了一点影响。所以目前还不完全清楚。

05:23-05:33 - Researcher: That makes sense – you're trying to hit that sweet spot where the instructions would actually matter. How much of your typical day is spent on this kind of research versus coursework or other PhD responsibilities? - 这很合理——你试图找到那个指令能真正发挥作用的“平衡点”。在你典型的一天中，有多少时间花在这种研究上，又有多少时间花在课程或其他博士职责上？

05:40-05:47 - Alex: Um, at this point, I'm not doing any coursework, but I am I - 嗯，目前我已经不修课了，但我有……

05:47-05:47 - Researcher: teaching. - 教学吗？

05:50-06:07 - Alex: hold on. I mean, I do this research, you know, I guess it's most of my time, then I also have my thesis that I'm working on. But I spend less of my time on that. So, like, most of my time is spent analyzing data, kinda just doing minutiae of lab tasks. Yeah. - 稍等。我是说，我做这类研究，大概占了我大部分时间，然后我还要准备毕业论文，但在那上面花的时间相对少一点。所以，我大部分时间都花在分析数据和处理实验室的琐碎任务上。是的。

06:09-06:21 - Researcher: Got it. So most of your day is data analysis and lab work. Now let's shift gears a bit – I want to talk about AI tools. What AI apps or web tools do you use most often in your life, and what's your experience with them been like? - 明白了。所以你的一天大部分是数据分析和实验室工作。现在让我们换个话题——我想聊聊 AI 工具。在你的生活中，你最常使用哪些 AI 应用或网页工具？你的使用体验如何？

06:34-07:41 - Alex: I use ChatGPT the most. I have the ChatGPT plus subscription, I think it's called. It's, like, the twenty dollar a month one. So that's my main one. Um, I use it as my kinda default and go to. I like the fact that they have projects in there so I can you know, kinda group things and it kinda helps me make it easier to find them later if I wanna go back to conversations. I do a lot of things just in the temporary chat, though, because I don't really wanna keep have them kinda clutter up all my other conversations. Also sometimes I just don't want it to use you know, memory of other conversations, so I just kinda wanna clean answer. But because I've been hearing really good things about Gemini, I've started using that a little more recently and I find that to be very good now, whereas it used to be pretty bad in the past. I used to use Claude a decent amount, especially before ChatGPT was able to give you PDF reading for free and before I had a ChatGPT subscription. Claude was kind of the first to do that for free. So that was really useful. I also find it's really good for coding but I tend to use ChatGPT more just because I kinda have everything on there. And if ChatGPT can't help me with my coding, then, you know, I move over to Claude. - 我用 ChatGPT 最多。我订阅了 ChatGPT Plus，应该是叫这个名，就是那个每月 20 美元的。所以那是我的主力工具，是我的默认选择。我喜欢它里面的“Projects（项目）”功能，我可以把东西分类关联，以后想回溯对话时更容易找到。不过很多事情我也直接在临时聊天（temporary chat）里处理，因为我不想让它们弄乱我其他的对话记录。有时我也不希望它使用其他对话的记忆，我只想要一个干净的回答。最近我也听说 Gemini 的评价非常好，所以也开始用得稍微频繁了一点，我觉得它现在非常好，而以前挺糟糕的。以前我还用过不少 Claude，特别是在 ChatGPT 还不支持免费读取 PDF 以及我还没买订阅之前。Claude 算是第一个提供免费 PDF 读取功能的，所以当时很有用。我也觉得它写代码非常好，但我还是更倾向于用 ChatGPT，纯粹是因为我所有的东西都在上面。如果 ChatGPT 帮不了我写代码，我才会转向 Claude。

07:42-07:51 - Researcher: You mentioned you have the ChatGPT Plus subscription at twenty dollars a month. Are you paying for any other AI tools, or is that your only paid subscription right now? - 你提到你订阅了每月 20 美元的 ChatGPT Plus。你还有为其他 AI 工具付过费吗？还是说这是你目前唯一的付费订阅？

07:51-07:52 - Alex: That's it. - 就这一个。

07:58-08:03 - Researcher: So ChatGPT Plus is your only paid AI tool. You mentioned using Gemini and Claude too – are those on free plans? - 所以 ChatGPT Plus 是你唯一的付费 AI 工具。你也提到了使用 Gemini 和 Claude——那些用的是免费版吗？

08:07-08:08 - Alex: Yep. - 是的。

08:10-08:15 - Researcher: What made you decide to pay for ChatGPT specifically but stick with free versions of the others? - 是什么让你决定专门为 ChatGPT 付费，而其他工具只用免费版？

08:16-08:51 - Alex: Uh, well, only need to pay for one. I just needed unlimited access or virtually unlimited You know, I never run out of messages with ChatGPT Plus. So there's really no reason to pay for more of them. And the reason I paid for ChatGPT plus at all was to, you know, not run into those bottlenecks. There's sort of limits around, you know, PDFs you can upload, I think there might have been other stuff at the time as well that was kind of a a bottleneck. And so it just made sense because I was using it a lot and wanted to use it more. So, yeah, I just you know, I upgraded. - 呃，我只需要给一个付钱。我只是需要无限访问，或者说实质上的无限访问。用 ChatGPT Plus 我从来没用完过消息额度。所以没理由给更多的工具付钱。我之所以买 Plus，就是为了不遇到那些瓶颈。比如上传 PDF 数量有限制，我记得当时可能还有其他限制构成了瓶颈。因为我当时用得很频繁，还想用得更多，所以买它是划算的。所以，我就升级了。

08:54-09:00 - Researcher: When did you first upgrade to ChatGPT Plus? Can you walk me through what was happening at that time that made you decide to pay? - 你最早是在什么时候升级到 ChatGPT Plus 的？你能跟我讲讲当时发生了什么让你决定掏钱吗？

09:12-09:16 - Alex: Um, probably about a year ago. I think it's kinda just what I just said. I was just running out of messages and wanted to kinda pull that bottleneck out. There might have been other features that weren't available yet, but I don't really remember. - 嗯，大概一年前吧。我想原因就跟我刚才说的一样。我当时频繁耗尽消息额度，想消除这个瓶颈。可能当时还有一些未开放的功能，但我不记得了。

09:18-09:24 - Researcher: You said you were running out of messages – what were you using it for so heavily that you hit those limits? - 你说你当时耗尽了消息额度——你主要是用它做什么，以至于达到了那些限制？

09:26-09:51 - Alex: Really anything in life that I just wanted information for. Was using it for, you know, coding in Python a lot. For, you know, my statistical work. I was using it just for, you know, general health questions, you know, you know, what food should I eat and what exercise is better. Using it just for, you know, information lookup. Questions about, like, relationships and social stuff and emotional stuff. Yeah. - 生活中任何我想获取信息的事情。我经商用它写 Python 代码，用于我的统计工作。我也用它处理一般的健康问题，比如该吃什么食物，什么运动更好。还有就是查资料。关于恋爱关系、社交问题、情感问题。是的，什么都问。

09:53-10:01 - Researcher: So it sounds like it became a go-to for pretty much everything. When you hit that message limit on the free version, did you upgrade right away or did you try anything else first? - 听起来它成了你大多数事情的首选工具。当你在免费版上达到消息限制时，你是立刻升级了，还是先尝试了别的方法？

10:03-10:27 - Alex: No. I was using you know, ChatGPT, Claude, Gemini or, you know, before that it was called Bard. So the way I would do it is just, you know, if I hit my limit on one, I would move over to the other. And then, you know, I used, like, temporary accounts or not temporary accounts, you know, another account. Yeah. That's kinda all I did. - 没有。我当时在混着用 ChatGPT、Claude、Gemini（之前叫 Bard）。我的做法是，如果一个达到上限了，我就换到另一个。然后我也用过临时账号——也不是临时账号，就是另一个账号。是的，那时候我就是这么干的。

10:29-10:36 - Researcher: So you were juggling between different tools and accounts for a while. What finally pushed you to just pay for it instead of continuing that workaround? - 所以你有一段时间在不同的工具和账号之间切换。最后是什么促使你直接付钱，而不是继续这种折中方案的？

10:46-11:22 - Alex: Don't really remember. I think I just either wanted to stop doing that or maybe there was other features of ChatGPT Plus that were available at the time that made it worth it. I don't really know if it's worth it now, but I saw that they had released a ChatGPT forget what it's called now, but it's, like, eight dollars a month. It's, like, their newest subscription model. And I'm actually set to you know, downgrade to that, you know, the next month. So I'm still on plus now, but once that comes, I'm gonna downgrade to that new version. - 不太记得了。我猜要么是我不想再这么折腾了，要么就是当时 ChatGPT Plus 有些功能让这个价格物有所值。我现在不确定它是否还值那个价，但我看到他们推出了一个（大概叫什么我不记得了）每月 8 美元的版本。那是他们最新的订阅模式。我其实已经准备好下个月降级到那个版本了。我现在还在 Plus 上，但等时间到了，我会降级到那个新版本。

11:24-11:29 - Researcher: Interesting – so you're planning to downgrade. What's making you feel like Plus isn't worth it anymore? - 有意思——你打算降级。是什么让你觉得 Plus 不再划算了？

11:32-11:56 - Alex: It's just that I get kinda everything I need with this lower tier subscription. Like, it has more messages. Like, I've never hit my message limit on plus. I don't think I've ever come close to it. So there's just no reason to be paying for that much, whereas I get pretty much everything I need with the, you know, downgraded subscription. It's I think there was, like, a couple of features I might lose that were kind of like, ah, that'd be nice to have. But to save twelve bucks a month, it's worth it. - 主要是那个低层级的订阅已经能满足我所有的需求了。比如它有更多的消息额度。我用 Plus 从未达到过消息上限，甚至没接近过。所以没理由付那么多钱，低配版订阅完全够用了。我觉得可能会损失一两个功能，那种感觉会有“啊，有这个功能当然更完美”的想法，但为了每月省下 12 刀，降级是值得的。

11:58-12:01 - Researcher: What are those features you'll lose that would be nice to have? - 那些你会损失、但“有则更好”的功能是什么？

12:03-12:03 - Alex: I don't remember. - 记不清了。

12:06-12:12 - Researcher: Fair enough. Besides ChatGPT, have you ever paid for any other AI tools in the past – even if you're not paying for them now? - 没关系。除了 ChatGPT，你过去付过其他 AI 工具的费用吗——即使现在没在续费了？

12:17-12:18 - Alex: Not to my knowledge. - 据我所知没有。

12:20-12:27 - Researcher: Got it. So thinking about your experience paying for ChatGPT Plus over the past year – would you say that's been a satisfying purchase overall? - 明白了。回想过去一年支付 ChatGPT Plus 的经历，你会说这总体上是一次令人满意的消费吗？

12:36-12:51 - Alex: Yeah. I mean, it's like two hundred forty bucks a year, so it's not small, but I think it's paid off. I really like having everything organized in my projects. And because I've had those high usage limits, enabled me to kinda put a lot into those. So, yeah, it's been worth it, but I'm very happy that I can now do eight dollars a month, which seems like a more fair price for what I need. - 是的。一年 240 美元，虽然不算小数目，但我觉得值得。我真的很喜欢在项目（Projects）里组织一切。正因为有高使用限额，我才能往里面投入很多。所以是有价值的，但我很高兴现在能选每月 8 美元的方案，这对我所需的功能来说是更合理的价格。

12:53-13:01 - Researcher: You mentioned liking the projects feature a lot. Can you give me a specific example of how you've organized something using projects that's been really helpful? - 你提到非常喜欢项目功能。你能给我举个具体的例子，说明你是如何利用项目功能来组织内容的吗？

13:17-14:02 - Alex: So one example is just that I just like having you know, all my conversations about, you know, econometric models together. And so, like, when I think about some conversation, I'm like, I need I wanna go back to that. It's just easy to find. And it's almost like a way of kind of archiving or favoriting them, which I'm I'm kind of annoyed that it doesn't have a favorite feature. Another thing I've done is, you know, I have to study for this kinda big exam in my third or fourth year of grad school. And you basically have a ton of readings you need to do and then questions on those readings. And so I've uploaded PDFs of, like, you know, a certain section of those readings into a project in ChatGPT and then I just have it, you know, give me, like, quiz questions based on it. That's very useful that you can have kind of a set of PDFs that you can just kinda automatically pull from without it also, like, contaminating everything you ask ChatGPT about, you know, stuff that in different projects. - 比如，我喜欢把所有关于计量经济模型的对话放在一起。当我想起某次对话，想“我要找回那个”时，很容易找到。这几乎成了一种存档或收藏的方式，其实我挺烦它没有“收藏”功能的。另一件事是，在读研的第三或第四年，我得准备一场大型考试。你要阅读大量的文献，然后回答相关问题。我把文献某个章节的 PDF 上传到 ChatGPT 的一个项目里，然后让它根据内容给我出测试题。这非常有用，你可以拥有一组特定的 PDF 供它自动调用，同时又不会污染你在其他项目里问 ChatGPT 的内容。

14:03-14:14 - Researcher: That's really helpful to understand. Now thinking more broadly – have you had any other AI purchases or subscriptions recently that felt particularly satisfying, or was ChatGPT Plus really the main one? - 这对我的理解非常有帮助。现在宏观地来看——最近还有其他让你感到特别满意的 AI 购买或订阅吗？还是说 ChatGPT Plus 就是唯一的重点？

14:16-14:18 - Alex: I haven't purchased any other AI tools. - 我没买过任何其他 AI 工具。

14:23-14:34 - Researcher: Alright, let's dive deeper into your experience with AI chat assistants specifically. You've mentioned ChatGPT, Gemini, and Claude. Are there any other chat assistants you use regularly, or are those your main three? - 好的，让我们深入聊聊你使用 AI 聊天助手的具体体验。你提到了 ChatGPT、Gemini 和 Claude。你还有其他经常使用的助手吗，还是主要就这三个？

14:36-14:38 - Alex: Those are the only ones I use regularly. - 经常用的就这三个。

14:40-14:44 - Researcher: Among those three – ChatGPT, Gemini, and Claude – which one would you say is your favorite overall? - 在 ChatGPT、Gemini 和 Claude 这三个中，综合来看你最喜欢哪一个？

14:52-15:00 - Alex: ChatGPT for sure. And then it used to be Claude, then Gemini. I've announced Gemini, Claude. - 肯定是 ChatGPT。以前第二是 Claude，然后是 Gemini。现在顺序变成 Gemini、Claude 了。

15:02-15:04 - Researcher: What makes ChatGPT your favorite right now? - 为什么 ChatGPT 目前是你最喜欢的？

15:15-15:51 - Alex: I just find it tends to have better answers, like, I tend to get less fluff. I haven't really played around with Gemini and Claude enough recently, so maybe I need to. But because I'm able to, like, customize the prompt style on ChatGPT, it kind of cuts through some of the fluff, gives more direct answers. And I generally just find that I get what I need more so than Gemini and Claude. Gemini tends to be too verbose. Claude is great for coding, but outside of that, you know, it doesn't have, like, the web search built in the same way, which is a very important feature that ChatGPT has. - 我觉得它给出的答案更好，废话更少。我最近还没怎么深入研究 Gemini 和 Claude，也许我该试试。但因为我可以在 ChatGPT 上自定义提示词风格，它可以过滤掉一些废话，给出更直接的回答。总的来说，比起 Gemini 和 Claude，我更容易拿到想要的结果。Gemini 往往太啰嗦。Claude 写代码很棒，但除此之外，它没有内置同样便捷的网页搜索，而这是 ChatGPT 的一个非常重要的功能。

15:53-15:58 - Researcher: You mentioned customizing the prompt style on ChatGPT – can you show me or describe what you've set up there? - 你提到在 ChatGPT 上自定义提示词风格——你能展示或描述一下你是怎么设置的吗？

16:02-17:03 - Alex: Sure. I I think of the things I tell it is, like, don't give intros and conclusions. Like, you know, if you ask it a question about, you know, you know, what is you know, the effect of, you know, you know, supplement a versus supplement b on heart health it'll usually start with something like many supplements can affect heart health, and it's important to understand what you know, is the best. And I kinda told, you know, don't do that. Just get straight to the answer, and then it kinda does the same thing at the end where it provides, like, this, you know, synthesis and conclusion at the end. And so I've kinda told it to avoid that. So that's kinda my overall, you know, messages to it. And then also, specifically for projects, you know, I'll tell it, like, you know, how I want maybe answers formatted or the types of things I want it to talk about. You know, with coding, I have, like, you know, certain ways of formatting my code. So if I'm asking it for suggestions, you know, I tell it to not do it certain ways and, you know, give it to me this way instead. Which just makes it easier. - 没问题。我告诉它的一些原则包括：不要给前言和结尾。比如，如果你问它“补充剂 A 和补充剂 B 对心脏健康的影响分别是什么？”，它通常会以“许多补充剂都会影响心脏健康，了解什么最合适非常重要”之类的话开头。我告诉它别这样，直接给答案。它在结尾也常会做总结陈词，我也告诉它避免这些。这是我总体的要求。此外，针对具体的项目，我会告诉它我希望答案如何格式化，或者我希望它讨论哪些方面。关于写代码，我有特定的代码格式。如果我向它征求建议，我会告诉它不要用某些方式写，而是按我的这种方式提供。这样就省事多了。

17:05-17:16 - Researcher: That's really specific and useful. Let's talk about Gemini for a moment – you said you've been using it more recently. What kinds of things are you using Gemini for these days? Can you give me a recent example? - 非常具体且有用。聊聊 Gemini——你说你最近用得更多了。你现在通常用 Gemini 做什么？能举个最近的例子吗？

17:20-17:24 - Alex: All the same stuff I use ChatGPT for, so coding life questions. - 和我用 ChatGPT 做的事情一样：写代码、问生活问题。

17:27-17:31 - Researcher: When you open Gemini instead of ChatGPT, what makes you choose Gemini in that moment? - 当你选择打开 Gemini 而不是 ChatGPT 时，那一刻促使你选择 Gemini 的原因是什么？

17:34-17:45 - Alex: It's usually just if ChatGPT didn't work, didn't didn't get me the answer, or if I wanna kinda triangulate between sources and see if I get consistent answers across sources. - 通常是因为 ChatGPT 没奏效，没能给我答案，或者我想在不同来源之间互证，看看不同工具给出的答案是否一致。

17:47-17:54 - Researcher: Got it. And with Claude – you mentioned it's great for coding. Can you walk me through a recent time you used Claude for a coding task? - 明白了。至于 Claude——你提到它写代码很棒。你能讲讲最近一次用 Claude 处理代码任务的情况吗？

17:56-18:03 - Alex: Not really. I can't remember any specific examples, but it's all just, you know, analyzing my survey data from my PhD. - 讲不出来。我想不起具体的例子，但都是些分析我博士课题调查数据的工作。

18:05-18:10 - Researcher: When did you first start using ChatGPT? Do you remember how you heard about it or what made you try it? - 你最早是在什么时候开始用 ChatGPT 的？你还记得是怎么听说它的，或者是什么让你想去尝试？

18:13-18:32 - Alex: Think I just heard about it through social media. Started using it. This would have been maybe the summer of no. Even before the summer of twenty twenty one. Spring spring of twenty twenty one, maybe, like, February or something. - 我想是通过社交媒体听说的。开始用它应该是……2021 年夏天？不对，比 2021 年夏天还早。可能是 2021 年春季，大概 2 月份左右。

18:35-18:41 - Researcher: Wait, February 2021? ChatGPT launched in late 2022. Do you mean February 2023, or were you thinking of a different tool? - 等等，2021 年 2 月？ChatGPT 是 2022 年底发布的。你是指 2023 年 2 月，还是说你指的是另一种工具？

18:45-18:47 - Alex: You're right. February twenty twenty three. - 你是对的。2023 年 2 月。

18:50-18:57 - Researcher: No worries. So February 2023 – when you first heard about it on social media, do you remember what people were saying about it that made you want to try it? - 没关系。那么 2023 年 2 月——当你第一次在社交媒体上听说它时，你还记得人们是怎么评价它的，让你想去尝试？

19:09-19:33 - Alex: People were just saying how good it was and how well it answered, like, a wide variety of topics. And I remember, like, one of the first things I did with it is I just, like, said, and and people were saying, like, you know, like, this is, like, the end of, like, you know, writing in high school. Like, you can't test kids on writing. And so I remember I just asked it, like, write me a five paragraph essay on, you know, the importance of the Industrial Revolution to the United States. It gave me just, like, a really impressive answer and probably one that was better than, you know, I would have written and it did it in, you know, twenty seconds. Yeah. - 大家都在说它有多好用，回答各行各业的问题有多出色。我记得我做的第一件事是，因为人们都在说“这是高中写作课的终结，你没法再考学生的写作了”。所以我记得我就问它：“给我写一篇关于工业革命对美国重要性的五段式散文。”它给出了一个非常令人印象深刻的答案，可能比我写的还好，而且只用了 20 秒。是的。

19:34-19:39 - Researcher: When you saw that essay, what was your reaction – were you excited, worried, something else? - 当你看到那篇散文时，你的反应是什么——兴奋、担心，还是别的？

19:42-19:48 - Alex: I think mostly excited that, you know, obviously it raised concerns, but it was obviously a major technological leap. - 我想主要是兴奋。当然它引发了一些担忧，但这显然是一个重大的技术飞跃。

19:52-19:57 - Researcher: Did you start using it right away after that first test, or did you wait a bit before really diving in? - 在那次首测之后，你是立刻就开始用了，还是等了一段时间才真正投入使用？

20:05-20:17 - Alex: I definitely started using it, and then it you know, gradually ramped up over time. You know, back then, it had lots of hallucinations and lots of things that it just couldn't really answer. And I also needed to learn how to communicate with it to, you know, get the sort of answers that were helpful. But I've learned all that over time. - 我肯定立刻就开始用了，然后使用量逐渐增加。那时候它还有很多幻觉，很多问题答不上来。我也需要学习如何与它沟通，以获得有用的答案。但我随着时间的推移学会了这些。

20:19-20:27 - Researcher: You mentioned learning how to communicate with it. Can you give me an example of something you used to ask that didn't work well, and how you changed the way you asked? - 你提到学习如何与它沟通。你能举个例子吗，比如你过去是怎么问的但效果不好，后来你又是怎么改变提问方式的？

20:36-20:38 - Alex: I can't really think of anything off the top of my head right now. - 我现在一下子想不起来具体的例子。

20:40-20:50 - Researcher: That's fine. So you started with the free version in early 2023, then upgraded to Plus about a year ago. What about Gemini and Claude – when did you first try those, and where did you hear about them? - 没关系。所以你是在 2023 年初开始用免费版，然后大约一年前升级到 Plus。那 Gemini 和 Claude 呢——你什么时候第一次尝试它们的，从哪听说的？

20:56-21:22 - Alex: Bard would have been fall of twenty three, I guess, and I don't remember where I heard about it. Claude, I heard like, early fall twenty twenty three from my PhD adviser had mentioned it. And then I started using that a fair bit, but the message limits were so low that I would have liked to use it more because I felt like it was better at the time. Especially with coding, especially because of the PDF uploads. But it just had so low limits that I barely used it. - Bard（后来的 Gemini）应该是 2023 年秋天，我不记得从哪听说的了。Claude 则是 2023 年秋天早些时候，我的博士生导师提到了它。然后我开始用得比较勤，但它的消息限额太低了，虽然我觉得当时它更好用，尤其是在写代码和 PDF 上传方面。但也因为限额太低，我用得很少。

21:24-21:29 - Researcher: When your adviser mentioned Claude, do you remember what they said about it or why they recommended it? - 当你的导师提到 Claude 时，你还记得他们是怎么评价它或为什么推荐它的吗？

21:31-21:35 - Alex: I just remember they just said they thought it was better, so I checked it out. - 我只记得他觉得那个更好用，所以我就去看了看。

21:39-21:44 - Researcher: You said Claude had really low limits back then. How many messages could you send before hitting the limit? - 你说 Claude 当时的限制非常低。你在达到上限前能发多少条消息？

21:57-22:28 - Alex: Oh, I don't remember, but I'd say maybe it was something like ten. It could have even been as low as five. I don't really remember, but I just remember it was frustrating because it felt like by the time I got into a conversation, like, I couldn't get any further. Especially because it wasn't just the messages, it was the context window. And so like, if I uploaded a PDF, I usually have to, a, cut it in half so it'd only be, like, you know, eight pages or something. You know, if it's like a paper for school that I wanted, like, a summary of. And then b, that would just take up so much of the context window that, like, I could only ask a couple questions after that. That was a little frustrating. - 噢，我不记得了，但大概也就 10 条左右。甚至可能低到 5 条。我不太记得了，只记得那让人很沮丧，因为感觉刚进入状态，对话就进行不分下去了。不仅是消息数量，还有上下文窗口。比如我上传一个 PDF，通常得先砍掉一半，只留个 8 页左右（如果是我想要总结的课业论文）。而且这一下就占用了太多的上下文空间，导致我之后只能问几个问题。那确实有点烦人。

22:30-22:43 - Researcher: That does sound frustrating. Let me shift gears a bit – thinking about your willingness to pay for ChatGPT going forward, on a scale of 0 to 10, where 0 is not willing at all and 10 is extremely willing, how would you rate your willingness to continue paying for it? - 那听起来确实令人沮丧。让我换个话题——考虑到你未来支付 ChatGPT 费用的意愿，在 0 到 10 分的范围内（0 表示完全不愿意，10 表示非常愿意），你会给继续付费的意愿打几分？

22:50-22:51 - Alex: I would say about six point eight. - 我会说大概 6.8 分。

22:54-22:56 - Researcher: What makes it a 6.8 specifically – not higher, not lower? - 为什么是 6.8 分？不是更高或更低？

23:21-23:35 - Alex: I think it's more of a question of just do I really need it? I was considering you know, before the eight dollar a month subscription was released, I was considering maybe after this month just eliminating, you know, my subscription completely and just you know, when I hit that usage limit on ChatGPT, just moving over to Gemini because I've just heard it's been doing really well. But when I saw the eight dollar a month one, I was like, okay. That's, you know, that's worth it. You know, ninety six bucks a year, you know, there's not really anything. So if it even helps me a little bit in my work, then, you know, it's good. You know? - 我觉得核心问题是我是否真的需要它。在那个每月 8 美元的订阅发布之前，我曾考虑月结后就完全取消订阅。如果 ChatGPT 达到上限，就直接转用 Gemini，因为听说它现在表现很好。但当我看到 8 美元那个方案时，我想，好吧，那值得。一年 96 美元不算什么。只要它对我的工作有一点点帮助，那就行。

23:39-23:44 - Researcher: So the $8 tier keeps you in, but if that didn't exist, you'd be okay just bouncing between free versions? - 所以 8 美元的档位挽留了你，但如果那个档位不存在，你也能接受只在各免费版之间切换？

23:52-23:55 - Alex: I think so. I would at least try it out. And then if, you know, I found out that wasn't working for me, then, you know, I might go back. But yeah, - 我想是的。我至少会去尝试一下。如果发现那样行不通，我可能再回来续费。但是的。

23:58-24:06 - Researcher: What would "not working" look like for you? Like, what would you notice in your day-to-day that would make you realize you need to go back to paying? - 对你来说，“行不通”会是什么样子的？在日常生活中，什么情况会让你意识到必须重新付费？

24:07-24:39 - Alex: Running out of messages. Anything with projects that, you know, made it not as useful, and especially if I'm hitting my message limits and now my content is spread across, you know, two different platforms. I'd rather just have it all on one. You know, just for my own kind of record keeping, essentially. And if ChatGPT or Gemini or Claude or some other tool, like, if one of them just, like, surpasses the others, like, I would wanna make sure I'm kinda spending most of my time on that tool and have full access to it. - 消息不够用。项目功能变得不好用。尤其是如果我达到了消息上限，导致我的内容分散在两个不同的平台上。为了记录方便，我更希望把东西都放在一个平台上。如果 ChatGPT、Gemini、Claude 或其他工具中有一个明显超越了其他工具，我会想确保我大部分时间都花在那个工具上，并拥有完整的访问权限。

24:41-24:48 - Researcher: Makes sense. You mentioned projects a couple times – if the $8 tier didn't include projects, would that change your decision to downgrade? - 有道理。你几次提到了项目功能——如果那款 8 美元的订阅不包含项目功能，会影响你降级的决定吗？

24:54-24:56 - Alex: Very good chance, but not definite. - 很有可能，但不确定。

24:59-25:03 - Researcher: What would you lose without projects that would make you hesitate? - 没有了项目功能，你会失去什么让你犹豫的东西？

25:06-25:21 - Alex: Uh, just a way of organizing information. So it's almost like my way of you know, being able to find stuff later. It's my way of organizing conversations around specific principles of responding that I want the AI to use. Yeah. - 噢，就是一种组织信息的方式。它几乎是我以后能找回资料的唯一途径。它也是我围绕特定响应原则来管理对话的方式，也就是我希望 AI 如何回应我的原则。是的。

25:25-25:35 - Researcher: When you think about paying for AI tools more broadly – not just ChatGPT – how do you feel about it? Does it feel like a normal thing to pay for, or does it still feel different from other subscriptions? - 当你宏观地思考为 AI 工具付费时（不只是 ChatGPT），你感觉如何？这感觉像是一件再正常不过的付费行为吗，还是说它和你其他的订阅服务不太一样？

25:38-25:39 - Alex: No. It feels normal. - 不，感觉很正常。

25:41-25:46 - Researcher: Is it the same as paying for, say, Spotify or Netflix, or does it feel different in some way? - 这和付 Spotify 或 Netflix 的钱一样吗，还是在某些方面感觉不同？

25:48-25:51 - Alex: I mean, feels better because I use it for work. - 我觉得感觉更好一些，因为我把它用于工作。

25:54-25:56 - Researcher: How so – what makes paying for a work tool feel better? - 怎么说呢——是什么让你觉得为办公工具付费更值得？

25:59-26:02 - Alex: Because I care more about work than other things. - 因为我比起其他事情，更在乎工作。

26:04-26:12 - Researcher: Got it. Thinking about all the AI products out there, do you have a monthly budget in mind for how much you're willing to spend total on AI tools? - 明白了。考虑到市面上所有的 AI 产品，你心里有没有一个关于每月愿意在 AI 工具上总共花费多少的预算？

26:13-26:17 - Alex: No. It's just if it brings me more value than I pay for it, it's good. - 没有。只要它带给我的价值超过我的支出，就是好的。

26:19-26:30 - Researcher: So there's no hard cap – it's just about whether each tool is worth it. If something amazing came out tomorrow at $50 a month, would you consider it, or is there a point where the price feels too high no matter what? - 所以没有硬性的上限——关键在于每个工具是否物有所值。如果明天出一款非常棒的产品，每月 50 美元，你会考虑吗？还是说无论如何，价格到了一定程度你就会觉得太高了？

26:35-26:36 - Alex: No. If it genuinely brought me more than fifty dollars a month in returns it's fine. - 不会。如果它每月确实能带给我超过五十美元的回报，那就没问题。

26:39-26:43 - Researcher: How do you measure that return – is it time saved, quality of work, something else? - 你如何衡量那种回报——是节省的时间、工作质量，还是别的什么？

26:49-27:06 - Alex: All those things, I mean, if I think, you know, if it cost me a thousand dollars a year, I would ask myself, you know, looking at what I do with the tool, what I'd do without it. Would I pay a thousand dollars for that, you know, bump. Whether that be increased productivity, more time saved, better health, whatever outcomes happen. - 所有这些。我是说，如果我觉得，你知道，如果每年要花一千美元，我会问自己，看着我用这个工具做的事，以及不用它我会怎么做。我愿意为那份提升支付一千美元吗？无论是生产力的提高、节省更多时间、更好的健康，还是任何产生的结果。

27:08-27:15 - Researcher: Can you think of a specific example where ChatGPT gave you that kind of return – where you thought, "Yeah, this was definitely worth the money"? - 你能想到一个具体的例子吗？就是 ChatGPT 给你带来了那种回报——让你觉得“是的，这钱花得绝对值”？

27:17-27:41 - Alex: I mean, there's just tons of stuff coding wise, like, I would not have figured out on my own or would have taken forever. You know, just using that instead of, like, you know, developer code forums, which is the old way. Just just so much better to get answers that way than, you know, just trying to send random messages online and scrolling through message boards, you know, custom tailored to you. So that just makes analysis so much quicker. - 我是说，在代码方面有太多例子了。比如，有些东西靠我自己是想不出来的，或者要花很长时间。你知道，用它来代替开发者代码论坛那种老办法。通过这种方式获取答案比在网上乱发信息或者翻帖子要好得多，你知道，它是为你量身定制的。这让分析变得快得多。

27:44-27:52 - Researcher: When you say "so much quicker," can you give me a rough sense – like, something that used to take you hours now takes minutes, or what's the difference? - 当你说“快得多”时，能不能给我一个粗略的感觉——比如，以前需要几小时的事情现在只需要几分钟，或者区别是什么？

28:14-28:18 - Alex: I would say, like, a lot of time, it's, you know, having my analysis time. But then, there are times where it's probably, you know, cutting it by yeah, hours to minutes because I would have had to, like, Google and Google and Google and find something that's not the right one. Why isn't this code working? We're now at just, like, I just type in ChatGPT, get it instantly, copy paste, and it works. - 我会说，很多时候它让我的分析时间减半。但也有一些时候，它可能会把时间，没错，从几小时缩短到几分钟，因为我本来不得不不停地谷歌、谷歌、再谷歌，最后还没找到对的那个。为什么这段代码跑不通？而现在，我只需要在 ChatGPT 里输入，即刻得到答案，拷贝粘贴，就能运行了。

28:19-28:31 - Researcher: That's a huge difference. Outside of work stuff, you mentioned earlier using it for health and relationship questions. Does that factor into how you think about the value, or is it mostly the work benefits that justify the cost? - 那确实是巨大的差异。在工作之外，你之前提到用它来咨询健康和情感问题。这些会影响你对价值的判断吗？还是主要是工作上的收益在支撑这笔费用？

28:34-28:41 - Alex: Mostly the work benefits You know, the other stuff is nice, but I don't know if I'd pay for it. - 主要是工作收益。你知道，其他那些东西也很好，但我不知道我是否愿意为之付费。

28:43-28:48 - Researcher: If you had a magic wand and could change one thing about ChatGPT right now, what would it be? - 如果你有一根魔杖，现在可以改变 ChatGPT 的一件事，那会是什么？

29:02-29:21 - Alex: I think I would still like to make it better at tailoring responses. I still find that even with my instructions, it doesn't always follow them, and it just you know, gives too much information sometimes. And doesn't always follow my coding format that I give it So it just needs to be a little better at that. - 我想我还是希望能让它在定制化回复方面做得更好。我发现，即使我有自定义指令，它也并不总是遵守，而且有时候给的信息太多了。它也并不总是遵循我给它的代码格式。所以它在这方面需要变得更好一点。

29:24-29:32 - Researcher: When it doesn't follow your instructions, what usually happens – does it revert to being verbose again, or mess up the code formatting, or something else? - 当它不遵循你的指令时，通常会发生什么——是又变得啰嗦了，还是搞乱了代码格式，或者别的什么？

29:34-29:35 - Alex: Yeah. Those two things. - 对，就是这两件事。

29:38-29:42 - Researcher: How often does that happen – like, is it every few conversations, or pretty rare? - 这种情况发生的频率有多高——是每隔几次对话，还是非常罕见？

29:51-29:53 - Alex: I would say yeah, every few conversations. - 我会说，是的，大概每隔几次对话就会发生。

29:55-30:00 - Researcher: And when that happens, do you just re-prompt it, or does it derail the whole conversation? - 当发生这种情况时，你是重新给提示词（re-prompt），还是说整个对话就被带偏了？

30:11-30:20 - Alex: Usually just reprompt it reprompt it, and for those things, it does well. You know? Other things where you try to get it to correct a wrong or bad answer, it doesn't do well. But generally, for those things, it will correct. It's just kind of annoying to have to manually ask it after I've already had in the instructions. - 通常只是重新提示它，重新提示它，在这些事情上它表现得很好。你知道吗？但在你想让它纠正错误或糟糕的答案时，它表现得不好。但总的来说，对于那些问题（指格式)，它是会纠正的。只是在我已经在指令中写好了之后，还要手动去要求它，这有点烦人。

30:23-30:31 - Researcher: On a scale of 1 to 5, where 1 is not at all unique and 5 is very unique, how unique do you think ChatGPT is compared to other AI tools like Claude and Gemini? - 以 1 到 5 分为标准，1 是完全不独特，5 是非常独特，你觉得 ChatGPT 与 Claude 和 Gemini 等其他 AI 工具相比，其独特性是多少？

30:36-30:38 - Alex: Two point three. - 2.3 分。

30:40-30:44 - Researcher: What keeps it from being a 1 – what does ChatGPT have that the others don't quite match? - 是什么让它不至于得 1 分——ChatGPT 有哪些是其他工具不太能比得上的？

30:48-31:01 - Alex: Projects, Canvas, web search, What's the other feature they have? Agent, which I haven't really used much of, deep research. Prompting and instructions. - Projects（项目）、Canvas（画布）、网页搜索，他们还有另一个功能是什么？Agent（智能体），虽然我还没怎么用过，Deep Research（深度研究）。还有提示词和指令功能。

31:03-31:11 - Researcher: Got it. And on content quality – how satisfied are you with ChatGPT's generated content on a scale of 1 to 5, where 1 is very dissatisfied and 5 is very satisfied? - 明白了。关于内容质量——你对 ChatGPT 生成内容的满意度是多少？同样是 1 到 5 分，1 为非常不满意，5 为非常满意。

31:16-31:17 - Alex: Four. - 4 分。

31:19-31:21 - Researcher: What would it take to get to a 5? - 怎样才能达到 5 分？

31:23-31:49 - Alex: Eliminate hallucinations, especially thing that happens, like, when it gets the wrong answer, it just, like, does not correct. So it's, if I at this point, I kinda know that if, you know, on first answer, it doesn't really get what I'm trying to ask it, like, it tends to not ever make the correction and get there. So being able to better understand context and when it's wrong and kinda make those adjustments would be good. - 消除幻觉，尤其是那种当它给出错误答案时，它就是不改正的情况。所以，到了现在，我基本知道如果它在第一次回答时没能真正理解我想问的内容，它往往就永远无法通过纠正来达到目的。所以，如果能更好地理解语境、发现错误并做出相应的调整，那就太好了。

31:51-31:58 - Researcher: Last rating – how satisfied are you with ChatGPT's value for money, 1 to 5, where 1 is very dissatisfied and 5 is very satisfied? - 最后一个评分——你对 ChatGPT 的性价比满意度是多少，1 到 5 分，1 为非常不满意，5 为非常满意？

32:09-32:10 - Alex: For the twenty dollar a month subscription, I would say I am a three point eight. - 对于每月 20 美元的订阅，我会说我是 3.8 分。

32:12-32:14 - Researcher: And for the $8 tier you're switching to? - 那对于你要转到的 8 美元档呢？

32:22-32:23 - Alex: Haven't tried it yet, so I I'm not so sure about the usage limit but I'm gonna guess I'm gonna feel it's like a four point three. - 还没试过，所以我不确定使用限制，但我猜我会觉得大概是 4.3 分。

32:25-32:33 - Researcher: That makes sense. Last question on this – where do you usually hear about new AI products? Like, how did you find out about Claude, or when new features launch? - 这很有道理。关于这个的最后一个问题——你通常从哪里听到新的 AI 产品？比如，你是怎么知道 Claude 的，或者是如何获知新功能发布的？

32:38-32:45 - Alex: I feel like it's just headlines. I hear people talking about it. social media, podcasts, wherever. - 我觉得就是看新闻头条。听到别人在谈论。社交媒体、播客之类的。

32:47-32:56 - Researcher: That's really helpful. We've covered everything I wanted to ask – thank you so much for taking the time to walk me through all of this. Your insights have been really valuable. - 非常有帮助。我想问的内容都涵盖了——非常感谢你花时间为我讲解这些。你的见解非常有价值。