THE ROBOT’S TORSO and limbs are chubby and white. It seems to be naked except for blue briefs below its pot belly, although it does not have nipples. It is only 2 feet tall. Its face, a rectangular screen, blinks on. Two black ovals and a manga smile appear.

“Hello! I am QT, your robot friend,” it says. It says this to everyone, because that’s its job. QT raises both arms in a touchdown gesture. The motors whir. They sound expensive.

It might look and sound sort of familiar if you know anything about humanoid social robots—contraptions built to respond to us in ways we recognize. You may also remember their long history of market failures. RIP Kuri, Cozmo, Asimo, Jibo, Pepper, and the rest of their expensive, overpromising metal kin. QT is not like them. It is not a consumer product; it’s a research device equipped with microphones, a 3D camera, face recognition, and data recording capabilities, built by a Luxembourgian company for scientists like Šabanović to deploy in studies. She’s using QT to explore ikigai, a Japanese word that roughly translates to a reason for living or sense of meaning in life, but also includes a feeling of social purpose and everyday joy. Doing a favor for a neighbor can create ikigai, as can a hard week’s work. Even reflecting on life achievements can bring it on. Her team, funded by Toyota Research Institute, is tinkering with QT to see what kind of robot socializing—reminiscing, maybe, or planning activities, or perhaps just a certain line of conversation—might give someone a burst of that good feeling.

To see QT in action, I drive Šabanović and graduate student Long-Jing Hsu over to Jill’s House, a small memory care home in Bloomington less than 2 miles from the university. Šabanović and her teams of students and collaborators have worked alongside residents here for years. It’s now September, and all summer Hsu has visited every week, leading workshops, testing small adjustments to QT’s demeanor and functionality, collecting data on how people react to the robot—whether they smile, mirror its gestures, volunteer parts of their life story, or get bored and annoyed. Making decisions about how and what the robot should do is not up to the researchers, says Šabanović. “It’s up to this deliberative, participatory process, and engaging more people in the conversation.”

A living area at Jill’s House, a small memory care home in Bloomington, Indiana. Photograph: Kayla Reefer

One challenge is that dementia is never the same for any two people. There are different varieties, such as Alzheimer’s, frontotemporal dementia, and Lewy body disease, and they are dynamic, changing with time. Some people have no problem with memory but struggle with words; others make strange decisions. Many say their perception of time changes, or their senses become more acute. Some people are angrier, some calmer, and others lose all filters and say whatever they think. (I’m looking at you, Mom.) In my experience, dementia doesn’t erode individuality. It sharpens and twists it. As the disease advances, it can steal the ability to drive a car, operate a microwave, change clothes. It also layers on new habits and quirks. My mother, for instance, has always loved natural history. With dementia she now can become completely absorbed in a leaf, a flower, or a pattern of light and shadow in the trees. A good, useful, helpful robot must adapt to each person at each phase.

Today, Hsu will demo a storytelling game between person and machine. Eventually QT will retain enough information to make the game personalized for each participant. For now, the point is to test QT’s evolving conversational skills to see what behaviors and responses people will accept from a robot and which come across as confusing or rude. I’m excited to see how this plays out. I’m expecting spicy reactions. People with dementia can be a tough audience, with little tolerance for encounters that are annoying or hard to understand.

After I park the car, we enter a large common room with cathedral ceilings and high windows, bright with sunlight. As we chat with the director, Hsu crosses at the back, a small figure pushing a white robot on a cart past the wingback chairs and overstuffed love seats to a small side conference room. It’s a spectacle, but the residents who are hanging out this afternoon take little note. To them, robots are old news.

We follow her into the side room and pull chairs into a semicircle around the robot on his table. Šabanović will mostly just watch today as Hsu troubleshoots the tech and leads the session. Soon, Maryellen, an energetic woman in a red IU ball cap, walks in and takes a seat across from the robot. Maryellen has enjoyed talking to QT in the past, but she’s having an off day. She’s nervous. “I’m in early Alzheimer’s, so sometimes I get things wrong,” she apologizes.

The robot asks her to select an image from a tablet and make up a story. Maryellen gamely plays along, spinning a tale: A woman, maybe a student, walks alone in the autumn woods.

“Interesting,” says QT. “Have you experienced something like this before?”

“I have,” Maryellen says. “We have beautiful trees around Bloomington.” The robot stays silent, a smile plastered across its screen. QT has terrible timing, pausing too long when it should speak, interrupting when it should listen. We all share an apologetic laugh over the machine’s bad manners. Maryellen is patient, speaking to QT as if it were a dim-witted child. She understands that the robot is not trying to be a jerk.

Today’s robot-human chat is objectively dull, but it also feels like a breath of fresh air. Everyone in this room takes Maryellen seriously. Instead of dismissing her pauses and uncertainty as symptoms, the scientists pay careful attention to what she says and does.

Next enters Phil, a man with a tidy brush mustache, neatly dressed in chinos and a short-sleeve button-down printed with vintage cars. After taking a seat across from the robot, he chimes in with QT to sing “Take Me Out to the Ball Game.” He faces the machine, but he’s playing to us, mugging and rolling his eyes. Song over, he first teases Hsu, then another resident, then pretty much every woman in the room. In other circumstances he’d be patronized or “diverted”—someone would attempt to distract him. Instead, we join him in being silly, joking about the situation and the robot.

QT pipes up with another round of awkward conversation (“I love the song. Do you?”), and Phil replies with a combination of graciousness and sass (“You sing very well. Did you have that recorded, maaaybe?”). Hsu asks Phil how he felt talking to the machine. “Like I’m a fool talking to nothing,” he says sharply. “I know it’s not a real person.” Theatrically, he turns to the robot. “You’re not real … are you?” He winks, and laughs uproariously.

Long-Jing Hsu, a graduate student who works with Sabanovic. Photograph: Kayla Reefer

He likes the robot? He doesn’t? It’ll be the team’s job to figure out these enigmatic yet relatable reactions. The three of us plus robot pack up and head back to Šabanović’s R-House Lab at the university. In the big conference room there, her team will converge, students of informatics, data science, computer vision, and psychology. They’ll pick apart Maryellen’s kindness and hesitation and Phil’s glee and annoyance, looking for their next task, the next skill QT needs to learn.

I drive around for a while, thinking. The demonstration session was playful and lighthearted, but the room also had an unfamiliar energy that I hadn’t felt before in the context of dementia. Activities for people with dementia are usually perfused with nostalgia, turning away from the difficult present to relive what remains of the past. But today was all about the future. We were imagining, playing at new kinds of relationships, trying to brainstorm applications and uses that are still hard to picture. It felt creative. More precisely, it felt alive.

THE R-HOUSE LAB at Bloomington seems to be an ordinary conference room: white walls, large wood conference table, and desks, chairs, and monitors ringing the walls. But it is littered with the artifacts of Selma Šabanović’s research career, a menagerie of robots and robot parts. Perched on a windowsill are two butter-yellow Keepon units, an early invention that’s little more than two spheres with eyes and a button nose. A fluffy white seal named Paro recharges on a filing cabinet, its power supply a pink pacifier plugged in where its mouth should be. In the back of the room lurk two of Honda’s 2018 Haru, something like a desk lamp crossed with a crab. Three more QTs (2017) doze on tables. It’s like an unnatural history museum, and Šabanović is the resident David Attenborough.

Šabanović, 46, is tall and slender with a cloud of dark hair and an aura of sly humor that suits this strange place. Now she sits at the back of the room as her team hashes out ways to solve QT’s social problems. At one point, the students get stuck on QT’s habit of interrupting. Maybe that’s OK, Šabanović proposes. “We can play off the fact that it will be, inevitably, to some degree, stupid,” she suggests. What the researchers need to figure out is “where stupidity is harmful.”

She may be in her element, but the defunct and sleeping robots give the room a weird energy, as if it were full of ghosts. This, I learn, is the superpower of social robots: not strength, not speed or precision, but vibes. They grab hold of our psyche. They get under our skin. Even though we know better, we respond to them as if they were alive. The tech critic and author Sherry Turkle calls it pushing people’s “Darwinian buttons.” Unlike most other gadgets, robots get our social instincts tingling. Of course, explains Šabanović, “what distinguishes robots is that they have a body.” She adds, “They can move, show they’re paying attention, trigger us.” Children learn more from a robot than a screen. Adults trust robots more readily than computers. Dogs obey their commands.

Šabanović is fascinated by these reactions, which makes sense, because she has lived her life in the company of robots. Her parents were both engineers, and her father worked on industrial robots. Back in those years, the only social robots were in fiction. Her father’s contraptions were serious machines for heavy industry. When she was 9, in 1987, the family spent a summer in Yokohama. An only child, she would often tag along to work and curl up in the lab with a book. In Japan, she noticed, fictional robots were friendly and helpful: more Astro Boy than Terminator. In college in the mid to late ’90s, she puzzled over those differences and wondered why some cultures assumed future robots would be sweet and cute, while others imagined them as villains or thugs.

A resident at Jill’s House, a memory care facility, tests QT’s conversational prowess

By then she was also dropping in at conferences with her parents, where she heard about real social robots—machines designed to interact with humans on our terms. This seemed like a strange idea. The industrial robots she knew so well were unattractive and untouchable, doing dangerous tasks on assembly lines. “It was super intriguing to me—how do they think this is going to work?” she says.

She wanted to understand these new relationships forming between people and machines. In graduate school, she visited and observed the people who were developing the field of social robotics. In 2005 she spent time with the pioneering roboticist Takanori Shibata at Japan’s National Institute of Advanced Industrial Science and Technology and his robot seal pup Paro. Handcrafted, the little critter responded to speech and touch by bleating—it was programmed with actual seal pup cries—closing its eyes, and flipping its tail and flippers. It was one of very few robots at the time that could be used outside the lab without expert assistance.

Even at this early stage, elderly people were the target audience. The researchers took the machine to care homes, and Šabanović was startled to see the effect. “People would suddenly light up, start talking to it, tell you stories about their life,” she says. Shibata’s studies, then and later, showed that the cuddly seal improved quality of life; it got people to interact more, reduced stress, and eased depression.

So Šabanović joined the emerging field of human-robot interaction. Her experiments since have explored how we project our “techno-scientific imaginaries”—our cultural baggage, fears, and fantasies—onto these hunks of metal and plastic. Sort of like if Isaac Asimov became an experimental psychologist.

In one early study, she brought Paro into a nursing home to study how the device turned wallflowers into butterflies. Most residents would ignore the seal pup until other people showed up—then it would become an icebreaker or a social lure. They’d gather to touch it. They’d comment on its sounds and movements, laughing. The robot, she saw, seemed to open a door to other people.