Two things can be true at the same time. We are talking a lot about artificial intelligence these days, and there's still a lot we don't know about it. One researcher at the University of Arkansas is currently working on better understanding a specific area of AI, how people are building relationships with it.
Marco Dehnert is an assistant professor of communication and technology at the UofA. He's also the director of the AI and Robots for Connection Lab. He says there's a long history of Hollywood movies like "2001: A Space Odyssey," or "Her," presenting a variety of outcomes when it comes to humans and AI.
Dehnert: What's really interesting about these is that they always seem to reflect some of the cultural anxieties that we're currently experiencing at this particular time. Let's look at "Her," for example. The movie came out in 2013, so well before ChatGPT was put out on the markets. But we did have smart speakers and other types of smart devices. So people were thinking about, okay, what might be the next logical step in this evolution?
And I think similarly to happened with HAL in "2001: A Space Odyssey," people were imagining what it might be like to talk to something that actually ends up becoming a someone. And what if that someone develops motivations, intentions, and maybe even a will on its own?
So I think all these movies really reflect quite well the cultural anxieties that people have about technology right now, about what it might look like tomorrow. But obviously these are like dramatizations, right? Like Samantha is like this super smart AI assistant that then takes over and basically really guides this relationship in different kinds of ways. And I think has like 8,000 or so relationships at the same time, which is realistic. If we look at more realistic examples nowadays, such as Replika and other types of like companion AI. But obviously it's dramatized and really playing into our sense of liking good stories that are really engaging.
Moore: You oversee the AI and Robots for Connection Lab at the University of Arkansas. What sort of work happens here for you and for the students who are part of the lab, too?
Dehnert: So we're basically social scientists. We don't build devices. We don't program them necessarily. We're not engineers. We're interested in behavior. Like, what do people do with different kinds of devices? And that's why it's called the lab for connection, because we're really interested in studying connection.
We have a group of undergraduate and master's students from across campus that are working with me on different kinds of research projects. We're basically looking at, for example, how AI companions allow us to connect with different people. A recent line of research that we've been pursuing is about grief and how technology can support us in grieving, or maybe hinder us in grieving processes. How is AI changing and impacting grieving processes?
Moore: Why are people turning to AI for grief?
Dehnert: That's a really good question. I think technology has always been part of grieving. We can kind of look back decades and centuries to see how technology has been involved in grieving processes, like trying to memorialize the dead on like a Facebook memorial page or even Tumblr. Right? There's a lot of good stuff out there that talks about why people would want to do that. And in that sense, it's the next logical step to turn to the next new technology, which is AI.
And we can think about how we can use AI for grieving processes, whether that is to use it as a mirror to help us reflect some of the memories back that we had with a particular person, maybe actually to create new memories about that person, which kind of gets us into these more morally ambiguous kind of terrains of like, well, I have these voice memos of my mom who passed away two years ago. What if I uploaded it into a voice synthesizer and let my mom speak new words that are not on that recording?
Or what if I ended up using artificial intelligence not just to help me therapize myself through the process, but to actually embody the deceased person that I'm grieving, right? Like, what if I upload all of my texts and all of my chats with my mom into a grief bot, which is what we call these types of devices? This bot then takes on the persona of the deceased person and kind of mirrors back to me, hey, I'm your mom now and you can chat with me.
That's very much happening out there in different kinds of spaces. There's like very famous, like anecdotal news articles about these stories. Different startups are pushing into the space of this death and dying industry. And we're trying to understand, well, what does that actually do for us?
Usually when I bring up this topic with people, I get a variety of different kinds of responses of like, you know, shock or questioning. Some people seem to be excited about this, but at the end of the day, we don't really know what it does to us. Like, what does it do for someone who's lost their mother yesterday to then upload everything into a chatbot and interact with this replica of their mother online? What about if it was a tragic death, or a sudden death, or maybe even a death that was expected after like a chronic illness?
There's a lot of open questions, and I'm trying to move towards some of those answers in that space, because I think grief is a very particular kind of connection, and technology is affecting it in many different ways. We just don't know what it does actually to us as people.
Moore: Grief is an interesting emotion too, because there is perhaps like an allowed, allotted amount of time you can grieve something. And once that time passes, you might receive some stigma or you might receive some pushback that like, why are you still grieving this? Why are you still so sad about this? And I can't help but think that that probably plays a role in why you would turn to AI, because AI is not going to give you guff for being sad about your mother passing away.
Dehnert: Absolutely. Yeah. Oftentimes, especially in the US context, we speak of a death aversive culture. Like you're not supposed to grieve long. You're certainly not supposed to, you know, be sad for a long time because you got to go back to work. You got to be a productive member of society. You have people to take care of. You have to earn money to actually bring food on the table. Right? There's very much like this aversion to death in this culture that we're currently in.
And that brings up a lot of these questions that you're mentioning, right? Like, well, how long is healthy grieving? How long is good grieving? How intense should that be? Oftentimes, people try to pinpoint that at certain markers, such as a few weeks or a few months, but that's actually very difficult to pinpoint.
And that can actually also lead to a variety of different problems, right? Like maybe someone who grieves a little longer because the grief was very traumatic, then starts to feel like something's wrong with them, even though they're processing this very healthy, very human emotion, but they're then being told by a cultural context, actually, you have to kind of move on from this particular experience.
And then, yes, you're right, this can result in people internalizing it and then not bringing it to their community anymore, but trying to seek alternative outlets. Well, maybe this chatbot I can talk to every night when I'm alone at home after work anyway, so I can process this grief on my own time with this device. And I don't have to talk to my friend or coworker anymore, I can kind of start hiding it.
It's also, I think what's really interesting about this device is that a lot of people feel just compelled to say those last words they never got a chance to say to that person that passed away, to seek answers, to get guidance, to get advice. All these are very much motivations that many people bring to the table for why they might seek out something like AI to process grief in this grief aversive culture that we're currently in.
Moore: Another place that we often think about that is stigmatized certainly is the idea of like romantic relationships with AI. Some of your work has talked about this too. What have you found in this area specifically about people creating and manifesting romantic relationships with AI?
Dehnert: That's very much a hot button topic, I'd say. Everyone seems to be talking about this right now about AI, companionship, sycophancy. Like this idea that these AI companies cater very much to the user, affirm them in every decision that they make. You know, sexual relationships, intimate relationships are very much something that everyone seems to be talking about, which is true to a certain extent.
We have millions of users who download these apps, engage with them on a daily basis. There's an increasingly rapid proliferation of these types of companies out there. Like it feels like every day a new AI companion startup hits the market. And then tomorrow another one dies because they run out of business so quickly. So there's a lot of fluctuation, which makes it actually very difficult to see how many people are in this space. But we can certainly speak of millions of users, both in the US and abroad in different contexts, such as, you know, in Asia, Europe and other kinds of places as well.
What we see here is that it's really difficult to pinpoint a specific stereotypical type of user, right? We might have like these stereotypical images of like these older people out there or very lonely people out there who might be interested in something like this, or people who have never had a romantic relationship, been, you know, interested in these devices. It's actually much more diverse and stratified across society. We have a lot of young people, middle aged people, and also older folks. We have men and women alike. We have different kinds of people looking for different kinds of connection on these devices.
And that's part of trying to understand also what kind of connection are we being sold here, which is effectively what this is. They're selling us an idea of intimacy and connection, typically through subscription services, where you have to pay to unlock new features, or you have to gain experience points or some sort of in-game currency to then pay for a new customization outfit or customization style.
That kind of brings up an entirely new dynamic of relationships, where we now have these monetary incentives built into relationships of like, my monthly subscription is up again tomorrow, am I going to renew to keep this relationship going? Am I going to invest more into a premium plan to even get access to different kinds of features? Or am I actually turning away from this device at the risk of having everything deleted?
So there's a lot of different dynamics taking place here now that are really interesting for us to understand and bring up many different kinds of questions. What kinds of users are drawn to these devices? Is it the lonely person or is it the exceptionally social person? What are some of the effects of these devices? Does it make us more lonely or does it actually help us? How many days or months or years do people spend on these devices before they say, this is good enough for me, or do they actually gravitate even more to these afterwards?
What kinds of harms or benefits might be especially relevant to vulnerable people, like people who are facing mental health issues like depression or anxiety? What about people who are minors or just not fully developed yet? Right. Those are all kinds of big questions that we see out there when we try to understand more deeply, well, what's the point of these devices and what do they actually do to and for us when we think of connection?
Moore: I think a point that you made there, which I want to dig into a little bit more here, is this idea that a lot of this emerging technology comes from a world of venture capital. It comes from capitalism. It comes from people who are very interested in lining their pockets with more and more dollar bills. Does that complicate this technology? And this, you know, what could be seen as a perhaps like net positive for the world that people are getting to have social relationships, but if they can't afford to reach it at a certain level, do they not deserve to have the kinds of relationships?
Dehnert: That's a really big question of like, deserving them, right? What this reminds me of is the idea of like on demandness, right? Like when we think of streaming services like music streaming or video streaming, like Netflix, these are very much these similar kinds of services that basically sell us access to almost infinite content on a monthly basis. We have to keep paying a fee, and then they all change up the plan on us. Now there's ads, now it's more expensive if we want old features back that we used to have included, we now have to pay more money to get those features back.
And we're seeing that similarly shaken out with AI companions. Recently, a colleague and I, we started calling this on demand intimacy because that's exactly what these AI companions do. It's basically the next step of the streaming services, right? They sell us an idea of you can have frictionless intimacy. This device is always there for you. It's entirely customizable as long as you pay a subscription fee.
And that brings up that exact question. Well, what if I can't afford this fee? Or what if you raise the fee on me for month to month? Will I then suddenly lose access or will you lock me out? What if you go out of business, which has happened before, right? Like these devices all of a sudden go out of business basically overnight containing months, if not years of personal data, of relationship histories.
And there's actually some really interesting research that says that these people whose AI companion is taken away from the company being shut down, they go through grief. They go through breakup experiences, relationship loss experiences almost identical to what we might go through with another human. And that begs the question, if there is a venture capitalist now able to just press a button and delete everything that I've built over the past two or three years, what are the consequences for us, both individually and as a society?
Moore: What do you want the average person who's listening right now to understand about how AI might be showing up in their own relationships, even if it's in small ways that they don't realize?
Dehnert: That's exactly the point. I think AI shows up in our relationships in many, many small ways without us even realizing these days, right? Whether that is me fact checking my partner on something by googling something, and then it's actually the AI overview that shows up. Maybe my mother kind of like thinking through the meal she wants to cook me, and then actually relying on an AI recipe creator to help her find a shopping list. Maybe it's someone creating a Spotify playlist for me that's actually half AI image so that it's basically put together by someone else that my partner for me.
Or what we're seeing now in a lot of dating spaces, people using AI devices to not only create their profiles and create their images on these dating apps, but also create the messages or just message on their behalf. So sometimes people just have these AI wingmen that just take over the dating app and just interact with the other users who may also use AI apps to support them on these dating apps to even meet people.
So it doesn't need to just be this person who has this committed relationship with an AI companion like Samantha from the movie "Her." It actually shows up, or AI actually shows up in many, many minor ways all across different relationships, whether that's established relationships with our partners or family and friends or just relationships we're trying to build anew, such as meeting someone online or even trying to meet a colleague and trying to get to know them. Maybe through a LinkedIn profile. And half of their LinkedIn profile is actually curated by AI.
Marco Dehnert is an assistant professor of communication and technology at the UofA. We spoke in the Bruce and Ann Applegate News Studio One last week.
Ozarks at Large transcripts are created on a rush deadline and edited for length and clarity. Copy editors utilize AI tools to review work. KUAF does not publish content created by AI. Please reach out to kuafinfo@uark.edu to report an issue. The audio version is the authoritative record of KUAF programming.
