Designing Comfort and Trust in Caregiving Robots

The path to physical assistance runs through something unexpected: a reassuring face

(Source: Piya W./stock.adobe.com; generated with AI)

Based on an interview with Hod Lipson

Robotics is increasingly transforming healthcare, supporting everything from minimally invasive surgery and hospital logistics to diagnostics and rehabilitation as intelligent machines assist medical professionals across diverse clinical settings. As the design and functionality of robots continues to evolve, new robot models are serving as healthcare assistants, helping to augment the work of overburdened human caregivers.

To understand where caregiving robots are headed, we spoke with someone who’s spent his career exploring what makes machines feel human. Hod Lipson, a Columbia University robotics and artificial intelligence (AI) researcher known for his groundbreaking work in soft robotics and expressive machines, argues that the future of caregiving will not hinge on strength or dexterity alone. Instead, it will depend on whether robots can earn our trust through subtle, deeply human signals—like a smile delivered at the right moment. In this blog, we explore how Lipson’s perspective reframes the entire caregiving roadmap and reveals why emotional comfort may be the real frontier of robot design.

Robotics’ Moving Target

Lipson runs a robotics lab that has spent years developing compliant materials, adaptive actuators, and force-sensing systems that would let robots safely lift, steady, and assist humans in need. The most physically demanding caregiving tasks, such as lifting someone from a wheelchair, helping them bathe, or steadying them through transfers, are also among the most technically complex. They require exactly the kind of soft, compliant bodies roboticists have spent years trying to build.

But those tasks also require extraordinary trust. And trust, Lipson has concluded, doesn’t start with a gentle touch. It starts with whether a robot’s face makes you feel safe enough to let it stay in the room.

“For 10 years, we were thinking soft robots are going to be soft hands, arms like an elephant trunk,” Lipson says. “But if you look around, you don't see any of that. Faces are the soft robot that we did not see coming.” The potential reason for this comes from an observation that emerged from Lipson’s research: “When people talk, we measured about 50 percent of the time they are staring at the [speaker’s] lips.”

Hod Lipson is a Columbia University professor who works in the areas of robotics and AI. He is an award-winning researcher, teacher, and communicator. He has received many recognitions, including Esquire magazine’s “Best & Brightest,” Popular Science’s “25 Most Awesome Labs in the US,” and Forbes “Top 7 Data Scientists in the World,” to name a few. His TED talk is one of the most viewed on AI, and he was centrally featured by the New York Times in their 2023 piece, “What’s Ahead for A.I.”

Why Lifting Comes Last—Not First

It’s a common assumption that caregiver robots step in for the most physically demanding moments: lifting a person, steadying them down steps, helping with transfers, and assisting with bathing. Tasks that strain human caregivers and carry real risk when something goes wrong.

At first, Lipson also expected soft robots to tackle physical caregiving first. Now, he believes those tasks will come last.

"Originally, that's what I thought soft robots would do," he says. "But I think that's actually going to be the hardest thing."

This is not because the technology is impossible, but because the stakes are too high to rush. "It's dangerous because it has to lift the whole body," Lipson explains. "It requires a lot of force and a lot of trust."

One mistake during a transfer could lead to a serious fall. One misstep in shifting weight on a staircase could lead to an injury. Even if engineers perfect the mechanics, society will need time to accept robots performing personal, high-risk care. Lipson estimates that it could take a decade for this trust and comfort to develop.

The path to that future begins with something simpler than physical labor.

Getting to Know Caregiver Robots

What Lipson discovered in his robotics work is that the first applications will be situations where the caregiver robot isn't touching anyone. "Tidying things, bringing food, taking away things, clearing stuff," he says. These tasks may sound mundane, but they solve real problems and, more importantly, let people get used to a robot moving through a home without having to trust it at the level of physical contact.

Only then does Lipson expect caregiving robots to move to the second stage: companionship.

"You're already seeing robots that are like a pet," Lipson says. "It can just be there."

This might sound peripheral—until you consider what companion robots actually provide.

Lipson saw this firsthand with his own grandmother. At 100 years old, she loved talking but would cycle through the same stories, the same questions, the same observations. "It was tough for the caregivers to have conversations—over and over the same conversation," Lipson says. "But she would enjoy that very much."

Human caregivers can struggle with this kind of repetition. Robots don't. "AI has infinite patience," Lipson says. "I think that's a compelling thing."

He puts it another way: "The part where the robot sits down and plays Scrabble, okay, and listens to the same joke 50 times—that's the part that AI can do very well."

Companionship matters beyond loneliness because it's where robots earn permission to be present. If people can't tolerate a robot during a simple conversation, they won't accept it helping them get dressed, guiding them down stairs, or standing beside them in a bathroom.

And that tolerance depends heavily on whether the robot's face and behavior feel right—whether it looks attentive without being intense, whether its timing matches the moment, whether its expression signals safety rather than uncertainty.

The 50-Degree-of-Freedom Problem

Walk through any robotics lab, and you'll see humanoid robots doing impressive physical feats. What you won't see are convincing faces.

"A lot of people are building humanoids—they do backflips, they're running, they're walking," Lipson says. "None of them has a face. They're all missing a face because faces are really, really hard."

The reason comes down to a measurement roboticists use called degrees of freedom, which is the measure of how many independent movements a system must coordinate. A leg has about ten. A hand has about twenty.

"Human faces have about fifty degrees of freedom," Lipson says. "Everything we do in terms of human communication has to do with moving those 50 muscles in a particular way: smiling, nodding, looking, talking, moving your lips, moving your eyes, frowning, everything."

Coordinating all those parts is extraordinarily difficult. But the real challenge isn't only mechanical—it's psychological.

"If you move your hands incorrectly, it's okay," Lipson says. "But if it smiles incorrectly, it's really, really creepy."

Humans read faces automatically and unforgivingly. A smile from a humanoid robot that's slightly mistimed or asymmetrical doesn't register as "almost right." It triggers the uncanny valley—that visceral sense that something is amiss. And in caregiving, where trust is the whole game, a single wrong expression can undo weeks of confidence.

That's why getting the face right isn't just a nice-to-have feature. It becomes a prerequisite for everything else.

Learning How to Look Normal

Lipson's lab has built a robotic face with twenty-six degrees of freedom, which is fewer than a human face, but still provides the ability to offer some facial expression. The robot studies authentic human expression by training on large collections of human video data, including online video sources, to learn statistical patterns of facial motion.

Still, the challenge isn't just making the mechanics work. It's about timing, context, and the thousands of micro-adjustments humans make without thinking. A smile has to arrive at the right moment, match the emotional tone of the conversation, and fade naturally.

Get it right, though, and something remarkable happens.

"Once robots move their lips and use them in the correct way," Lipson says, "they almost become alive. There's something almost magical about that moment."

That "magic" is what unlocks the door to physical caregiving—not because a face makes a robot stronger, but because it makes people willing to trust what comes next.

Compliance Without Softness

Although Lipson’s team has yet to achieve an authentic robotic smile in the way they set out to, they remain confident that compliant materials and adaptive control will become essential, just later in the timeline.

When robots do start performing physical tasks, they'll need fundamentally different behaviors than factory robots. "You want your tractor to plow through, even if it faces an obstacle," Lipson says. "But you want your caregiving robot to back off if you push back."

That principle—yielding instead of insisting—is what roboticists call compliance. And here's Lipson's twist: compliance doesn't necessarily require soft materials.

"You can have a rigid robot that almost feels soft," he explains, "just because the software makes it sort of ... timid." With enough sensors feeding back resistance data, even a metal-framed robot can back off the moment it detects pushback, behaving less like a machine forcing its way through a plan and more like a cautious helper waiting for permission.

Whether achieved through physical materials or sophisticated software, compliance will matter enormously once physical assistance begins. But physical contact comes after trust, and trust is earned through repeated, uneventful success—day after day, interaction after interaction—until the robot feels normal enough to keep around.

Building Trust Through Behavior, Not Just Materials

Ask engineers about barriers to caregiving robots, and they'll list complex technical challenges: actuators, sensors, power systems, and cost. Lipson returns to something else.

"The psychology of soft robots is going to be the biggest obstacle."

And the stakes only climb as the setting gets more sensitive. Elder care is often treated as the main proving ground; however, Lipson sees it as the first step toward even harder domains, like childcare and education, where acceptance and safety standards are even less forgiving.

Society doesn't grant trust in new technologies quickly, especially when vulnerable populations are involved. Driverless cars are going through this now. "They have to prove themselves, and people are still uncomfortable," Lipson says. "But once you sit in a driverless car a couple of times, you become comfortable with it."

Caregiving robots will follow a similar path, but with higher stakes and slower acceptance. The most capable robots won't make a difference if people refuse to be near them. That's why the deployment roadmap doesn't match what engineers might assume. The most technically demanding tasks—lifting, bathing, transfers—are also the ones that require the most trust. And trust has to be built from simpler interactions first: a robot that moves around the home without incident, a robot that doesn't crowd you, a robot that can hold a conversation without unsettling you.

That's why Lipson keeps coming back to the same deceptively simple standard.

"It's going to move in a normal way," he says. "Getting it to be normal is the challenge."

Normal for humanoid robots means a smile that doesn't make you uncomfortable. A gaze that feels attentive without being intense. Lips that move naturally with speech. The tiny signals humans use to decide, in a split second, whether someone—or something—is safe to be near.

Robots that can do backflips still can't smile naturally. And until they can, the path to physical caregiving—with all its sophisticated soft robotics technology—remains blocked by something as small as an expression.

The most complex caregiving tasks will eventually happen. But they'll happen only after robots master something simpler: making a person feel comfortable enough to stay in the room.

That starts with a smile.