Transcript

Rachid 00:00:00 I think we have a bear, a dolphin, a hedgehog, a giraffe, and the parrot, which I am apparently is one better than the other.

Voiceover 00:00:08 Welcome to the Made by Google podcast, where we meet the people who work on the Google products you love. Here's your host, Rashid Finge.

Rachid 00:00:15 I'm wide awake for this one. Today we're talking to Google Sleep Scientist Logan Schneider and senior staff research scientist Conor Heneghan about sleep.

Voiceover 00:00:25 This is the Made by Google Podcast.

Rachid 00:00:27 Logan, let me start with you. Why does a world-class sleep scientist work at Google?

Logan 00:00:33 Well, I think that's a, a great question. I have my own personal reasons why I work at Google. One is I appreciate all of the capabilities that Google brings to bear on being able to understand people's sleep and then help them sleep better. But I think the question really is why does Google invest in having a sleep clinician involved in its product development? And I think the answer is because they truly wanna make people live healthier lives at the scale of billions. And so the way to do that is obviously bring experts in such as myself or other clinicians who work on other Google products to ensure that people are getting the best value out of their investment in Google.

Rachid 00:01:10 And Conor, you're not a sleep scientist but an electrical engineer by trade. So I sort of think of Logan and you like Batman and Robin, maybe James Bond and Q working together uh, to bring those great sleep tracking products out. Is that fair to say?

Conor 00:01:24 I actually think of us as a bit more like the characters from the Incredibles movie where you have lots of different heroes with superpowers and that's a lot like how our team works. We have people who are super deep in things like large language models, psychology, sleep science, neurology, like Logan himself, lots of modern machine learning technology. How do you make devices? We're a team of Incredibles.

Rachid 00:01:47 That's an amazing way to put it. So it is almost World Sleep Day as we record this world. Sleep Day is on March 15th. I think I found the perfect way to prepare for this episode because I had a terrible night. Maybe a little bit too much wine. And then we have a four month old and a 3-year-old. So that combination really can hurt your sleep. I think most people are listening have some basic understanding of what good sleep looks like. So still curious though, Logan, in your experience, what is the biggest misconception that people seem to have about what great sleep looks like?

Logan 00:02:19 I think the misconception comes out of our deliberate efforts to help raise awareness in the sleep medicine community and highlight that many people are not getting adequate sleep. The emphasis often is on getting more mm-Hmm or getting enough. And I think that's where a lot of people have been striving for increasing the amount of sleep that they get. And I think that's admirable. Our goal obviously is really to allow for enough time to get the sleep you need.

Rachid 00:02:45 Those people who tell me like I only need five hours of sleep where I was like, you know, you're crazy. But that could actually be the case then

Logan 00:02:52 It could be. But we also have to be cautious, right? Because our brains are very good at deceiving us as to our sleep needs. And so from that standpoint, there's a very famous study from 2003 from the UPenn group that highlighted that the awareness of your sleep related impairments from insufficient sleep very quickly plateau even though you continue to accrue deficits, right? Meaning that you become unaware of how insufficient your sleep has been either from a quantity or quality standpoint. And you just highlighted in your prior experience why that likely is evolutionarily if kids continue to disrupt our sleep and we were aware of how sleep deprived we were through that first period of time, I don't think many of us would have a second kid. Yes. So in that sense, our brain has to kind of habituate to this insufficient sleep over time. This is my theory on this and just like all stimuli we habituate to it, a cold temperature slowly becomes normal for us and we don't feel as cold, right?

Logan 00:03:45 So same thing with sleep deprivation and sleep insufficiency. So we have to be aware, we can be tricked into how well we feel relative to how impaired we might actually be. And there is no sleep elzer like a breathalyzer to tell us how impaired we are. But you can look to aspects of your day. Are you drowsy dozing off at inappropriate times? Are you needing to consume caffeine in order to stay awake? Are you taking naps? Like all of those signs can be indications or sometimes if your mood is just off can be indications that you're not getting enough sleep. And so it might be helpful to help you calibrate what that amount is, but it's hard. There's no perfect answer.

Rachid 00:04:17 So I guess Conor, this could also be potentially where sleep tracking comes in because rather than faring on your perceptions of sleep, it gives you some sort of more objective measure of how you slept, right?

Conor 00:04:28 Yeah. And that's kind of been sort of the most fascinating aspect of the last few years as we begin to think more about the subjective aspect of sleep as well as objective. So as you, as you well know, the trackers and smart watches are doing a good job of tracking objective sleep metrics like how long you slept or how often you woke up during the night. And what we've been doing the last few years is then begin to see how that relates to subjective perceptions of sleep. And in fact, we're doing a really nice study right now with the University of Oregon called the Digital Wellbeing Study. And as part of that we ask people about their perceptions of sleep. So there are two NIH approved metrics called the sleep disturbance index and the sleep related impairments, which are kind of your subjective sense of, you know, is my lack of sleep affecting my day-to-day activity?

Conor 00:05:14 Am I waking up a lot? And what we found is the objective metrics that we measure, they're relatively poor predictors of those perceived uh, metrics. So I think that kind of opens up the next avenue where this technology goes to really bring in those more subjective and qualitative experiences. So you, you brought up the issue being woken up by a child at nighttime and that's something that a wearable device isn't gonna capture necessarily. Uh, we, we can see you woke up, we don't know if it was because your dog barked. Is it because you were worried about work? Is it because the child was crying? What we're really excited about internally in the technology side is you've probably heard of large language models which can really begin to interact with humans in a more naturalistic way. So potentially in the future, products like ours will include the ability to to talk, to journal, to give some feedback on what's actually going on in a person's life. And we'd use that to do a much better job of not only capturing objective sleep metrics but also potentially giving a little bit more insight into the subjective side of your sleep and you know, hopefully guiding people towards a better understanding in that way.

Rachid 00:06:14 So people who sort of make a joke out of a sleep tracker saying, you know, I can feel how I slept, right? I don't only need to look at my Pixel watch telling me how many hours or what my sleep score is, but there's actual valuable information in there because it's not perception based. Yeah,

Conor 00:06:28 I, I mean I think the future giving people both the objective information of how they slept is very important, but without taking away from their subjective experience. So I think it would be crazy for a person to say, Hey, our watch said you slept eight hours, don't worry about you, you should be fine. But yet they feel terrible, they fall asleep all the time. That subjective experience is very important and should not be undervalue. And I think maybe turning over the Logan, I mean when you, you deal with patients who have exactly those types of experiences where objectively things look okay, but subjectively they're feel like they're not sleeping as well. I assume from a clinical point view, the objective information can help guide a person towards what the underlying issues might be.

Logan 00:07:06 Yeah, you're right on point. Uh, Conor is the sense that in the clinical domain, obviously I'm looking for objective evidence of diseases, but also there is that potential for understanding somebody's actual experience. And sometimes I can't just rely upon my metrics, I'll check the boxes, you're okay. Typically in medicine if I can't treat something, obviously I have to suggest that they explore other avenues. But certainly in this space it's an opportunity to say you're experiencing something is sleep at the core of that. And we have an unprecedented ability to look at that in clinical medicine. I have that one night in my arbitrary sleep lab experience and hopefully there's a disease if I, if there is one that I can catch it and treat it. If there isn't a disease then I, I have to say, well you're okay, but in this opportunity it's unique. 'cause every day the individual can reflect on how they're feeling and then we can provide them a unique insight with objectively track data that they get to look at on a regular basis to say is the sleep that we're tracking associated with how you're feeling? And if it is, then we can help them identify what that is and then if it's something that they can improve, that's great.

Rachid 00:08:10 In your experience, what is the most surprising thing that people get wrong when they try to improve their sleep?

Logan 00:08:17 I think the most surprising thing is that people focus too much on the sleep itself. Uh, and maybe are not as focused on the behaviors and habits that affect that sleep, particularly when they're trying to strive for more. There's one thing that one of my teachers taught me in sleep medicine is that you can't try to get more sleep, right? You have to allow your body the opportunity it needs to sleep. Hmm. Um, but that doesn't just mean giving yourself enough time in bed. It actually means setting yourself up from the minute you wake up until that next sleep period. Establishing habits that promote healthy sleep and sometimes those habits are things that are bad habits that have accrued over your entire lifetime. Essentially you've built things up when you were more resilient. You know, as a teenager you can sleep anywhere, anytime and so caffeine won't affect your night, but now that caffeine is suddenly something you're more sensitive to. So it's really reflecting on your whole day looking at the healthy behaviors that you're acting on during the day in order to ensure that you get adequate sleep when you need it and when you want it.

Rachid 00:09:15 So Conor, I guess tracking sleep is just one part of the puzzle because you know, once you have that sleep data of someone, you need to actually present it in a way to that person that they can make sense of it, maybe can learn something from it, can compare it easily to nights before, maybe get some sort of recommendation on what to do next. How do we tackle that? How do we give those insights to our users?

Conor 00:09:36 Some of the ways we present insights is firstly by benchmarking relative to the population. Um, we have to be fair to people. So because sleep changes you age, we compare against your age group and also men and women have a slightly different sleep phenomenon. So in general, women sleep a little bit longer. So again, we wanna compare men with men, women with women, to give you a sense of where you lie. We don't want people to over index on that. But obviously if you are uh, at the first percentile or the 99 percentile, that's interesting to you to, to know that you're one of the extremes. The other ways we've taught about presenting information is the concept of what we call a sleep profile. What we've come up with is six different profiles and in medical terms it's called a phenotype, which is like how does your sleep get expressed?

Conor 00:10:18 You know, some people naturally kind of get very consolidated sleep with very little wake time in the middle. Some people have a longer sleep period but wake up a lot. So what we are trying to capture with those six different types is characteristic patterns of sleep. And again, what we're trying to give assurance to people is that they might think that their sleep is like out of whack or something really strange, but it turns out that 15% of the population sleep the same way they do. And we think that can be very beneficial to someone to understand their overall sleep pattern across more than just a single night, but across a whole month of data. And then in the future where we'd like to get to is to start to tie it to daytime performance. So we did an experiment recently where we were collecting people's vigilance levels, so using basically a response task on the phone to see if we could figure out the impact of a person's sleep on their daily vigilance. That ties back to the question you had earlier about how much sleep do I need? So some people might need, you know, nine hours to kind of be at their optimal vigilance during the day. Some people the five and a half hours may be sufficient. So I think in the future we wanna also use our sleep trackers to tie better to what the daytime outcome might be, whether it's cognitive, physical and so on.

Rachid 00:11:30 Just wanted to quickly circle back to the sleep animals. I think we have a bear, a dolphin, a hedgehog, a giraffe, and the parrot, which I am apparently is one better than the other. Is that a fair way of looking at it?

Logan 00:11:43 The question that you have to ask is are you the right animal for you? We don't have any evidence to suggest that there is one specific animal that is preferred. What we're trying to help people do as Conor highlighted, is really just give you context. Where do you fall in relation to multiple other individuals? And does that sleep pattern that we just described based on how your data assigned an animal to you, does that seem to fit with what your experience is? And if it does, that's hopefully helpful insight and then you can focus on the aspects that helped identify the animal. Was it the length of your sleep? Was it how long it took you to fall into a sound sleep? Those things can give you insight into, is that ideal for me or is that something where maybe I should move to a different spot and see if I'm that still that animal?

Logan 00:12:31 Right? If you become healthier in any of those dimensions, if you achieve your goals in any of those dimensions and you remain that animal, it might suggest that yeah, okay, you're fitting more into that mold if not, and you actually end up shifting to a different animal. Maybe that's the suggestion that by getting myself to where I need to be now I've identified where I ought to be. So it's really just an opportunity for like a guide on the journey. Where do I fall relative to others? And based on the scientific evidence of what patterns of sleep might help you identify what style of sleeper you are really.

Rachid 00:13:01 Logan, you mentioned your sleep lab earlier and I'm wondering how sort of different it is to tracking sleep on a wearable, like a Pixel watch. You know, when when you get to the end of someone's sleep, how different is the data you get from your sleep lab versus what you see in the Fitbit app?

Logan 00:13:15 So from a clinical standpoint, I think that's the foundation of all devices that are out there. They always benchmark against this gold standard, which is called the polysomnogram or the in lab sleep study that folks might be familiar with if they've ever been into a sleep clinic or a sleep lab, there is,

Rachid 00:13:31 That's where you would stick things on my, on my head

Logan 00:13:33 Probably. That's right. Yeah. That's where I deliberately try and interrupt your sleep to get an estimate of how it is , uh, which is its own aspect of highlighting the fact that that's not a natural sleep and not maybe reflective of your true sleep. And so there are limitations even to that gold standard. Maybe it's not quite gold or standard, but that being said, almost all devices that track sleep try and benchmark against those. And to some degree, I mean we're measuring different signals, right? The sleep is a brainwave based phenomenon and we're using surrogate signals based on how those brainwaves filter through the motor system and the, what we call the autonomic nervous system. Basically how your heart beats and you breathe and things like that on autopilot. And so we can infer quite readily what stage of sleep your brain is in based on how your body is your body state at that time.

Logan 00:14:18 It's never gonna be a in lab clinical poly polysomnogram on your wrist. It's never gonna be at your bedside or under your bed. You really have to be watching brainwaves to get that perfect level of accuracy, but there's kind of like an event horizon almost nobody can get to a hundred percent accuracy, but almost all of the algorithms out there today, particularly with the sophisticated machine learning models that can actually estimate what state of sleep your brain is in at that moment are pretty good. And we're, you know, we're basically everybody in the pack is, has pushed the bounds of what is possible with these devices. So you, you get a reasonably good estimate. I think the most valuable part though is not tracking to an actual polysomnogram, but it's actually saying, one, we get the benefit of getting real world sleep, right? We're not deliberately interrupting your sleep to get those signals and uniquely we get to watch how they change over time, which is something I don't get to do in clinical medicine. I don't get to track what you did that day against a polysomnogram. Oh, you had caffeine today, you you had a drink today. What does a polysomnogram look like tonight? Right? I don't get that in the real world though. We get to leverage the longitudinality of data in the real world, undisturbed nature of it to really be able to make those strong and meaningful associations that impact your day-to-day sleep from what your behaviors are that day.

Rachid 00:15:32 What I found a fun fact is that this year mark's 15 years since the first Fitbit came out, which had sleep tracking right out of the gate. Do you know Conor, like how sleep tracking changed and improved in those 15 years?

Conor 00:15:45 Yeah, so the, the very first, uh, version of sleep tracking relied only on the movements of your wrist. So it had what we call an accelerometer to measure your movement and then what the next generation of Fitbit devices may be going back nearly 10 years now, added in the heart rate signal and by adding in the heart rate signal, you began to get a little bit of information to the different stages of sleep. So it wasn't just whether you were asleep or awake, it actually decided whether you were in potentially in deep or rem or light sleep. Over the years we've refined that algorithm, but that's where the current, uh, algorithms are working. And what we're beginning to look at now, and we see this in other research papers as well, is not just looking at the heart rate itself and looking at each individual heartbeat, but potentially there's even deeper information in the individual signals underlying that. So I think the next few years I think us and other manufacturers will see even a little bit more increase in, in accuracy and precision and how well we can track the different aspects of sleep.

Rachid 00:16:44 And then that's where we talk about Fitbit and something on your watch. You know, my my experience at Google is that every year we launch at least one thing where I'm like, how is that even possible? And a few years ago that definitely was the second generation nest hub, which did sleep tracking without actually, you know, wearing anything. It was just using a, a radar, I think, to get my sleep data. I'm still puzzled by that. Can any one of you explain to me how you make that work?

Logan 00:17:08 Yeah, well, I mean fortunately I worked on that project so I can speak to that, but it's basically what Conor alluded to already. It's tracking coincidentally the same signals, the same biological phenomena. It's just tracking them differently. And so the solely radar that was integrated into the nest hub with sleep sensing was able to look at a similar signal of body movement, so what we call actigraphy, but instead of just doing it on the wrist, it did whole body actigraphy, which is actually quite insightful, at least from a sleep medicine perspective. We put all these biosensors all over your body to capture all of the muscle movements. This gave a unique opportunity to look at how all of the limbs are moving independently instead of one limb as a measure of most of the body movements, which is a reasonable measure of that.

Logan 00:17:49 And then also based on how small of a movement the solely radar was able to pick up on. It not only captured the breathing, but also the heartbeats percussion on the chest wall. So essentially what we call a cardio ballistic artifact. Basically every time the heart beats you can see the change in the chest wall movement. And so those signals, those vital sign signals essentially are similar signals to what the, uh, wrist-worn devices are capturing. And so you can just look at the, what we call the autonomic nervous system. Basically it's is it steady during a slow wave deep type of sleep or is it highly volatile during a dream sleep period when your body is seemingly active but paralyzed. And so we, we basically captured a similar set of signals, body movements and the autonomic nervous systems impact on breathing and heart rate and then could actually estimate sleep stages based on that. Yeah, it was, it was amazing technology and when we looked at these signal plots to actually try and start building the algorithm, it was like, oh my gosh, we can totally tell this is a person breathing right here. And then we saw them move and it, it was, it, it truly was amazing and beautiful to be able to actually just track that not knowing who the person was or any information about them, we could actually see that they were sleeping and, and for what stages they were in.

Rachid 00:19:05 We're now at the part of the Made by Google podcast where we love to offer our listeners a top tip. Maybe we can do it a little bit differently this time and I'll ask both of you gentlemen. Logan, I'll start off with you. What is, when it comes to sleep, one thing I should start doing and one thing I should stop doing to get the best possible sleep?

Logan 00:19:23 So I think the thing that I would suggest is what Conor already highlighted is start by looking at how you feel and then reflect upon that to then look at your data and start from where you're feeling to then reflect. Does my data relate to that? And a grid opportunity that many folks may not even realize even if they have premium services, is that they can tap deeper and deeper into that experience to gain more understanding, teasing apart the overall sleep perspective, like the sleep score, and even go into those sub scores and even tap into those sub scores and find out what went into that sub score. So it's really helpful to first say, how am I feeling? And then look at the data dive deeply and say, is one of these things associated with how I'm feeling? The other thing that I would suggest people stop doing is over-indexing on those metrics, right? On a given day, right? Sleep is a long-term health perspective and so everybody is gonna have a bad night as you highlighted. It's really in the overall view that you need to actually look at and say, am I achieving my goals broadly? And if I'm reinforcing healthy habits on a, on the regular basis, one-off nights aren't gonna be too much of a burden for me. And so, you know, it's really trying to say like, am I overall achieving my health goals?

Rachid 00:20:32 Conor, what would you say? Starting and stopping?

Conor 00:20:34 Would say a habit to start if you don't already, is to really focus on the pre-bedtime routine, particularly around work. I think for people who are doing emails at 10:30 at night or who are kind of thinking about the meeting tomorrow, it's really difficult to get your brain to switch from that on mode to turning off. So finding your sweet spot for you personally is what will make your brain calm down a little bit. Whether that's reading, it could be listening to music, you know, reading a story to your child, you know, I think that's people Made by

Rachid 00:21:08 Google podcast. Yeah,

Conor 00:21:09 Yeah . If people prioritizing that, that like hour in before they're trying to temp seat, it'd be like trying to sprint a hundred meter race where you literally just got out of the couch a second ago. So trying to switch your brain from full on to full off is really not a fair expectation. I think on the stop, I, this is more a personal one, and this may be, you know, I'm getting a little bit older. Uh, I would say I love naps. I think naps are a great thing, but you have to be really careful if you take a nap quite late in the day. So let's say a lot of us come in from work at maybe 6:30 or 7 in the evening and we're tired. I say, I'll just have a quick lie down for 20 minutes, half an hour.

Conor 00:21:48 Then that night you can't fall asleep. So you kind of dissipate your, what they call your sleep pressure, your desire to sleep because you're kind of resetting your body a little bit. So sometimes you just have to, at the end of the day, if you're tired, you're probably outta a push through and then go to bed at 9:30 or 10 rather than saying, oh, I'm so tired after my commute, I gotta lie down and see. But I've noticed that more as I've got a little bit older, like a nap is a little bit more likely to mess my falling asleep later that night up.

Rachid 00:22:14 Logan Conor, thank you so much and can't wait to see what's next in the world of sleep

Conor 00:22:18 Tracking. Thank you very much. Yeah, thank you.

Voiceover 00:22:20 Thank you for listening to the Made by Google podcast. Don't miss out on new episodes. Subscribe now wherever you get your podcasts to be the first to listen.