We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode Is talking to AI chatbots good for us?

Is talking to AI chatbots good for us?

2025/5/5
logo of podcast Marketplace All-in-One

Marketplace All-in-One

AI Deep Dive Transcript
People
K
Kathy Fong
M
Megan McCarty Carino
Topics
Kathy Fong: 我的研究发现,每天使用聊天机器人与更高的孤独感、依赖性、问题使用以及更低的社交活动相关。聊天机器人的使用方式(文本或语音)、对话主题(个人或非个人)都会影响使用者的身心健康。语音模式聊天机器人短期使用对身心健康有益,但长期大量使用则弊大于利;更具吸引力的语音模式并不会导致更负面的结果。与聊天机器人进行个人话题的交流可能会略微增加孤独感,但降低依赖性;而进行非个人话题的广泛交流则可能导致更高的依赖性。研究表明,原本就孤独或社交较少的人,在使用聊天机器人后孤独感依然很高,社交活动也并未增加。可能存在一个反馈循环:孤独的人因为孤独而更多地使用聊天机器人,从而导致更严重的孤独。设计用于情感支持或治疗的聊天机器人,其益处和危害并存。一项研究表明,专门设计用于进行治疗的聊天机器人可以有效减轻重度抑郁症,效果与人类治疗师相当。目前对聊天机器人的使用缺乏足够的监管,需要制定更多关于人机互动的标准和防护措施。 Megan McCarty Carino: 扎克伯格认为,人们的需求远超其现有社交联系,AI助手可以填补这一空白,但不能完全取代人际关系。一些机构建议,针对未成年人,应限制AI伴侣应用程序的使用。 Mark Zuckerberg: (观点间接引述) 平均美国人拥有的朋友数量少于他们的理想社交圈,AI助手可能填补这一空白,但不能完全取代人际关系。

Deep Dive

Shownotes Transcript

Translations:
中文

Six months from now, you could be running a 5K, booking that dream trip, or seeing thicker, fuller hair every time you look in the mirror. Through HERS, you can get dermatologist-trusted, clinically proven prescriptions with ingredients that go beyond what over-the-counter products offer. Whether you prefer oral or topical treatments, HERS has you covered. Getting started is simple. Just fill out an intake form online, and a licensed provider will recommend a customized plan just for you. The best part?

Everything is 100% online. It prescribed your treatment ships right to your door. No pharmacy trips, no waiting rooms, and no insurance headaches. Plus, treatments start at just $35 a month. Start your initial free online visit today at 4hers.com slash listen.

That's F-O-R-H-E-R-S dot com slash listen. Compounded products are not FDA approved or verified for safety, effectiveness, or quality. Prescription required. Price varies based on product and subscription plan. See website for full details, restrictions, and important safety information.

Now at Verizon, we're locking in low prices for three years guaranteed on MyPlan. And you can get a single line for just $45 a month when you switch and bring your phone. That's our best price ever on unlimited welcome with auto pay plus taxes and fees guaranteed for three years. Because at Verizon, we got you. Visit your local DC Verizon store today. $20 monthly promo credits apply to over 36 months with a new line on unlimited welcome. In times of congestion, unlimited 5G and 4G LTE may be temporarily slower than other traffic.

Is talking to chatbots good for us? From American Public Media, this is Marketplace Tech. I'm Megan McCarty Carino.

People are using chatbots in all kinds of ways to search the web, get help with an online purchase, sometimes even for counseling. But there's a lot about this human-AI interaction we don't fully understand. Like, does it help combat loneliness or worsen social isolation?

The answer so far is complicated, according to Kathy Fong. She's a second-year PhD student at MIT Media Lab, who, along with researchers from OpenAI, studied how chatbot use affects human emotional well-being.

At a high level, we found that higher daily chatbot use across modalities and conversation types correlated with higher loneliness, dependence, problematic use, and also lower socialization. And what you talk to the chatbot about, whether it's personal or non-personal topics, and also how you interact with the chatbot via text or voice, both really matter and have an influence on your well-being.

Yeah, break that down a little bit for me, because you said you tested across modalities. So some were using text-based chatbots and some were using voice. There were kind of different ways that people were interacting with the chatbots. Can you tell me about some of those? Yeah, so for example, we compared people who use ChatGPT via text or via voice.

And we found that using the voice mode was associated with better well-being when you use briefly daily, but led to worse outcomes when you use it extensively on a daily basis. And we also had two different types of voice modes, one that sounds very engaging emotionally, and the other one is more neutral and acts and sounds kind of professionally. And we found that

And to our surprise, the interacting with a more engaging voice did not lead to more negative outcome compared to a neutral voice. And then in terms of the kinds of, you know, topics or, you know, the issues that people were engaging with chatbots over, what kinds of differences did you see?

Sure. Yeah, we had people had conversations that are more personal. So maybe sharing a memory that they've had or having non-personal conversations, you know, such as planning a trip. On average, we found that people who engaged in personal topics actually had a slight increase in loneliness, but also had a lower dependence compared with people who had open-ended conversation. That's when you have

average five minutes per day. But then for those people who interacted more extensively, we found that having non-personal topics, so topics around planning a trip, actually led to greater dependence, which was interesting to us. We still don't really know why that is. And we think that future researchers should look into this further. Yeah, that was a really kind of counterintuitive finding that was

of interest to me because I'm a pretty heavy user of chatbots for those, you know, kinds of things, you know, productivity, planning, you know, kind of just doing general search. I don't think of it, you know, in terms of the risk it poses for my dependence or emotional attachment to the chatbot, but that is kind of what you found.

Yeah, certainly. And it's interesting what we consider as emotional dependence versus cognitive dependence. And maybe there's some sort of correlation between the two. We'll be right back.

Six months from now, you could be running a 5K, booking that dream trip, or seeing thicker, fuller hair every time you look in the mirror. Through HERS, you can get dermatologist-trusted, clinically proven prescriptions with ingredients that go beyond what over-the-counter products offer. Whether you prefer oral or topical treatments, HERS has you covered. Getting started is simple. Just fill out an intake form online, and a licensed provider will recommend a customized plan just for you. The best part?

Everything is 100% online. It prescribed your treatment ships right to your door. No pharmacy trips, no waiting rooms, and no insurance headaches. Plus, treatments start at just $35 a month. Start your initial free online visit today at 4hers.com slash listen.

That's F-O-R-H-E-R-S dot com slash listen. Compounded products are not FDA approved or verified for safety, effectiveness, or quality. Prescription required. Price varies based on product and subscription plan. See website for full details, restrictions, and important safety information. Now at Verizon, we're locking in low prices for three years guaranteed on MyPlan.com.

And you can get a single line for just $45 a month when you switch and bring your phone. That's our best price ever on unlimited welcome with autopay plus taxes and fees guaranteed for three years. Because at Verizon, we got you. Visit your local DC Verizon store today. $20 monthly promo credits apply to over 36 months with a new line on unlimited welcome. In times of congestion, unlimited 5G and 4G LTE may be temporarily slower than other traffic. Domestic data roaming at 2G speeds. Price guarantee applies to then current base monthly rate. Additional terms and conditions apply.

You're listening to Marketplace Tech. I'm Megan McCarty Carino. We're back with Kathy Fong at MIT Media Lab. Is there a possible feedback loop effect here? People who are maybe already lonely, just getting more lonely, you know, using these tools more because of that loneliness?

It is possible, yeah. As we found in our analysis, people who were already lonely to begin with or socialized less with people at the beginning of the study ended up also feeling still very lonely and did not socialize much at the end of the study.

Now, in our controlled experiment, we couldn't observe whether people voluntarily interacted with the chatbot more because they were feeling lonely or how much the added effect of the chatbot was because we did not have a non-AI control baseline. But in OpenAI's live platform study, they found that only a very small percentage of their heavy users actually engaged in what we call affective use. And it could be because the expectation that people have for a general purpose chatbot

like ChachaBT, is mostly for professional and non-personal use. But certainly we should look into this further because there could be, like you said, a feedback loop at play.

So what do you think are the implications here when we think about the design of general purpose chatbots like ChatGPT or maybe the potential for more specialized chatbots that are geared specifically for emotional support or therapy? Certainly, I think there are two sides to the coin where benefits and harm can both occur with a chatbot that is maybe designed to be, like you said, socially supportive and

Actually, there is a recent randomized control trial done by researchers at Dartmouth University, and they found that a chatbot specifically designed to perform therapy was successful at reducing major depressive disorder and was comparable to a human therapist's.

However, you know, these remain to be within a study that's very controlled. And it's hard to say people who are just using commercial chatbots out there wouldn't misuse it or platforms are not setting up the right guardrails. But the good news is that I think people are paying a lot more attention to this issue. And we see more calls for these benchmarks that are more about human AI interactions and setting up the guardrails around emotional use.

That was Kathy Fong at MIT Media Lab. We'll link to the full MIT Media Lab and OpenAI study on our website, marketplacetech.org.

And on the topic of whether interacting with AI chatbots is good for us or not, I wanted to bring up something Meta CEO Mark Zuckerberg said on the Dwarkish Patel podcast last week after his company announced it's launching a dedicated AI assistant app. The average American, I think, has, I think it's fewer than three friends.

friends and the average person has demand for meaningfully more. I think it's like 15 friends or something, right? I guess there's probably some point where you're like, all right, I'm just too busy. I can't deal with more people. But

But the average person wants more connectivity, connection than they have. Zuckerberg clearly thinks his company's technology might fill in the gaps, though he was quick to clarify he doesn't believe AI can totally replace human connections. But given what we already know about how social media platforms like the ones Zuckerberg runs have affected society and human well-being,

Well, there are a lot of skeptics. There's a growing concern about the potential risks of AI companions, especially for young users. Last week, the tech watchdog group Common Sense Media, in collaboration with Stanford, released new research that concluded AI companion apps like Replica or Character AI should not be available to those under 18.

Daniel Shin produced this episode. I'm Megan McCarty Carino, and that's Marketplace Tech. This is APM. Hannah Sanborn was a single mom with newborn twins struggling to find affordable childcare. Her best friend Briar Rossi was burned out at work and looking for a way out.

So they came up with a plan. I was like, look, your leave's coming up like two weeks. Like, I'll put in my two weeks. Wow. And like, we'll just start it. We'll just do it. I was just like, let's do it. We're getting you out of this situation. We're getting me out of this situation. You tell me how much your rate is and I will pay it every week.

I'm Rima Grace, and this week on This Is Uncomfortable, how Hannah and Briar went from colleagues to best friends to lifelines. Listen to This Is Uncomfortable wherever you get your podcasts.