fbpx

My Baby Won’t Sleep. Help Me AI, You’re My Only Hope.

21st century parents hope data and tech interventions can help when they exhaust options — and themselves. Illustration by Matt Panuska

Sleepless babies can drive modern parents to desperation. Can bots, algorithms, and apps actually help?

On a recent Tuesday afternoon, my toddler lay facedown on the kitchen floor and wailed. When I left the room, she got up and followed me so she could cry closer to where I was. A quarter-hour later she was now flat on the dining-room floor, still screeching while her drool puddled. She had only napped 20 minutes that day, and it was driving us both crazy.

Lately, ever since my daughter turned two, most days have been like this: She naps for only 20 minutes, and then unravels before bedtime. My usual tricks to solve sleep trouble — basically, putting her to bed extra-early and waiting for her to grow out of it — haven’t worked. So I thought maybe technology could help. About two weeks into the napping crisis, I downloaded an app called Nod in the hopes its algorithms could puzzle out a solution for me.

Each day, I enter the start and end times of my daughter’s sleeps into the app, and log how difficult her bedtime was on a sliding scale of 11 gradations, from happy baby face to crying baby face. Nod analyzes when and how she sleeps, and shows me data about her patterns (long orange bar overnight, tiny orange blip midday). Most appealing of all, it promises daily AI-driven recommendations to improve her sleep.

I chose Nod because it was a finalist in the 2018 Best of Baby Tech Awards and offers help for toddlers up to 36 months old. Similar sleep-coaching apps include Huckleberry, Hatch Baby, and Cradle — all part of a wave of technologies that aim to help babies and toddlers sleep better. In addition to apps that comb through data to offer advice for better rest, there are smart monitors that track infants’ sleeping, waking, and wiggling, and robotic baby-jigglers that deliver carefully calibrated bouncing to help a baby pass out.

“I think a lot of the buzz around this space is capitalizing on tired parents who are thinking, ‘I’d give anything for a good night’s sleep,’” says Michael Pistone, who leads a technology accelerator at Cincinnati Children’s Hospital Medical Center. It’s difficult to prove that products like these really work, he says. And as I would learn, these technologies are not magic, just because they’re digital — they’re only as smart as the information and people behind them. Some of them have the potential to help babies sleep more safely, or to educate parents so they can figure things out for themselves. But if their experience is anything like mine, parents who hope that the arms of a robot or the brain of an app will soothe their child to sleep may eventually find themselves seeking human help.

Apps for naps

Rest Devices Inc., the company behind Nod, first created a gizmo called Mimo, a turtle-shaped monitor babies could wear that tracked sleep along with body temperature and breathing. The product, which launched in 2014 to mixed reviews, is no longer for sale. But it provided useful feedback for the founders: Information alone wasn’t enough. “Parents were saying, ‘Data’s great, but I actually want you to tell me specifically what to do for me to get my baby to sleep better,’” co-founder Dulcie Madden says.

So in 2017 the group launched Nod for people like my tired toddler and me. Users can choose from 10 goals such as “Teach her to sleep in her own room” and “Reduce her nighttime feedings.” Sadly, there is no goal for making a toddler’s naps longer, as I discovered. Still, I hoped the algorithms would notice that my daughter slept a paltry 10 or 20 minutes at a stretch during naps, and offer some advice.

Screenshots of the baby sleep app Nod.
Realizing data alone is not enough, the Nod app added 10 sleep goals to help parents. Courtesy of Nod

The app’s artificial smarts weren’t up to the challenge in my household. A glitch in the app initially made it think my daughter was only six months old, and it gave me congratulatory messages about how well she slept. I contacted the Nod customer support team, who fixed the age glitch but — despite pooling their actual human knowledge — had no suggestions about the naps.

Madden groans in sympathy when I describe our problem. She says the team is developing 30 to 40 more goals, including lengthening naps and other goals specific to toddlers. Eventually, they also hope to add real-time coaching driven by machine learning.

In the meantime, parents like me have many alternatives to choose from, especially if their kids are still infants. The baby sleep tech market is a busy place right now, providing many new options beyond the video monitors that first appeared in the 1990s and early 2000s.

Some products go a step beyond an app with physical devices. Nanit combines its app with a smart baby monitor; the camera tracks a baby’s movements to analyze its sleep.

There are smart monitors that track infants’ sleeping, waking, and wiggling, and robotic baby-jigglers that deliver carefully calibrated bouncing.

Ana Natera, a mom in Somerville, Massachusetts, started using Nanit when her son was about five months old and she and her husband were sleep training him. (For the uninitiated, sleep training includes any plan parents follow to try to teach babies to sleep on their own.) They thought the monitor was helpful — they appreciated the wide view from Nanit’s camera, as well as the data it gave them about how many times the baby had woken up each night and when they had gone into the room. But they didn’t always pay attention to the recommendations Nanit emailed once a week, like reminders not to put their son in his crib fully asleep. “It’s almost like when your boss scolds you,” Natera says.

Natera works in clinical research at a biotech company and is comfortable sifting through a lot of data. As sleep training progressed, though, the parents found they needed less information. Now their son is 15 months old. Although Natera still gets the weekly emails, she says, “I don’t think I’ve opened most of them.”

The $1,160 robotic jiggler

Parents may want data, but in that terrible moment when a baby is wailing instead of sleeping, a precisely plotted graph isn’t much use. Another set of arms to rock or bounce the baby just the way it likes would be ideal — and that’s now something you can buy.

Meet the robotic baby holder. The popular mamaRoo was launched in 2010 and has now sold more than a million units. For $220, the current version, which looks something like a throne, can sway a seated baby with the motion of a swing, a figure-eight, or the (allegedly soothing) hopping of a kangaroo. (Because it doesn’t let a baby lie flat, it’s intended for naps rather than overnight sleeping).

Product shot of the SNOO baby bassinet, made of wood, white mesh, and metal legs.
When the sleek (and pricey) SNOO detects a fussy baby, its rocking accelerates. Courtesy of Happiest Baby

Then in 2016, the SNOO made its debut by shimmying onto the scene. Its inventor, Harvey Karp, is best known for simple, accessible methods for calming babies. He described these strategies, including swaddling infants or letting them suck on something, in his book The Happiest Baby on the Block.

But the SNOO is anything but homespun. As soon as an infant is secured in a special swaddle that hooks into its frame, this sleek bassinet starts rocking and playing white noise. If the SNOO’s microphones detect the baby fussing, the bassinet cranks up its own volume and starts rocking more vigorously. (If this high-tech approach alone fails to calm the baby, suggests the SNOO website in FAQ tip, try putting a bag of rice on its belly.)

Natera bought a secondhand SNOO for her next baby. Her first was “a horrible sleeper,” she says. Since she’ll soon have two kids under age two, she says, “I just can’t afford to have a baby that doesn’t sleep at all.”

Natera researched the SNOO intensively beforehand to find out what people thought about it. Everyone can agree on one thing: it costs a lot. The most common word that people interviewed for this story used to describe the SNOO was “expensive.” At full price, it sells for $1,160. But since Natera found more positive feedback than negative — and was able to score a discount — she decided to take the plunge. When her second baby is born in January, she’ll find out whether it was worthwhile.

“I’m normally a very skeptical person,” says Megan Wilson, another Somerville mom. “I was probably the last person to get an iPod.” But she bought a SNOO (on sale) to use with her second child. Like Natera, she was afraid to relive the sleep troubles she had with her colicky firstborn. She felt “completely mystified” about sleep for the first year of her daughter’s life, she says. But five weeks in, her new son is sleeping well in the robotic bassinet. “I feel like this is a miracle worker,” she says, and then admits: “It could just be a different kid.”

Pediatric nurse practitioner and Boston Children’s Hospital sleep consultant Jennifer Gingrasfield doesn’t think it’s a miracle worker. Her job is to advise parents on how to help their babies sleep better, and she says she’s consulted with families who are using the SNOO and still need help getting their infants to sleep. She notes that the SNOO soothes babies by doing basically what a determined parent would — jiggling and shushing. But relying on all that intervention to fall asleep can become a bad habit that a baby needs to unlearn later, whether the help was robotic or human.

The device does have an upside, from her perspective. The SNOO has the potential to improve sleep safety, she says: A baby hooked into the bassinet will be flat on its back (as recommended) and not snuggled with a desperate parent on the couch.

Sleep apps could pose another problem, she suggests: The abundant data these apps provide could create a new source of stress for parents. Even with ordinary video monitors, she says, people get obsessed.

However, she thinks AI-driven baby sleep apps could, in principle, be helpful, depending on the quality of the information feeding the algorithms. She notes that much of the scientific research on how much sleep people need, for instance, is low-quality.

AI vs. human intelligence

Right now, new parents are on their own to seek out information about baby sleep in books, Google searches or apps. Ideally, Gingrasfield says, sleep education would be standard in prenatal care. All parents would learn some basics, such as that waking up during the night is normal for babies and adults.

Madden agrees that parents need more information. Parenting is changing, she says, and our own mothers and fathers might have had very different experiences than ours. Any 1980s snapshot of a baby sleeping on her stomach in a drop-side crib lined with soft bumpers illustrates a few ways things have changed.

Parents need more information. Parenting is changing, she says, and our own mothers and fathers might have had very different experiences.

Technologies like hers can help fill that gap, she suggests: “Any technology product that comes to the market should always be used to help a mom or a dad build their instincts,” she says. “The parent’s instinct is always going to be the number-one most important thing. I don’t think technology can replace that.” Even though she’s been using Nod with her own two children, she doesn’t expect her invention to be wiser than she is about sleep.

Since Nod couldn’t help me, I ask Gingrasfield if she (oh please) has any ideas about my daughter’s naps. I describe the pattern she’s fallen into. “Well, that’s actually the easiest problem,” Gingrasfield says. Startled, I press the phone to my ear and listen like she’s about to pass on a state secret.

She thinks my daughter needs less sleep now than she did earlier in life, and her lengthy nighttime slumbers are preventing her from napping. I should put my daughter to bed later and get her up promptly in the morning, she says, no matter how little she naps.

As of this writing, we’ve been trying the new schedule for almost a week. My daughter’s eyes are weary and pink-rimmed by evening. A few nights ago around 6:30 she asked, “Blue blankie yellow blankie?” in a hopeful tone, meaning the blankets in her crib. I told her she wasn’t allowed to go to sleep for another hour. But then one day she napped for almost 50 minutes, so maybe there’s hope. And if this doesn’t work, maybe another mind somewhere out there — made of synapses or software — has the answer.

Go Deeper