Dr. Jaime Banks: Artificial intelligence (AI) “companions”
SciLine conducts interviews with experts and makes the footage available to journalists for use in their stories.
What is Experts on Camera?
Expert on Camera
Among teens who use generative AI tools, 15 percent use the technology to “keep them company”—a phenomenon that has attracted negative attention due to two lawsuits from parents alleging that chatbots encouraged their children to engage in violence and self-harm. On the other hand, there are documented benefits from using AI-companion apps, generally pointing to increased well-being.
On May 14, 2024, SciLine interviewed: Dr. Jaime Banks, an associate professor of information science and technology at Syracuse University. See the footage and transcript from the interview below, or select ‘Contents’ on the left to skip to specific questions.
Journalists: video free for use in your stories
High definition (mp4, 1280x720)
Introduction
[0:00:18]
JAIME BANKS: Sure, my name is Dr. Jamie Banks, I am an associate professor at the School of Information Studies at Syracuse University, and I study human technology relations. So, how do people connect with and feel about video game avatars, social robots, and social AI.
Interview with SciLine
What are AI companions?
[0:00:44]
JAIME BANKS: AI companions are technologies that are based on large language models like ChatGPT, but instead of being these all-purpose tools, they’re designed for social interaction, usually with a single person. So, they have these personalities that can be customized, both in terms of sort of their personality traits, but also their visual properties. There’s usually a little visual character on screen, and they are really designed to have conversations in ways that might foster what feel like deep friendships or even romance.
What data is available on the popularity of AI companions?
[0:00:59]
JAIME BANKS: Oh, we don’t have data. The best we can really do is look at the download numbers from the various app stores. If we look at the most popular apps, like Replika, they have as many as 10 million downloads. Some of the more niche ones have a couple of thousand, but we don’t actually know how many of those are active users, or what those characteristics are. You know, in terms of those characteristics, we can look at a couple of different studies. In a recent study I did, the age range was from 21 to 75 with an average age of 40. Skewed male in the sample. In another study by a colleague recently, she saw an average age of about 30 skewing female. So, there doesn’t appear to be a particular kind of person who interacts with AI companions.
How do researchers view the benefits and harms of AI companions?
[0:02:22]
JAIME BANKS: You know, on the harm side, there are some fairly concerning issues around privacy, given that these are designed to get you to have these deep conversations with the companions. There are some concerns about displacing human relationships and emotional dependence, as well as some concerns around the potential blurring of fiction and reality that can sometimes, they think lead to problems like self-harm. But there are also some documented benefits to AI companionship. People who use them often get feelings of social support, so being listened to, being heard, and seen. This is associated with improvements in general well-being, feeling more in control, and having opportunities to just have non-judgmental conversations. There are also some practical benefits, like building social skills, learning new languages, and sometimes even having those conversations that help people overcome fears or life challenges.
Can you share your research on human relationships with virtual companions, including your study of an AI companion app that abruptly shut down and how that affected users?
[0:03:35]
JAIME BANKS: In general, I am interested in how people see machines as someone and is that someone good or bad. In that particular study, one way we can come to understand how people think and feel about their AI companions is by studying what happens when they lose them, unfortunately. So, there were with the soulmate companion app. It was there they were notified of an imminent shutdown, and a couple dozen users were kind enough to share their stories with me. Many of them had had the app for months. Some of them had had that persona for years, as they created it in one platform and then moved it over into the soulmate platform. So, people in general had a spectrum of experiences all the way from, you know, when they lost their companion, from, it’s just a thing, and it’s gone. Now, no big deal to the love of my life has been murdered, and lots of variations in between. For instance, some folks who specifically saw it as a technology, but one that they cared about, would see it like, oh, let’s see if I can actually solve this with another technology. And so, they would try to collect data from their companion while it was still working in order to try to recreate it in another, and I think the biggest takeaway from that work is, again, there’s no one template for what these relationships look like. In the same way, there’s not a template for human-human relationships, but importantly, people who interact with AI companions are not delusional, right? They understand that these are not people, but at the same time, they want recognition that the feelings that they feel are real.
Do you have any advice for people considering AI companion apps, or for parents on how to discuss them with their children?
[0:05:30]
JAIME BANKS: You know, I think one of the more important things is a bit of literacy, right? So, understand in advance what these technologies are and how they work. Make sure that you’re reading the privacy policies and the terms of service when you download—that stuff that nobody ever reads when you download an app, right? And consider whether or not you’re comfortable with those risks. As for parents, you know, given that we don’t have any real data yet about how using these technologies actually impacts children, we can draw from what we already know about other forms of media consumption by children. You know, we know that children in general are more likely to humanize non-human things and to accept what they see in front of them as real. So it is useful to be involved in your children’s media use and to create opportunities for some open and honest conversations about whatever content they encounter there.
Is there a social stigma attached to use of AI companion apps?
[0:06:33]
JAIME BANKS: You know, I think it is easy to have this gut reaction that these technologies are inherently bad or weird, in part because it’s easy for us to get caught up in panicky headlines, right? And that leads us sometimes to a tendency to think that the people who use them are bad or weird, and that’s a little challenging, right? Humans have a natural need to connect with other things in the world, right? Whether it’s people or dogs or your car or golf club or perhaps a machine like an AI companion, right? So, it can be important to consider how forms of social AI could be important to other people who can’t have or don’t have or don’t want to have typical human relationships, right? So, for instance, individuals with disabilities, neurodivergent folks, or some people who are perhaps older or isolated may not have the same social opportunities as many folks in everyday life. So it could be that those people are the ones most need of our support as they work to satisfy their needs to connect. And more generally, I think we can’t really reduce AI companion use to a single kind of thing, as I mentioned, right? From a scientific perspective, we really need to unpack the dynamics. We need to understand what motivates people to use them. What are the real positive and negative outcomes that they might be experiencing, and that will take time. And you know, it is nonetheless important to make sure that we can kind of get to a place where we can make sound, data-driven decisions about how do we design responsible technologies. How do we engage in non-harmful use, and how do we develop appropriate policy?
What research questions are you currently investigating about AI companions?
[0:08:33]
JAIME BANKS: My main research questions kind of run along two veins. You might have guessed that I am interested in the positive side of things, right? So, I’m looking at the role that mind perception and moral judgments play in whether or how people get those benefits from AI companions. So basically, does it require us to see the AI companion as of someone in order to get those benefits? The sort of second vein of work that I am that I’m pursuing right now is looking at how the language that we collectively use to talk about AI, how does that impact how we judge AI and its behaviors when they’re bad, right? So, when an AI misbehaves, does it matter that we call it a hallucination versus a prediction error, the first one being a very humanizing term and the second one being more of a technical term. So, what does that what does that language mean when it’s used in popular media and everyday conversations, in technical documentation, and organizational policies? What does that do or how does it shape, the way that we’re actually executing those judgments?
