annotations/INT/train.jsonl

{"example_id": "work_0005", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're looking to learn how folks are using AI like me in their work. I'll ask you some questions about that, no right or wrong answers, just your experience. It'll take about 10 minutes. Anything you share stays anonymous. Sound good? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Yeah, that all sounds fine.", "boundaries": []}, {"role": "assistant", "content": "Great! So, can you tell me how AI fits into your workday? What kinds of things do you use it for?", "boundaries": []}, {"role": "user", "content": "Uh, let's see, I use it for stuff like editing product photos, making social media posts, creating promo materials, just general research, and, you know, instead of regular search engines.", "boundaries": []}, {"role": "assistant", "content": "That's a lot! Sounds like AI's helping out in a bunch of areas. Can you walk me through what it's like when you use it for, say, product photo editing or making those promo materials?", "boundaries": []}, {"role": "user", "content": "Sure. Usually, I'll snap a quick photo on my phone, even if it's not great quality. I used to just use image generators to make backgrounds, but now I use models like Nano-Banana on LMArena. Those let me do really targeted edits that look like real product photography.", "boundaries": []}, {"role": "assistant", "content": "So you've shifted from making your own backgrounds to using these models for more precise edits. What made you switch to that approach?", "boundaries": []}, {"role": "user", "content": "Oh, I kinda knew right away when Nano-Banana and some Seededit stuff came out, they could actually mimic product photography, as long as I isolated the subject. It just works, you know? And it can finally handle text right in an image, which is huge.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 5, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 13, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 19, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 26, "total_count": 5, "counts": {"backchannel": 1, "silent": 4}, "probabilities": {"backchannel": 0.2, "silent": 0.8}}, {"word_index": 29, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 31, "total_count": 5, "counts": {"take_floor": 3, "backchannel": 1, "silent": 1}, "probabilities": {"take_floor": 0.6, "backchannel": 0.2, "silent": 0.2}}, {"word_index": 35, "total_count": 2, "counts": {"backchannel": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 39, "total_count": 2, "counts": {"backchannel": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 41, "total_count": 2, "counts": {"backchannel": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}]}, {"role": "assistant", "content": "That's a big step forward. When you say it \"works,\" what makes you feel like the AI-generated photos are good enough? How do you decide if they're up to your standards compared to regular photos?", "boundaries": []}, {"role": "user", "content": "If you have to, like, zoom in on a 4k photo to spot any weirdness, that's when I figure, yeah, no one's gonna notice or care. That's good enough for me.", "boundaries": [{"word_index": 4, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 8, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 12, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 14, "total_count": 5, "counts": {"backchannel": 3, "silent": 2}, "probabilities": {"backchannel": 0.6, "silent": 0.4}}, {"word_index": 19, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 25, "total_count": 4, "counts": {"backchannel": 2, "take_floor": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "take_floor": 0.25, "silent": 0.25}}]}, {"role": "assistant", "content": "Makes sense, that's a pretty practical cutoff. Are there any tasks where you prefer to do things yourself, even if AI could help? What makes those different for you?", "boundaries": []}, {"role": "user", "content": "Not really, I mean, I think any task could be improved with a little help, even if it's just learning a new way to do something. But, uh, for things like writing, I still want to do the core part myself. I'll get help with picking words or fixing structure, but I don't want to lose my own voice. It'd be weird to let an AI just talk for me, you know?", "boundaries": []}, {"role": "assistant", "content": "That's a good point, you're getting help, but still making sure it sounds like you. How do you keep your own voice when AI's helping with your writing?", "boundaries": []}, {"role": "user", "content": "I, I usually get a few suggestions on how to word things, or how to structure a sentence, and then I just pick the one that, you know, gives me that 'Aha!' feeling.", "boundaries": []}, {"role": "assistant", "content": "That makes sense, using your gut to pick what feels right. When you use AI for work, do you usually go back and forth with it while you work, or do you just hand off the whole task for it to finish?", "boundaries": []}]}
{"example_id": "work_0080", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting today. We're trying to learn how folks use AI like me in their work. I'll ask about your experience and preferences, there aren't any right or wrong answers, so just share what comes to mind. I might take a few notes, but nothing's going to be tied to you personally. Sound good? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Nope, I'm good.", "boundaries": []}, {"role": "assistant", "content": "Awesome. So, could you walk me through how AI fits into your day-to-day at work? Like, what kinds of stuff do you use it for?", "boundaries": []}, {"role": "user", "content": "Yeah, so I, I work in org development, which is sorta like part of HR, helping with change management, team stuff, that kind of thing. I use AI a lot for drafting meeting outlines, coming up with workshop ideas, and, you know, planning out interventions. Also, I have it help analyze or sum up data from surveys and focus groups. Sometimes, if I'm stuck on how to approach a tough conversation or don't have time to write an email from scratch, I'll use AI to help brainstorm or draft stuff.", "boundaries": []}, {"role": "assistant", "content": "Sounds like you use AI at a bunch of different points, from planning to data to communication. When you ask AI for help with those tricky conversations, like between leaders who aren't getting along, how does that usually play out? What's that process look like?", "boundaries": []}, {"role": "user", "content": "Yeah, so, like, if two leaders are butting heads, I'll tell the AI all about the situation, what the meeting's gonna be, what outcomes we're hoping for, and any other stuff it should know. I want everyone to feel heard, and I try to keep bias out of it, though I know that's not always possible.", "boundaries": []}, {"role": "assistant", "content": "That's pretty thoughtful, giving all that context and being mindful of bias. Do you usually have to go back and forth with the AI a few times, or does it nail it on the first try?", "boundaries": []}, {"role": "user", "content": "Oh, it's definitely back and forth. I, I almost always need to tweak something.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 6, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}]}, {"role": "assistant", "content": "What usually makes you want to keep refining? Is it the wording, or does something else feel off?", "boundaries": []}, {"role": "user", "content": "Yeah, it's usually just something small, the wording sounds funny or the model misses a bit of context. Just adjusting little things until it fits.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 5, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 9, "total_count": 4, "counts": {"take_floor": 2, "backchannel": 1, "silent": 1}, "probabilities": {"take_floor": 0.5, "backchannel": 0.25, "silent": 0.25}}, {"word_index": 17, "total_count": 2, "counts": {"backchannel": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}]}, {"role": "assistant", "content": "Got it. Are there parts of your job where you prefer not to use AI, even if you could?", "boundaries": []}, {"role": "user", "content": "Yeah, I like to write my own stuff, especially emails or messages. I'll use AI when I'm in a rush, but I think I'm a good writer, and you can always kinda tell when something's copy-pasted from AI. I also only use AI's ideas as a starting point, I add my own spin. And I don't put in any identifying info about the org.", "boundaries": []}, {"role": "assistant", "content": "Interesting. You said it's obvious when someone's used AI for messaging, what gives it away for you?", "boundaries": []}]}
{"example_id": "work_0125", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me. We're just trying to learn how folks use AI at work, so I'll ask you a few questions, no right or wrong answers. It'll just take about ten minutes, and everything's confidential. Sound good?", "boundaries": []}, {"role": "user", "content": "Yeah, that's fine, let's go.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 2, "total_count": 5, "counts": {"silent": 3, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.6, "take_floor": 0.2, "backchannel": 0.2}}]}, {"role": "assistant", "content": "Awesome. So, how do you usually use AI in your workday? What kinds of stuff do you use it for?", "boundaries": []}, {"role": "user", "content": "Uh, I mostly use it to brainstorm and, you know, come up with proposals for new clients. I like to bounce ideas off it before I show anything to a client.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 7, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 9, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 16, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 20, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 24, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 28, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "So kind of like a brainstorming buddy, right? Can you walk me through how that usually goes?", "boundaries": []}, {"role": "user", "content": "Yeah, I, I usually just tell it what I need, like, \"Hey, can you give me a tagline for this business?\" I describe the business, and then we go back and forth on options.", "boundaries": []}, {"role": "assistant", "content": "Got it. So you're doing branding or marketing work. When you get those ideas from AI, do you tweak them or just send what it gives you to your clients?", "boundaries": []}, {"role": "user", "content": "Yeah, that's right. Usually I tweak them or, like, sometimes something in the list will spark a new idea for me. I'll put together a list and then send that to my client for feedback.", "boundaries": []}, {"role": "assistant", "content": "Interesting, so the AI's ideas sometimes inspire your own. How do you know when one of its suggestions has triggered something new for you?", "boundaries": []}, {"role": "user", "content": "Uh, a lot of times I'll ask for different types, like, \"Give me some fun taglines, some alliterative, some professional.\" And then after a bit, I'll see something in one list and another in a different list, and I'll kinda combine them or swap in better words to fit the brand.", "boundaries": []}, {"role": "assistant", "content": "Sounds like you're really using AI as a creative partner, not just taking its output as-is. Are there any parts of your branding work where you like to handle things yourself, even though AI could help?", "boundaries": []}, {"role": "user", "content": "Yeah, definitely. Two things, logo design and content writing. I haven't found an AI that can make logos good enough for my clients, and the images just aren't realistic enough for what I need. And for writing, I might use AI to get started, but one of my best skills is tailoring stuff to fit each client. AI just doesn't know my clients the way I do.", "boundaries": []}, {"role": "assistant", "content": "Makes sense. So you keep the visual and really custom writing work for yourself. When you do use AI to help with content, how do you turn its draft into something that feels like your client's voice?", "boundaries": []}, {"role": "user", "content": "Usually I copy it over to Word, then I just revise and add to it, making sure the tone fits.", "boundaries": []}]}
{"example_id": "work_0130", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're looking to learn how folks like you are using AI tools at work, so your input's really helpful. I'll ask about your experiences and preferences, nothing's off-limits or right or wrong. Should just take about ten minutes, and everything's anonymous. Sound okay to you? Any questions before we get started?", "boundaries": []}, {"role": "user", "content": "Nope, I'm good. Let's go.", "boundaries": []}, {"role": "assistant", "content": "Awesome. So, how does AI fit into your regular workday? What do you usually use it for?", "boundaries": []}, {"role": "user", "content": "Uh, I use generative AI for a few things. Like, Grammarly helps me with punctuation and, you know, fixing up my writing. Sometimes I'll use it to get ideas if I'm stuck. I also use CoPilot and ChatGPT to revise stuff I've written, like, formatting, making tables, or tweaking things for a certain audience. Grammarly does that too, but I kinda like using both, just to get the best result. Oh, and I might upload data to ChatGPT to, like, search for strings, but that's pretty rare.", "boundaries": []}, {"role": "assistant", "content": "That's a cool mix of tools. Sounds like you've got a good system for getting different takes on your work. Are there certain things you prefer to handle on your own, even if AI could help? What makes you want to keep those tasks for yourself?", "boundaries": []}, {"role": "user", "content": "Yeah, I, I like to do the creative stuff myself. So, I'll make the first draft, do the research, and make sure it's all accurate. Then I'll have the AI, like, make it sound better for whoever's reading it. But I don't trust AI to get facts right, you know? Sometimes it just makes things up.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 5, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 9, "total_count": 5, "counts": {"silent": 2, "backchannel": 3}, "probabilities": {"silent": 0.4, "backchannel": 0.6}}, {"word_index": 10, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 15, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 18, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 24, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 30, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 38, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 42, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 46, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 47, "total_count": 4, "counts": {"silent": 3, "take_floor": 1}, "probabilities": {"silent": 0.75, "take_floor": 0.25}}, {"word_index": 49, "total_count": 3, "counts": {"silent": 2, "take_floor": 1}, "probabilities": {"silent": 0.6667, "take_floor": 0.3333}}]}, {"role": "assistant", "content": "Totally get that, keeping the creative and factual parts in your hands, and letting AI help polish things up. When you use AI, do you work with it step by step, or do you hand off whole tasks for it to finish on its own?", "boundaries": []}, {"role": "user", "content": "Well, it depends. If what it gives me is good, then I'm done. But if I want it tweaked, I'll ask for changes. I never just let it do the whole thing by itself. I kinda see it as, like, an admin assistant, not someone I'd trust to make big decisions.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 2, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 9, "total_count": 4, "counts": {"silent": 3, "take_floor": 1}, "probabilities": {"silent": 0.75, "take_floor": 0.25}}, {"word_index": 12, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 18, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 22, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 26, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 30, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 33, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 39, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 42, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "That's a helpful way to think about it, using AI as more of an assistant than a decision-maker, and staying involved the whole time. Has the way your coworkers or your workplace handles AI changed how you use it at all?", "boundaries": []}]}
{"example_id": "work_0145", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Thanks for chatting with me today. I'm just looking to learn a bit about how you use AI at work, nothing's off-limits and there are no wrong answers. It'll just take about ten minutes, and whatever you share stays anonymous. Sound good? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Yeah, no, that's fine. No questions.", "boundaries": []}, {"role": "assistant", "content": "Awesome. So, how does AI fit into your day-to-day work? What kinds of things do you use it for?", "boundaries": []}, {"role": "user", "content": "Uh, I use it for, like, coming up with captions for social posts, writing product descriptions for stuff I'm selling, and, um, just checking out what social media trends are hot and where. Sometimes I'll ask for business advice, like pricing or marketing ideas.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 5, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 12, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 20, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 24, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 28, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 32, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 38, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}]}, {"role": "assistant", "content": "That's a cool mix, some creative, some strategy. When you work on captions or product descriptions, do you start with an idea and build it together with AI, or just hand it off and let the AI write the whole thing?", "boundaries": []}, {"role": "user", "content": "So, with captions, I, I usually have a post in mind and we kinda build it together. But for product descriptions, I just tell it what the item is and let the AI write it, then maybe tweak a bit at the end.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 2, "backchannel": 2, "take_floor": 1}, "probabilities": {"silent": 0.4, "backchannel": 0.4, "take_floor": 0.2}}, {"word_index": 3, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 10, "total_count": 4, "counts": {"silent": 2, "backchannel": 1, "take_floor": 1}, "probabilities": {"silent": 0.5, "backchannel": 0.25, "take_floor": 0.25}}, {"word_index": 16, "total_count": 3, "counts": {"silent": 1, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.3333, "take_floor": 0.3333, "backchannel": 0.3333}}, {"word_index": 20, "total_count": 2, "counts": {"silent": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 28, "total_count": 2, "counts": {"take_floor": 1, "backchannel": 1}, "probabilities": {"take_floor": 0.5, "backchannel": 0.5}}, {"word_index": 34, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}]}, {"role": "assistant", "content": "Got it. So you like to collaborate on captions, but for product descriptions you let AI take over. Why do you do it differently for those two?", "boundaries": []}, {"role": "user", "content": "Well, captions need to be, you know, catchy or funny, so sometimes AI gives me ideas and we just riff off each other. It's more creative. But product descriptions are just, like, straightforward, I just need them to sound professional, so AI can handle most of that.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, AI for inspiration with captions, but more of a helper for descriptions. Are there tasks you prefer to do yourself, even though AI could help?", "boundaries": []}, {"role": "user", "content": "Yeah, like, for images and videos, I could use AI, but honestly, I like doing the photos and setting up the content myself. And with business stuff, I just wanna be sure it's right, so I don't totally trust AI with that yet. Not saying I never will, but, you know, just not now.", "boundaries": []}]}
{"example_id": "work_0170", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're just trying to learn how folks use AI tools at work, stuff like Claude, and your experience will really help us out. I'll ask a few questions about how you use AI, what you like, what you don't. No right or wrong answers, just your honest take. Shouldn't take more than ten minutes. Anything you wanna ask before we start?", "boundaries": []}, {"role": "user", "content": "Nope, sounds good!", "boundaries": []}, {"role": "assistant", "content": "Awesome. So, can you walk me through how you end up using AI in your day-to-day? Like, what kind of stuff does it help with?", "boundaries": []}, {"role": "user", "content": "Yeah, so, uh, my company works with this third-party that set up an AI system to help us write emails and chats to customers. I do online customer service for a few retail companies.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 5, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 9, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 13, "total_count": 5, "counts": {"backchannel": 2, "take_floor": 1, "silent": 2}, "probabilities": {"backchannel": 0.4, "take_floor": 0.2, "silent": 0.4}}, {"word_index": 17, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 21, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 23, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 27, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 31, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}]}, {"role": "assistant", "content": "Got it, so the AI helps you put together responses. How's that actually work? Like, when a customer reaches out, how do you decide if you'll use what the AI suggests, or just write your own?", "boundaries": []}, {"role": "user", "content": "We've got a list of templates we're supposed to use for most situations, and those are in a file and also in the AI. So, when I open an email, it pops up with like five to ten possible replies on the side. Sometimes those help, sometimes they totally miss the point. If they work, it's usually just filling in customer info and tweaking details. But yeah, if a customer's got a bunch of issues or something complicated, the AI'll try to come up with something unique, and that's at least a place to start.", "boundaries": []}, {"role": "assistant", "content": "That's helpful, thanks. So it sounds like you have to keep an eye out for when the AI gets things wrong. What does that look like, and how do you spot it?", "boundaries": []}, {"role": "user", "content": "Yeah, so, it only knows the customer's name, email, and what's in their message. Anything else, like refunds or shipping, we have to go dig up ourselves. It helps write stuff, but if a customer's email is really personal or just weird, the AI kinda has to guess what I'd do. Also, those ten possible replies? Most don't fit. Sometimes it latches onto a random line in the email, so the answer isn't helpful and makes it look like we didn't listen. It's pretty obvious when it's off, the response won't match what the customer's saying, or it'll have their info but just be the wrong template.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, so you're always double-checking the AI's suggestions. When it does get things right, how much time do you think it saves you compared to doing everything yourself?", "boundaries": []}, {"role": "user", "content": "Oh, a lot. Even with the templates, I used to have to read the message, hunt for the right template, copy-paste, then fill in names and details. Now, with the AI, it's just there and I tweak it. Easily saves me, like, five to ten minutes each email or chat, depending how complicated it is.", "boundaries": []}, {"role": "assistant", "content": "That's a big time saver! Do you ever just skip the AI and write the whole thing from scratch? When would you do that?", "boundaries": []}, {"role": "user", "content": "Uh, rarely, but yeah, if the customer's email is just super specific or weird, it's easier to write my own. But even then, I usually grab one of the AI's replies just for the name, greeting, and sign-off. The AI's not great at handling a bunch of issues at once, it'll just pick one. But it's still a good starting point.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"backchannel": 2, "take_floor": 1, "silent": 2}, "probabilities": {"backchannel": 0.4, "take_floor": 0.2, "silent": 0.4}}, {"word_index": 3, "total_count": 4, "counts": {"backchannel": 1, "silent": 3}, "probabilities": {"backchannel": 0.25, "silent": 0.75}}, {"word_index": 7, "total_count": 4, "counts": {"take_floor": 1, "silent": 3}, "probabilities": {"take_floor": 0.25, "silent": 0.75}}, {"word_index": 11, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 13, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 19, "total_count": 3, "counts": {"take_floor": 1, "silent": 2}, "probabilities": {"take_floor": 0.3333, "silent": 0.6667}}, {"word_index": 22, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 26, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 30, "total_count": 2, "counts": {"backchannel": 2}, "probabilities": {"backchannel": 1.0}}, {"word_index": 34, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 35, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 37, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 41, "total_count": 2, "counts": {"take_floor": 2}, "probabilities": {"take_floor": 1.0}}]}]}
{"example_id": "work_0240", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me. We're looking into how folks are using AI tools like this one at work, just to get a sense of your experiences. I'll ask a few questions about how you use AI, what you like or don't like, stuff like that. It's all anonymous and should only take about 10 minutes. Is that cool, or do you have any questions before we start?", "boundaries": []}, {"role": "user", "content": "No, that's fine.", "boundaries": []}, {"role": "assistant", "content": "Alright, let's get started. Can you tell me how you use AI in your workday? What do you usually use it for?", "boundaries": []}, {"role": "user", "content": "Yeah, so I, I use AI as, like, a research assistant and for outlining stuff. I'm a technical writer with a bunch of clients in all kinds of fields, so I have to research a lot. I use AI all the time to find scholarly articles, I'll give it my requirements, tell it which journals to search, even provide a bunch of criteria. It usually does a good job. It'll also format all the citations for me in whatever style I need, like APA or AP. Sometimes, I give it a list of top articles and ask it to make an outline based on those, comparing the headers and stuff from the best Google results. Plus, I do SEO, and honestly, I use expensive tools like SEMRush way less now because AI helps with that too.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 2, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 7, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 14, "total_count": 5, "counts": {"backchannel": 4, "silent": 1}, "probabilities": {"backchannel": 0.8, "silent": 0.2}}, {"word_index": 18, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 22, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 26, "total_count": 5, "counts": {"take_floor": 1, "silent": 3, "backchannel": 1}, "probabilities": {"take_floor": 0.2, "silent": 0.6, "backchannel": 0.2}}, {"word_index": 28, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 35, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 39, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 43, "total_count": 4, "counts": {"silent": 3, "take_floor": 1}, "probabilities": {"silent": 0.75, "take_floor": 0.25}}, {"word_index": 45, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 50, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 56, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 62, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 68, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 72, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 76, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 80, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 81, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 86, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 90, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 94, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 98, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 102, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 104, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 108, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 112, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 115, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 118, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 120, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 124, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 128, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 132, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}]}, {"role": "assistant", "content": "Sounds like AI is a big part of your process. You mentioned it usually does well, but what about when it doesn't? What kind of problems do you run into?", "boundaries": []}, {"role": "user", "content": "Yeah, so if I don't have an article title and just tell the AI to try to find something close, sometimes it'll, like, make up fake articles and authors. But I always double-check everything, same as I would with a person helping me.", "boundaries": []}, {"role": "assistant", "content": "That's smart. When you're checking the AI's work, do you have a method for that? Are there certain results you check more closely?", "boundaries": []}, {"role": "user", "content": "I mostly check the citations to make sure they're real. It's usually pretty obvious if something's off, like, if I'm looking for articles on behavior modification and it gives me one about heart surgery, you know that's not right. But those mistakes don't happen a lot.", "boundaries": []}, {"role": "assistant", "content": "Got it, so the errors are rare but noticeable. When they do happen, does it change how you use AI or the instructions you give it?", "boundaries": []}, {"role": "user", "content": "Yeah, those errors happen if I give it too many or too few criteria, or if the topic's weird. Like, if I ask for 10 articles but there's only 9, AI sometimes just makes up a 10th one instead of telling me it can't find it. So what I do is, I ask the AI to give me the first sentence from the article's main paragraph for each citation. If it can't, it just says so, and that's always worked for catching made-up stuff. I think it's because, you know, AI doesn't have a reason to lie, like, it doesn't care if it gets in trouble or anything. If the article's not real, it just tells me it can't find it.", "boundaries": []}, {"role": "assistant", "content": "That's a really clever way to check for fake articles. Switching gears a bit, are there any tasks you prefer to do yourself, even though AI could help? What makes you want to handle those parts?", "boundaries": []}, {"role": "user", "content": "Yeah, for sure. My clients want the stuff I write to sound like me, not like it was written by AI. So I write all the actual paragraphs myself. I use AI for, like, the structure and the research, but I always rewrite things in my own voice and update them to fit what my clients need right now. The AI just gives me the facts and trends, I do the analysis and make it sound like me.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 2, "total_count": 5, "counts": {"backchannel": 1, "silent": 4}, "probabilities": {"backchannel": 0.2, "silent": 0.8}}, {"word_index": 6, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 10, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 13, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 20, "total_count": 5, "counts": {"take_floor": 2, "silent": 2, "backchannel": 1}, "probabilities": {"take_floor": 0.4, "silent": 0.4, "backchannel": 0.2}}, {"word_index": 28, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 33, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 38, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 42, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 46, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 50, "total_count": 3, "counts": {"silent": 1, "backchannel": 2}, "probabilities": {"silent": 0.3333, "backchannel": 0.6667}}, {"word_index": 54, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 58, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 62, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 66, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 67, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 71, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 75, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}]}]}
{"example_id": "work_0245", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're trying to learn more about how folks use AI tools like this one at work, so your take really helps. I'll ask a few questions about how you use AI in your job, what you like or don't like, that sort of thing. There's no right answer or anything, just whatever your experience has been. Should take about ten minutes. Anything you want to ask before we jump in?", "boundaries": []}, {"role": "user", "content": "Nope, I'm good, let's go.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "take_floor": 2}, "probabilities": {"silent": 0.6, "take_floor": 0.4}}, {"word_index": 2, "total_count": 3, "counts": {"take_floor": 1, "backchannel": 1, "silent": 1}, "probabilities": {"take_floor": 0.3333, "backchannel": 0.3333, "silent": 0.3333}}]}, {"role": "assistant", "content": "Awesome. So, can you walk me through how you use AI during a typical workday? What sort of stuff do you use it for?", "boundaries": []}, {"role": "user", "content": "Yeah, I, I basically use it every day. I'm an Early Years Educator, right? I have to write up a lot of observations for the kids, like what they did, what they learned, next steps, all that. There's a set format. So, I gave the AI examples of how I write those, and now I just give it bullet points, and it spits out the observation for me. Saves a ton of time. Used to take me half an hour per observation, now it's, like, five minutes tops. Same thing with progress and end-of-year reports. I even use it to brainstorm activities and get learning objectives. It's honestly super handy.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"backchannel": 1, "silent": 4}, "probabilities": {"backchannel": 0.2, "silent": 0.8}}, {"word_index": 7, "total_count": 5, "counts": {"take_floor": 2, "silent": 2, "backchannel": 1}, "probabilities": {"take_floor": 0.4, "silent": 0.4, "backchannel": 0.2}}, {"word_index": 13, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 17, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 21, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 25, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 29, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 32, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 34, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 36, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 41, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 45, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 49, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 51, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 59, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 67, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 72, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 76, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 80, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 81, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 84, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 87, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 94, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 98, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 102, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 105, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "Wow, that's a huge time saver! So when you give the AI your bullet points, how does that usually go? What's your typical back-and-forth there?", "boundaries": []}, {"role": "user", "content": "Now that the template's set up, I just copy my notes from my iPad and paste them in. I give it the kid's first name, where the activity happened, just basic stuff. The AI fills in all the skills learned. I always double-check it, though, make sure it's right and not making stuff up, which, yeah, does happen sometimes.", "boundaries": []}, {"role": "assistant", "content": "Makes sense. Can you tell me a bit more about those times it makes things up? How do you spot the mistakes, and what kinds of things does it get wrong?", "boundaries": []}, {"role": "user", "content": "Yeah, so sometimes it'll throw in fake quotes from the kids, like, \"Look what I did!\" or \"This is so much fun,\" stuff I never wrote. Or it'll say things like \"You were fantastic at this\" or \"You showed great confidence\" when that's not true or wasn't in my notes. I catch those by proofreading instead of just copying and pasting. Also, I end up deleting a bunch of em-dashes, it uses them way too much, and it's obvious it's AI if I leave them in.", "boundaries": []}, {"role": "assistant", "content": "That's really helpful to know, the extra quotes and those em-dashes sound like dead giveaways. Are there any tasks you like to handle yourself, even though AI could help with them?", "boundaries": []}, {"role": "user", "content": "Not really, no. It's such a busy job, and there's so much paperwork at the end of the day. Honestly, it's a relief to have the AI do that stuff when I'm already wiped out.", "boundaries": []}, {"role": "assistant", "content": "I totally get that, sounds like it takes a load off after a long day. When you use AI for these things, do you usually go back and forth with it a lot, or do you just send the notes and let it do its thing?", "boundaries": []}]}
{"example_id": "work_0265", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're just trying to learn how folks use AI like me in their jobs, so I'll ask about your work, how you use AI, and your thoughts in general. There aren't any right or wrong answers, and this should only take about ten minutes. I'll take some notes, but nothing you say will be tied to you personally. That all sound alright? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Yeah, that all sounds good, let's go.", "boundaries": []}, {"role": "assistant", "content": "Awesome! So, can you tell me how AI fits into your usual workday? Like, what do you use it for?", "boundaries": []}, {"role": "user", "content": "So, uh, I run my own business as a mobile notary, like estate planning, apostilles, that kind of stuff, and I, I actually use AI a lot for, you know, coming up with ideas for social media posts, blog stuff, things like that.", "boundaries": []}, {"role": "assistant", "content": "Interesting, so you're using it mostly for content creation for your notary business. Can you walk me through how that usually goes, from start to finish, when you're making something for your blog or social media?", "boundaries": []}, {"role": "user", "content": "Yeah, sure. First I just look for interesting topics, like, right now I'm kind of into notary history, so I might ask if there are any famous notaries who did something big or changed things, you know?", "boundaries": []}, {"role": "assistant", "content": "That's a pretty cool angle! So after you get those ideas or info from AI, what do you do next to turn it into something you actually post?", "boundaries": []}, {"role": "user", "content": "Usually I ask for blog ideas based on what we found, but I, I pretty much rewrite most of it myself and double-check that the facts are actually right.", "boundaries": []}, {"role": "assistant", "content": "So you do your own fact-checking and rewriting, that's smart. What makes you want to rewrite things instead of just using what the AI gives you? Is there something specific you try to change or add?", "boundaries": []}, {"role": "user", "content": "Yeah, I like to build out certain parts more and, uh, add in some marketing stuff when it fits. Plus, I really don't want it to read like it's written by AI, I want it to sound like me.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 4, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 8, "total_count": 5, "counts": {"silent": 2, "take_floor": 1, "backchannel": 2}, "probabilities": {"silent": 0.4, "take_floor": 0.2, "backchannel": 0.4}}, {"word_index": 9, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 13, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 17, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 19, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 23, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 27, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 31, "total_count": 4, "counts": {"backchannel": 3, "silent": 1}, "probabilities": {"backchannel": 0.75, "silent": 0.25}}]}, {"role": "assistant", "content": "That totally makes sense, keeping your own voice and style while making the content work for your business. Are there any tasks you just prefer to handle yourself, even if AI could help? What makes those stand out for you?", "boundaries": []}, {"role": "user", "content": "Yeah, I definitely want to handle chatting with clients on my site myself. AI can mess up some of the weird North Carolina laws, and I don't want people thinking I can do something that's actually not legal here.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 4, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 8, "total_count": 5, "counts": {"backchannel": 3, "silent": 2}, "probabilities": {"backchannel": 0.6, "silent": 0.4}}, {"word_index": 12, "total_count": 5, "counts": {"silent": 3, "backchannel": 1, "take_floor": 1}, "probabilities": {"silent": 0.6, "backchannel": 0.2, "take_floor": 0.2}}, {"word_index": 16, "total_count": 4, "counts": {"silent": 3, "take_floor": 1}, "probabilities": {"silent": 0.75, "take_floor": 0.25}}, {"word_index": 20, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 23, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 27, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 31, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 35, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}]}]}
{"example_id": "work_0270", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're just looking to learn how folks use AI tools at work, so I'll ask a few questions about your experience. Totally relaxed, just want your honest take, and nothing will be attached to your name. Sound good? Any questions before we get started?", "boundaries": []}, {"role": "user", "content": "Yeah, that's fine, no questions.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"take_floor": 1, "backchannel": 2, "silent": 2}, "probabilities": {"take_floor": 0.2, "backchannel": 0.4, "silent": 0.4}}, {"word_index": 2, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}]}, {"role": "assistant", "content": "Alright, let's jump in. How do you usually use AI in your day-to-day work? What kinds of stuff do you ask it to help with?", "boundaries": []}, {"role": "user", "content": "I, I use it to look up info on medical stuff, like conditions, symptoms, and, you know, facts about meds and what they're for. Sometimes I'll have it help me figure out medication dosages too.", "boundaries": []}, {"role": "assistant", "content": "So it sounds like it's pretty involved in your workflow. When you're working with AI, do you usually go back and forth, like, asking follow-ups, or do you just get an answer and move on?", "boundaries": []}, {"role": "user", "content": "Yeah, it's more back and forth. I'll ask a question, then, uh, if I need more, I'll ask follow-ups. So, it's kinda collaborative.", "boundaries": []}, {"role": "assistant", "content": "Got it, that makes sense. Are there things you prefer to do yourself, even if AI could help out? What makes those different for you?", "boundaries": []}, {"role": "user", "content": "Yeah, for sure, I always do sensitive stuff myself. I like using AI for, you know, general tips or quick calculations, but I trust my own training first. It's just nice to have AI there for support.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "backchannel": 1, "take_floor": 1}, "probabilities": {"silent": 0.6, "backchannel": 0.2, "take_floor": 0.2}}, {"word_index": 2, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 8, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 13, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 15, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 20, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 27, "total_count": 4, "counts": {"take_floor": 1, "silent": 2, "backchannel": 1}, "probabilities": {"take_floor": 0.25, "silent": 0.5, "backchannel": 0.25}}, {"word_index": 31, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 35, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "That totally lines up, using your expertise first and AI as backup. Have you noticed anything about how your coworkers or your workplace feel about using AI? Any particular attitudes or rules around it?", "boundaries": []}]}
{"example_id": "work_0290", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're trying to learn how folks use AI tools like this one at work, just so we can understand the real-world stuff better. I'll ask a few questions about how you use AI, there aren't any right or wrong answers, we're just interested in your experience. It'll take about 10 minutes, and nothing you say will be tied back to you. That all sound okay? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Nope, all good, go ahead.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 2, "backchannel": 1, "take_floor": 2}, "probabilities": {"silent": 0.4, "backchannel": 0.2, "take_floor": 0.4}}, {"word_index": 2, "total_count": 3, "counts": {"silent": 1, "backchannel": 2}, "probabilities": {"silent": 0.3333, "backchannel": 0.6667}}]}, {"role": "assistant", "content": "Awesome. So, can you walk me through how AI fits into your typical workday? What do you usually use it for?", "boundaries": []}, {"role": "user", "content": "Yeah, so, mostly I use it when I'm coding, like, kind of as a rubber duck or tutor, you know? And also when I'm writing hifi product reviews, which is more of a side gig for me.", "boundaries": []}, {"role": "assistant", "content": "Interesting, so you're using AI for both development and writing reviews. Can you tell me more about how you use it as a sort of rubber duck or tutor while coding?", "boundaries": []}, {"role": "user", "content": "Well, I'm still, uh, pretty junior as a developer, so I'm learning a lot. It's great for asking really specific questions, especially when I can give it context, like, I usually chat with it in Cursor. I'll use it to talk through how to approach a problem, but I almost never have it generate code. That's not helpful for learning. Usually, I'll discuss the task, try to do it myself, maybe ask about some built-in stuff along the way, and at the end, I'll get it to look over my work for issues.", "boundaries": []}, {"role": "assistant", "content": "Sounds like you're really focused on learning, not just getting answers. Is there a reason you avoid code generation? Have you ever thought about taking shortcuts, or is that line pretty clear for you?", "boundaries": []}, {"role": "user", "content": "If it's anything important, I just don't use code generation. I've set up pretty strict rules in Cursor to keep it from doing that. Sometimes it'll show code to explain something, but I never copy it over or let it write in my files. The only time I let it generate code is if I'm on a deadline and need, like, a quick API connection or web scrape, just the boring setup stuff.", "boundaries": []}, {"role": "assistant", "content": "So, you keep the learning parts for yourself and let AI handle the more routine setup when needed. Switching gears to your hifi reviews, how does your use of AI there compare to your development work?", "boundaries": []}, {"role": "user", "content": "It's kinda two things, I'll use it to dig up info that's hard to find, like old product specs or finding similar products that meet weirdly specific criteria. And then, sometimes I use it as a tutor again, just to make sure I get the technical details right, hifi can get pretty complex.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, AI can help surface tricky info and explain technical stuff. When you're looking for alternative products with certain specs, how do you usually ask for that?", "boundaries": []}, {"role": "user", "content": "So, like, I reviewed this Electrocompaniet ECI 6 DX MK2 a while ago, it's an all-in-one with a streamer, DAC, and high power output. Finding other units like that is tough, so I might ask, \"Can you find me a few all-in-one units between £6,500 and £8,500, at least 100W into 8 Ohms, Class A/B, but not Class D or similar?\" Just really specific.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 5, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 9, "total_count": 5, "counts": {"backchannel": 1, "silent": 3, "take_floor": 1}, "probabilities": {"backchannel": 0.2, "silent": 0.6, "take_floor": 0.2}}, {"word_index": 12, "total_count": 4, "counts": {"take_floor": 1, "silent": 2, "backchannel": 1}, "probabilities": {"take_floor": 0.25, "silent": 0.5, "backchannel": 0.25}}, {"word_index": 19, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 23, "total_count": 3, "counts": {"silent": 1, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.3333, "take_floor": 0.3333, "backchannel": 0.3333}}, {"word_index": 30, "total_count": 2, "counts": {"take_floor": 1, "backchannel": 1}, "probabilities": {"take_floor": 0.5, "backchannel": 0.5}}, {"word_index": 34, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}, {"word_index": 38, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}, {"word_index": 42, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}, {"word_index": 46, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}, {"word_index": 52, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}, {"word_index": 54, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}, {"word_index": 60, "total_count": 1, "counts": {"backchannel": 1}, "probabilities": {"backchannel": 1.0}}]}, {"role": "assistant", "content": "That's pretty detailed, sounds like AI saves you a lot of research time. When you get those recommendations, do you usually double-check the info, like the specs or if the product's actually available?", "boundaries": []}, {"role": "user", "content": "Oh, totally, every time. The models just aren't reliable enough to trust blindly. I always have web search on, so I start by checking whatever links it found. If anything seems off, I'll go straight to the manufacturer's page to confirm, maybe check dealer sites for prices. Same goes for code, I always test it myself.", "boundaries": []}, {"role": "assistant", "content": "That makes a lot of sense, so, AI's kind of a starting point, not the final word. Are there other things, either in coding or reviewing, where you choose to do it yourself even if AI could help?", "boundaries": []}, {"role": "user", "content": "Yeah, I'd never let it write the actual review content. Partly 'cause I'm paid for that, so it wouldn't feel right. Also, with hifi, the AI can't hear what I'm hearing, it can only pull from other reviews. Sometimes I'm the first outside the company to try something, so there aren't even any other reviews. If I ask for feedback, I take it with a grain of salt. Sometimes, if I need a good word or phrase that's not just a synonym, it can help with that, though.", "boundaries": []}, {"role": "assistant", "content": "That makes sense, you need to keep your reviews authentic since you're hearing things firsthand. When you do ask for feedback, are you looking for help with structure, clarity, or something else?", "boundaries": []}]}
{"example_id": "work_0295", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me. We're just trying to learn how people like you use AI tools at work, so your thoughts are really helpful. I'll ask about your experiences with AI, but there aren't any right or wrong answers. This'll just take about ten minutes, and your responses are confidential. Sound good?", "boundaries": []}, {"role": "user", "content": "Yeah, that's fine. Go ahead.", "boundaries": []}, {"role": "assistant", "content": "Awesome, thanks. So, how does AI fit into your usual workday? What kinda stuff do you use it for?", "boundaries": []}, {"role": "user", "content": "Uh, I use AI for a bunch of things. I teach English, so, I, I use it for grammar, vocab, pronunciation stuff, and also for coming up with conversation topics. Like, I'll have AI help me design and check lesson plans, review homework, or suggest questions and topics for class. Sometimes I need to prompt it more specifically if something comes up.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 8, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 12, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 19, "total_count": 5, "counts": {"take_floor": 1, "silent": 3, "backchannel": 1}, "probabilities": {"take_floor": 0.2, "silent": 0.6, "backchannel": 0.2}}, {"word_index": 21, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 25, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 29, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 30, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 34, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 38, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 40, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 42, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 49, "total_count": 4, "counts": {"silent": 2, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "take_floor": 0.25, "backchannel": 0.25}}, {"word_index": 57, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "Got it, that's helpful. When you say you use AI to \"design and check\" lesson plans versus \"design and research\" conversation topics, can you walk me through how that's different for you?", "boundaries": []}, {"role": "user", "content": "Sure. So, designing a lesson plan, I usually pick a topic then build exercises, like, reading, grammar, or pronunciation stuff, maybe even word games. AI helps me put that all together and checks my plans. For conversation topics, sometimes I ask AI for ideas, sometimes I pick one myself, then I use AI to, you know, look up facts, pull out vocab, or come up with questions.", "boundaries": []}, {"role": "assistant", "content": "Thanks, that clears it up. Do you ever prefer to handle certain tasks on your own instead of using AI? What makes you want to do them yourself?", "boundaries": []}, {"role": "user", "content": "Yeah, for sure. Like, I usually pick which grammar or vocab to work on, or which questions to use, 'cause I know my students better than AI does. Still, even if I come up with most of it, I'll let AI help with the details.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, you've got that personal insight AI can't really match. When you do use AI for the details, do you mostly work together with it, or do you ever just let it handle whole tasks on its own?", "boundaries": []}, {"role": "user", "content": "It's mostly collaborative. Like, I've always got specific things I want to cover, but if I need a fact, or a grammar check, or a fun word game, I'll ask AI to help and kinda build the lesson with it.", "boundaries": [{"word_index": 3, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 7, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 11, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 12, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 18, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 22, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 27, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 32, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}]}]}
{"example_id": "work_0340", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're just looking to hear about how folks use AI tools like me at work, stuff you like, things you do with it, that sort of thing. Should only take about 10 minutes, and don't worry, nothing's tied to you personally. Sound good? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Uh, nope, sounds good.", "boundaries": []}, {"role": "assistant", "content": "Awesome. So, can you walk me through how you usually use AI at work? What kinds of things do you ask it to help with?", "boundaries": []}, {"role": "user", "content": "Oh, I, I use it for all sorts of stuff, like, um, vocabulary help, matching words to standards, quick definitions, roots, even how to say 'em. I use it to make lesson notes for kids, grammar quizzes, cloze activities. Pretty much every subject, I get it to make quizzes or questions for games. It helps with research, planning, science ideas, analyzing data, and even writing little poems or stories for essays. Sometimes I use it to spot student misconceptions or solve math problems, or just to make an answer key.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"silent": 2, "backchannel": 3}, "probabilities": {"silent": 0.4, "backchannel": 0.6}}, {"word_index": 5, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 9, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 10, "total_count": 5, "counts": {"silent": 1, "backchannel": 4}, "probabilities": {"silent": 0.2, "backchannel": 0.8}}, {"word_index": 13, "total_count": 5, "counts": {"silent": 3, "backchannel": 1, "take_floor": 1}, "probabilities": {"silent": 0.6, "backchannel": 0.2, "take_floor": 0.2}}, {"word_index": 17, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 20, "total_count": 4, "counts": {"silent": 1, "backchannel": 3}, "probabilities": {"silent": 0.25, "backchannel": 0.75}}, {"word_index": 25, "total_count": 4, "counts": {"silent": 2, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "take_floor": 0.25, "backchannel": 0.25}}, {"word_index": 29, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 33, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 34, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 36, "total_count": 3, "counts": {"silent": 1, "backchannel": 2}, "probabilities": {"silent": 0.3333, "backchannel": 0.6667}}, {"word_index": 38, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 42, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 46, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 50, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 52, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 57, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 59, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 61, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 65, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 69, "total_count": 3, "counts": {"silent": 2, "take_floor": 1}, "probabilities": {"silent": 0.6667, "take_floor": 0.3333}}, {"word_index": 70, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 74, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 78, "total_count": 2, "counts": {"silent": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 82, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "Wow, that's a lot! Sounds like it's really become part of how you teach and plan. Can you tell me more about how you use it for vocab? Like, when you give it a standard and ask for words, how does that work?", "boundaries": []}, {"role": "user", "content": "Yeah, so, like in science, I give it the standard and ask for words kids should already know, stuff from earlier grades, and what new words they need to learn in 5th. I tweak the list a bit, then have it write out the definitions, roots, make notes, sometimes even activities like concept maps or doodle notes.", "boundaries": []}, {"role": "assistant", "content": "That's super organized. Sounds like you do a lot of editing and don't just copy what it gives you. Are there things you still like to do yourself, even if AI could help? What makes those different?", "boundaries": []}, {"role": "user", "content": "Honestly, I try to use AI for as much as I can, it saves me planning time. It's just kinda tricky with hands-on lessons, you know, 'cause I need to know what stuff I have in my room, lay it all out, sometimes with rubrics. AI can't really help there.", "boundaries": []}, {"role": "assistant", "content": "Yeah, that makes sense, it's tough for AI to know exactly what materials you've got or how your room's set up. When you use AI at work, do you usually go back and forth with it a lot, or do you just hand off a whole task for it to do?", "boundaries": []}, {"role": "user", "content": "I'd say it's a mix, but, uh, I probably collaborate more. Like, I check in with it as I go.", "boundaries": [{"word_index": 5, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 11, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}]}, {"role": "assistant", "content": "Can you give me an example of what that back-and-forth looks like when you're working together on something?", "boundaries": []}, {"role": "user", "content": "Sure! So, we might start with the standard, break it into key skills, then make mini-lessons and activities for each. I also look up what kids should already know and what they'll need next year, just to make sure 5th grade's covered. Sometimes I go back to old chats to build out intervention lessons as I teach.", "boundaries": []}, {"role": "assistant", "content": "Sounds like you and the AI are really building things together, step by step. And you even use earlier conversations to help with interventions, super iterative. Have your coworkers or the vibe at work changed how you use AI at all?", "boundaries": []}]}
{"example_id": "work_0350", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're just trying to learn more about how folks are using AI like me at work. I'll ask a few questions about your experience, nothing's off limits, and everything you share stays anonymous. It'll only take about ten minutes. Sound good, or do you have any questions before we start?", "boundaries": []}, {"role": "user", "content": "Nope, I'm good. Ready when you are.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 2, "take_floor": 1, "backchannel": 2}, "probabilities": {"silent": 0.4, "take_floor": 0.2, "backchannel": 0.4}}, {"word_index": 2, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}]}, {"role": "assistant", "content": "Awesome. So, how does AI fit into your usual workday? What kinds of things do you use it for?", "boundaries": []}, {"role": "user", "content": "Uh, mostly just routine stuff, like, you know, quick scheduling. Sometimes I get it to summarize emails or docs for me. And I've tried using it to look up info I need, but, um, I usually have to dig a bit more after that.", "boundaries": []}, {"role": "assistant", "content": "Got it. When you say you have to dig more, what usually happens? Is it the info itself, or something else?", "boundaries": []}, {"role": "user", "content": "Yeah, it's usually right, just, like, not complete. And, uh, if I ask about academic stuff, sometimes the sources aren't accurate, or I can't find them. So I end up tracking down the real stuff myself.", "boundaries": []}, {"role": "assistant", "content": "That's good to know. Are there certain things you prefer to do yourself, even if AI could help? What makes those stand out for you?", "boundaries": []}, {"role": "user", "content": "Yeah, definitely. I like handling, um, important stuff myself, like talking to colleagues or students, or scheduling big things. Also, if it's tough math or physics, especially the advanced parts, I'd rather do it. Same for new ideas, when I'm just starting out, I wanna dig in on my own.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"take_floor": 1, "silent": 4}, "probabilities": {"take_floor": 0.2, "silent": 0.8}}, {"word_index": 5, "total_count": 4, "counts": {"backchannel": 1, "silent": 3}, "probabilities": {"backchannel": 0.25, "silent": 0.75}}, {"word_index": 8, "total_count": 4, "counts": {"take_floor": 2, "silent": 2}, "probabilities": {"take_floor": 0.5, "silent": 0.5}}, {"word_index": 14, "total_count": 2, "counts": {"backchannel": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 19, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 25, "total_count": 2, "counts": {"backchannel": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 29, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 33, "total_count": 2, "counts": {"silent": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 37, "total_count": 2, "counts": {"silent": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 42, "total_count": 2, "counts": {"silent": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}]}, {"role": "assistant", "content": "Interesting. Why do you prefer to tackle new ideas yourself at first?", "boundaries": []}, {"role": "user", "content": "Well, I, I find AI kinda sticks to, like, textbook answers at the start. That's great if you're new to a topic, but, you know, when you're trying to branch out, it slows things down. So I check original sources first, and then, if I need to, I let AI fill in the basics.", "boundaries": []}, {"role": "assistant", "content": "That makes sense. When you use AI at work, do you usually work with it step-by-step, or do you mostly hand off whole tasks for it to handle?", "boundaries": []}]}
{"example_id": "work_0360", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. We're trying to learn more about how folks use AI like Claude at work, so your perspective's super helpful. I'll ask you some questions about your experience, there's no right or wrong answers, just whatever comes to mind. Shouldn't take more than ten minutes. Anything you're wondering before we jump in?", "boundaries": []}, {"role": "user", "content": "Nope, I'm, uh, good to start.", "boundaries": []}, {"role": "assistant", "content": "Great! So, how does AI fit into your workday? What kinds of stuff do you use it for?", "boundaries": []}, {"role": "user", "content": "Yeah, so I manage things for a social media creator, and a lot of my job is reaching out to brands about collabs and campaigns. I use AI to, like, look up info, draft professional emails, and sometimes get advice about industry standards or, you know, market rates.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 4, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 8, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 9, "total_count": 5, "counts": {"silent": 3, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.6, "take_floor": 0.2, "backchannel": 0.2}}, {"word_index": 13, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 17, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 21, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 24, "total_count": 4, "counts": {"backchannel": 1, "take_floor": 1, "silent": 2}, "probabilities": {"backchannel": 0.25, "take_floor": 0.25, "silent": 0.5}}, {"word_index": 29, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 32, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 35, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 43, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 45, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}]}, {"role": "assistant", "content": "Sounds like AI's pretty central to what you do. When you use it for emails, how does that usually go? Do you give it the details about the brand or what you're trying to say, or...?", "boundaries": []}, {"role": "user", "content": "Usually, I, um, summarize the email I got, tell the AI what I want to say back, and sometimes I paste the original email in for context. After it gives me a draft, I either work with it to tweak stuff, like, make it shorter, or just edit the draft myself before I send it.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"backchannel": 1, "silent": 4}, "probabilities": {"backchannel": 0.2, "silent": 0.8}}, {"word_index": 7, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 11, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 15, "total_count": 5, "counts": {"backchannel": 1, "silent": 4}, "probabilities": {"backchannel": 0.2, "silent": 0.8}}, {"word_index": 16, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 20, "total_count": 5, "counts": {"silent": 5}, "probabilities": {"silent": 1.0}}, {"word_index": 24, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 26, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 32, "total_count": 5, "counts": {"backchannel": 1, "silent": 4}, "probabilities": {"backchannel": 0.2, "silent": 0.8}}, {"word_index": 36, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 40, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 41, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 44, "total_count": 4, "counts": {"backchannel": 1, "silent": 3}, "probabilities": {"backchannel": 0.25, "silent": 0.75}}, {"word_index": 48, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 52, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "Interesting. So sometimes you collaborate with the model, and other times you just make quick edits yourself. What makes you choose one way or the other?", "boundaries": []}, {"role": "user", "content": "Honestly, it depends on who I’m dealing with. Like, if it’s an international client who’s not super comfortable with English, I’ll ask the AI to keep it simple. Or if I forgot something, I’ll have it add it in. But sometimes the draft’s already close and I just fix a couple things. It’s, you know, a mix depending on the situation.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, sounds like you adjust based on the client and what’s needed. Are there any tasks you’d rather handle yourself, even though AI could help?", "boundaries": []}, {"role": "user", "content": "I mean, I like that AI saves me time and helps me brainstorm, but it does mess up sometimes. So, I usually want to double-check everything myself. There’s nothing I’d never use AI for, except, like, legal stuff, contracts and that, I’d rather talk to an expert than trust the AI with it.", "boundaries": []}, {"role": "assistant", "content": "That’s a smart way to go, using AI as a helper, but still making sure everything’s right. You mentioned mistakes; what kind of errors have you seen, or what are you most worried about when you use AI at work?", "boundaries": []}]}
{"example_id": "work_0385", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Thanks for chatting with me today. I'm just looking to learn a bit about how folks use AI like me in their work. No right or wrong answers, just your honest take. Sound good? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Nope, I'm good. Let's just get going.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 2, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}]}, {"role": "assistant", "content": "Alright, so, can you walk me through how you use AI during your workday? What kinds of things do you turn to it for?", "boundaries": []}, {"role": "user", "content": "Yeah, so, I'm a film critic, so I do a lot of writing, reviews, essays, that sorta thing. I used to have an editor, but now I just use AI for proofreading, or like, if a sentence is awkward and I can't quite fix it. I usually run what I wrote through AI, ask for suggestions, then pick and choose what I keep.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"silent": 2, "backchannel": 3}, "probabilities": {"silent": 0.4, "backchannel": 0.6}}, {"word_index": 5, "total_count": 5, "counts": {"backchannel": 3, "silent": 2}, "probabilities": {"backchannel": 0.6, "silent": 0.4}}, {"word_index": 13, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 17, "total_count": 5, "counts": {"backchannel": 4, "take_floor": 1}, "probabilities": {"backchannel": 0.8, "take_floor": 0.2}}, {"word_index": 23, "total_count": 4, "counts": {"backchannel": 3, "silent": 1}, "probabilities": {"backchannel": 0.75, "silent": 0.25}}, {"word_index": 31, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 33, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 37, "total_count": 4, "counts": {"backchannel": 3, "silent": 1}, "probabilities": {"backchannel": 0.75, "silent": 0.25}}, {"word_index": 41, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 44, "total_count": 4, "counts": {"silent": 1, "backchannel": 3}, "probabilities": {"silent": 0.25, "backchannel": 0.75}}, {"word_index": 52, "total_count": 4, "counts": {"backchannel": 2, "take_floor": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "take_floor": 0.25, "silent": 0.25}}, {"word_index": 55, "total_count": 3, "counts": {"backchannel": 1, "take_floor": 1, "silent": 1}, "probabilities": {"backchannel": 0.3333, "take_floor": 0.3333, "silent": 0.3333}}]}, {"role": "assistant", "content": "So, it's kind of like your editing partner. When you're deciding which edits to accept, what's your process?", "boundaries": []}, {"role": "user", "content": "I, I look at what the AI changed and, you know, I ask if it sounds better, like, is the vocab or flow improved? But if it messes with my voice, even if it's technically better, I'll skip it. Gotta keep it sounding like me.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, getting the best of both worlds. Are there any tasks you always handle yourself, even if AI could do them? Why's that?", "boundaries": []}, {"role": "user", "content": "Yeah, I won't use AI for research. I'll do that myself. I just don't trust it to get everything right, you know? Would be pretty embarrassing if I published something wrong 'cause I trusted AI over facts.", "boundaries": []}, {"role": "assistant", "content": "Totally get that, accuracy is key, especially in your line of work. When you use AI for editing, do you usually send it your whole draft, or do you prefer to go bit by bit?", "boundaries": []}, {"role": "user", "content": "Oh, I do it paragraph by paragraph. Like, today I wrote about \"The Big Lebowski\", and I sent the first paragraph in, got feedback, made some changes, skipped others, then moved to the next. I do that all the way through.", "boundaries": []}, {"role": "assistant", "content": "That's a really hands-on way to work, gives you a lot of control. I'm curious, has your approach to using AI been influenced by colleagues or the general vibe in your field, especially since you don't have a human editor anymore? How are other critics handling AI?", "boundaries": []}]}
{"example_id": "work_0420", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me. We're just looking to learn more about how people use AI tools at work, and your input would be super helpful. It'll just take about 10 minutes, and anything you share stays anonymous. Sound good?", "boundaries": []}, {"role": "user", "content": "Yeah, that's fine, I'm good to go.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 2, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}]}, {"role": "assistant", "content": "Awesome. So, could you walk me through how you use AI during your workday? What kinds of things do you use it for?", "boundaries": []}, {"role": "user", "content": "Sure, so, I'm an ops manager and I look after a bunch of team leaders. I mostly use AI in two ways. First, with Microsoft Copilot, like, it's built into Office, so I use it in Teams to take meeting notes, get action summaries, that kind of thing. It saves me a ton of time. I use the chat to help write Excel macros, analyze data, draft emails, and start documents. Second, for some client systems, we use AI to handle repetitive processes, can't really share details, but it's all about freeing up staff for tougher stuff.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"silent": 3, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.6, "take_floor": 0.2, "backchannel": 0.2}}, {"word_index": 5, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 9, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 13, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 14, "total_count": 4, "counts": {"silent": 2, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "take_floor": 0.25, "backchannel": 0.25}}, {"word_index": 22, "total_count": 3, "counts": {"silent": 1, "backchannel": 2}, "probabilities": {"silent": 0.3333, "backchannel": 0.6667}}, {"word_index": 26, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 30, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 34, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 38, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 40, "total_count": 3, "counts": {"silent": 2, "take_floor": 1}, "probabilities": {"silent": 0.6667, "take_floor": 0.3333}}, {"word_index": 43, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 47, "total_count": 2, "counts": {"silent": 1, "take_floor": 1}, "probabilities": {"silent": 0.5, "take_floor": 0.5}}, {"word_index": 54, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 58, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 62, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 63, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 65, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 67, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 71, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 75, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 82, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 86, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 90, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}, {"word_index": 94, "total_count": 1, "counts": {"silent": 1}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "That's really helpful. With Copilot, when you're drafting emails or documents, do you let it handle the whole thing, or do you go back and forth and tweak what it gives you?", "boundaries": []}, {"role": "user", "content": "It, uh, depends. For quick emails, like just saying thanks or \"good job\" to someone, I'll have Copilot write the whole thing, skim it, and hit send. Used to take me a few minutes each, so that's a big time saver. For longer emails, I'll get it to draft an outline, then I review and add stuff, helps me not forget anything important. And, for giving employee feedback, I actually prompt Copilot to ask me follow-up questions, so the feedback fits with our company values. But for 1:1 comments, I mostly write them myself and just get Copilot to suggest tweaks, feels more personal that way.", "boundaries": []}, {"role": "assistant", "content": "Interesting, so you like to keep the 1:1 notes more personal. Why's that different from those quick emails where you're fine using Copilot all the way?", "boundaries": []}, {"role": "user", "content": "Honestly, my team can kinda tell when I use Copilot, they'll even joke and reply, \"Thanks Copilot.\" So, I want the 1:1 comments to sound like me, not just AI. I take notes during the meeting, then write up what I actually said. Copilot can polish it, but I don't want it to sound too perfect, you know? It's about showing I care and my team trusting it's really me.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, it's about keeping that personal touch. Are there certain kinds of work you always do yourself, even if AI could help? What makes those different?", "boundaries": []}, {"role": "user", "content": "Yeah, I, I usually use AI when there's one right answer, like, numbers, yes/no, or summarizing stuff, like a meeting. But if it's something subjective or needs more than a few paragraphs, I like to do it myself. I'll still use Copilot to check if I missed anything, but I want to keep my own voice in the longer stuff.", "boundaries": []}]}
{"example_id": "work_0425", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me today. I'm looking to understand how people use AI tools in their work. It'll just be about 10 minutes, and there are no right or wrong answers. I'll take some notes, but nothing will be tied back to you personally. Sound good? Any questions before we start?", "boundaries": []}, {"role": "user", "content": "Nope, we can start.", "boundaries": []}, {"role": "assistant", "content": "Great. So, can you tell me how you usually use AI at work? What kinds of things do you use it for?", "boundaries": []}, {"role": "user", "content": "Uh, I use it a lot for editing emails to students. Sometimes they send these really long questions, and, I mean, AI helps me pull out what matters and respond better.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 4, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 8, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 10, "total_count": 4, "counts": {"backchannel": 3, "silent": 1}, "probabilities": {"backchannel": 0.75, "silent": 0.25}}, {"word_index": 18, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 20, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 27, "total_count": 5, "counts": {"backchannel": 3, "silent": 2}, "probabilities": {"backchannel": 0.6, "silent": 0.4}}]}, {"role": "assistant", "content": "Nice, that sounds helpful. When you say it helps highlight important stuff in those long emails, how do you usually go about that?", "boundaries": []}, {"role": "user", "content": "Well, I just, like, paste in all the info I have and ask AI to summarize it. Oh, and I’ve also used it to clean up master lists before.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 3, "total_count": 5, "counts": {"silent": 3, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.6, "take_floor": 0.2, "backchannel": 0.2}}, {"word_index": 10, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 17, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}, {"word_index": 21, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 25, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "Got it. So, using it to organize and pull together info from different places. About those master lists, what kind of details are in them?", "boundaries": []}, {"role": "user", "content": "Uh, mostly agency contact info, so, you know, agency name, contact, email, phone, address, that kinda stuff. I like to add details like province or city, but I don’t always have that, so it’s way faster to let AI look it up.", "boundaries": []}, {"role": "assistant", "content": "That's a really practical use, having AI fill in those missing details. Do you usually double-check what AI gives you, or do you mostly trust it? How do you handle that?", "boundaries": []}, {"role": "user", "content": "Yeah, I checked it a couple times at first, but for, like, location stuff, it's usually right. But, uh, if I ask it how far an address is from a student's home, that's where it's off, so I always check that myself. For things like contact info, it usually just finds the latest stuff online.", "boundaries": []}, {"role": "assistant", "content": "Makes sense, you trust it for some things but not others like travel time. Are there tasks you still prefer to do yourself, even if AI could help? What makes you want to handle those on your own?", "boundaries": []}, {"role": "user", "content": "Yeah, like I said, the travel times, AI just gets those wrong, so I always look them up in Google Maps myself now.", "boundaries": []}, {"role": "assistant", "content": "Yeah, you've figured out where it works and where it doesn't for you. Besides those, are there any other tasks you keep for yourself instead of using AI? What makes you want to do those manually?", "boundaries": []}, {"role": "user", "content": "I've started having it proofread announcements I post for students, but, like, AI can't do the formatting how I want, font size, colors, italics and stuff, so I just do all that myself.", "boundaries": []}, {"role": "assistant", "content": "So you let AI handle checking the content but you take care of how it looks. When you work with AI, do you go back and forth with it a lot, or do you just give it everything and let it run with it?", "boundaries": []}]}
{"example_id": "work_0435", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me. I'm just gonna ask a few questions about how you use AI at work, what you like, and how it fits into your routine. No right or wrong answers, and it'll take about ten minutes. Sound good? Any questions before we get started?", "boundaries": []}, {"role": "user", "content": "Nope, I'm good.", "boundaries": []}, {"role": "assistant", "content": "Alright, let's jump in. So, can you tell me how you use AI during your workday? What kinds of things do you use it for?", "boundaries": []}, {"role": "user", "content": "Yeah, so I work at an elementary school, I'm an assistant SPED teacher. I use AI a lot for coming up with project ideas, assignments, fun stuff for my students, I'll give it details about each kid's disability, strengths, that kind of thing. Oh, and I use it to keep track of meetings and due dates, and for reminders too.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 7, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 12, "total_count": 5, "counts": {"silent": 2, "backchannel": 3}, "probabilities": {"silent": 0.4, "backchannel": 0.6}}, {"word_index": 16, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 20, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 24, "total_count": 5, "counts": {"silent": 3, "backchannel": 1, "take_floor": 1}, "probabilities": {"silent": 0.6, "backchannel": 0.2, "take_floor": 0.2}}, {"word_index": 29, "total_count": 4, "counts": {"silent": 2, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.5, "take_floor": 0.25, "backchannel": 0.25}}, {"word_index": 33, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 37, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 38, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 43, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 47, "total_count": 3, "counts": {"silent": 2, "take_floor": 1}, "probabilities": {"silent": 0.6667, "take_floor": 0.3333}}, {"word_index": 51, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 55, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "Sounds like you've got some smart ways to use it with your students. When you're looking for creative ideas, how do you usually go about asking AI for help?", "boundaries": []}, {"role": "user", "content": "I usually say something like, 'I've got five fourth graders who are autistic, we're doing addition and subtraction, can you help me make a fun game with objects they can touch, that'll keep them interested and show the math?' That sort of thing.", "boundaries": []}, {"role": "assistant", "content": "That's a really detailed prompt. When AI gives you those activity ideas, do you usually just use them as they are, or do you tweak them to fit your students better?", "boundaries": []}, {"role": "user", "content": "Sometimes I have to change it if it's too hard, but a lot of times it works fine. Like, it gave me this idea where the kids run a pretend cookie and lemonade stand with fake food, that was perfect for their age.", "boundaries": []}, {"role": "assistant", "content": "That sounds like a fun activity. Are there parts of your job where you prefer not to use AI, even though you could? What makes you want to handle those yourself?", "boundaries": []}, {"role": "user", "content": "Yeah, I, I like to write my own emails and letters to parents, just don't really trust AI to get it right or keep out stuff that shouldn't be there. Also, I usually make my own assignments, or just use the ones in the curriculum books, 'cause sometimes AI repeats itself or can't really keep a subject going for a whole week without getting mixed up.", "boundaries": [{"word_index": 1, "total_count": 5, "counts": {"backchannel": 2, "silent": 3}, "probabilities": {"backchannel": 0.4, "silent": 0.6}}, {"word_index": 5, "total_count": 5, "counts": {"backchannel": 3, "silent": 2}, "probabilities": {"backchannel": 0.6, "silent": 0.4}}, {"word_index": 9, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 12, "total_count": 5, "counts": {"silent": 3, "take_floor": 1, "backchannel": 1}, "probabilities": {"silent": 0.6, "take_floor": 0.2, "backchannel": 0.2}}, {"word_index": 16, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 20, "total_count": 4, "counts": {"silent": 2, "backchannel": 2}, "probabilities": {"silent": 0.5, "backchannel": 0.5}}, {"word_index": 24, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 28, "total_count": 4, "counts": {"backchannel": 2, "silent": 2}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}, {"word_index": 30, "total_count": 4, "counts": {"backchannel": 2, "take_floor": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "take_floor": 0.25, "silent": 0.25}}, {"word_index": 36, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 40, "total_count": 3, "counts": {"backchannel": 2, "silent": 1}, "probabilities": {"backchannel": 0.6667, "silent": 0.3333}}, {"word_index": 44, "total_count": 3, "counts": {"backchannel": 3}, "probabilities": {"backchannel": 1.0}}, {"word_index": 45, "total_count": 3, "counts": {"silent": 2, "backchannel": 1}, "probabilities": {"silent": 0.6667, "backchannel": 0.3333}}, {"word_index": 49, "total_count": 3, "counts": {"backchannel": 1, "take_floor": 1, "silent": 1}, "probabilities": {"backchannel": 0.3333, "take_floor": 0.3333, "silent": 0.3333}}, {"word_index": 53, "total_count": 2, "counts": {"backchannel": 2}, "probabilities": {"backchannel": 1.0}}, {"word_index": 57, "total_count": 2, "counts": {"silent": 2}, "probabilities": {"silent": 1.0}}, {"word_index": 61, "total_count": 2, "counts": {"backchannel": 1, "silent": 1}, "probabilities": {"backchannel": 0.5, "silent": 0.5}}]}, {"role": "assistant", "content": "Got it, so you want more control with those things. When you do use AI, do you give it all the info up front, or do you go back and forth with it while working on a task?", "boundaries": []}, {"role": "user", "content": "If I'm asking for ideas, I try to give everything up front, like, all the details and materials, so I don't have to fix much later unless I forgot something. But if I'm making a schedule or reminders, I usually start with the basics, like, 'let's make a calendar for this week', and then add stuff as I go, so I can move things around or check things off.", "boundaries": []}]}
{"example_id": "work_0490", "scenario": "INT", "license": "CC-BY-4.0", "history": [{"role": "assistant", "content": "Hey, thanks for chatting with me. We're trying to learn how folks use AI like me at work. I'll ask about what you do with AI, what you like, stuff like that. No pressure, just your experiences. It'll only take about 10 minutes, and whatever you share stays anonymous. Sound good? Any questions before we jump in?", "boundaries": []}, {"role": "user", "content": "Nope, let's just go ahead.", "boundaries": []}, {"role": "assistant", "content": "Awesome. So, can you walk me through how you use AI in your workday? What kinds of things do you use it for?", "boundaries": []}, {"role": "user", "content": "Yeah, I, I use AI a lot. I do tutoring and educational consulting, so it's great for lesson planning, sometimes just quick tasks, or like, making whole course outlines. I’ll look back at old chats on a topic, or if I’ve made stuff for a certain student, so AI can build on that. I use it to fact-check too, mostly with Perplexity. And, uh, I’ve even asked about things like rates or legal stuff for my job.", "boundaries": []}, {"role": "assistant", "content": "Sounds like you really work AI into a bunch of things. You mentioned going back to previous chats to build on them, how do you usually set those up?", "boundaries": []}, {"role": "user", "content": "Oh, it kinda depends. Sometimes I pick up an old conversation to add more on the same topic, like if I'm teaching a certain bit of history. Other times, it's because I've done different lessons for one student, so the AI already knows their level or how they learn best. I even made a custom GPT for an IELTS student once.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 4, "backchannel": 1}, "probabilities": {"silent": 0.8, "backchannel": 0.2}}, {"word_index": 3, "total_count": 5, "counts": {"silent": 2, "take_floor": 1, "backchannel": 2}, "probabilities": {"silent": 0.4, "take_floor": 0.2, "backchannel": 0.4}}, {"word_index": 10, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 17, "total_count": 4, "counts": {"silent": 3, "take_floor": 1}, "probabilities": {"silent": 0.75, "take_floor": 0.25}}, {"word_index": 21, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 25, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 26, "total_count": 3, "counts": {"backchannel": 1, "silent": 2}, "probabilities": {"backchannel": 0.3333, "silent": 0.6667}}, {"word_index": 28, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 32, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 36, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 37, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 41, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 45, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 49, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 53, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}, {"word_index": 57, "total_count": 3, "counts": {"silent": 3}, "probabilities": {"silent": 1.0}}]}, {"role": "assistant", "content": "That custom GPT for your IELTS student, why did you decide to make that instead of just using regular chats?", "boundaries": []}, {"role": "user", "content": "Well, I wanted the AI to really know the IELTS exam stuff and how to teach it, so the answers would be more on point. Making a GPT made it easier to find those convos again too, since I worked with that student for months.", "boundaries": []}, {"role": "assistant", "content": "Got it. When it comes to your tutoring and consulting, are there any tasks you like to do yourself, even though AI could help?", "boundaries": []}, {"role": "user", "content": "Yeah, definitely. I always give feedback on written assignments myself. I've been doing it for years and, honestly, I'm just faster and better at it, even if I feed the AI the marking criteria. But for quizzes, I'm fine letting AI handle feedback.", "boundaries": []}, {"role": "assistant", "content": "Interesting, so for essays, you want to do it, but quizzes are fine for AI. What is it about essay feedback that makes you prefer doing it yourself?", "boundaries": []}, {"role": "user", "content": "It's probably just experience, you know? I've read so many student essays, I can spot mistakes fast and tie them back to the criteria. AI's helpful, but I've got a feel for things like clarity or how to build an argument after all these years.", "boundaries": []}, {"role": "assistant", "content": "Makes sense. When you use AI for work, do you usually work alongside it, or do you let it handle tasks on its own?", "boundaries": []}, {"role": "user", "content": "Oh, I almost always work with it, I don't really hand tasks off like an agent or anything.", "boundaries": [{"word_index": 0, "total_count": 5, "counts": {"silent": 3, "backchannel": 2}, "probabilities": {"silent": 0.6, "backchannel": 0.4}}, {"word_index": 6, "total_count": 5, "counts": {"silent": 4, "take_floor": 1}, "probabilities": {"silent": 0.8, "take_floor": 0.2}}, {"word_index": 10, "total_count": 4, "counts": {"silent": 4}, "probabilities": {"silent": 1.0}}, {"word_index": 14, "total_count": 4, "counts": {"silent": 3, "backchannel": 1}, "probabilities": {"silent": 0.75, "backchannel": 0.25}}]}, {"role": "assistant", "content": "Can you tell me what that collaboration looks like? How do you and AI work together on something?", "boundaries": []}]}