Eight years ago, I almost died. I’d had chest pains for a few days but thought nothing of it as I was (I believed) a relatively healthy, gym-going person. Okay, so the “gym-going” may be a slight exaggeration, but I had a membership and did drive past it on a regular basis. Occasionally I even went inside. But after a few days of chest pains, and a nice meaty pizza for dinner (ahem!) I started to get worried. So I went to my local urgent care, which initially considered nothing to be wrong with me as my pulse rate was low (athletic-level low – see, healthy!). But after running some more tests, they called me back and sent me to an ER.
A few days, and stents later – with a wonderful video of one stent being inserted just as my “widow-maker” artery was closing for the last time – and I was on the road to recovery. Being a bit of a geek, one of my first purchases was a smartwatch to warn me of future issues. My second purchase was a smart scale because, well, once you enter the Quantified Self rabbit hole, you may as well keep going. And going I did: somehow I ended up with a second smart scale. The same make – even the same model – yet one fat-shamed me while the other stroked my wannabe athletic ego.
The weight remained consistent across both devices but my Lean Body Percentage varied by six percentage points between the two (again, identical!) scales. A glance in the mirror suggested that the lower number was a little optimistic: if there was a glimmer of a six-pack lurking, I had it very carefully protected with some padding. But still, we listen to the number we want to, right?
And so here we are again, questioning smartwatch and related data. One smartwatch I owned used to tell me I was underground occasionally (see That Sinking Feeling) while another one lied to my face about my quality of sleep (see I Dream of Rats). Now I’m not even sure if I can believe the body fat percentage of my scale.
“So what?” you may say. After all, the mirror does a pretty good job of telling us the reality of our situation. But there are (at least) two issues that I see. The first issue is that the US Health Secretary has proposed heavily marketing fitness trackers/smartwatches to more consumers as part of the MAHA initiative, which is great for the industry if it happens, but if these consumers start to truly rely on the device’s stats as a health indicator then we need to make sure the data is accurate.
Even if that initiative comes to nothing, there’s the current base of smartwatch owners. As AI is increasingly added into devices, AI-based fitness coaches are beginning to set goals for us all based on the data at hand. If my scale is lying about my fat level, the AI will believe that I’m healthier than I am – just as the Urgent Care initially thought I was fine based on my heartrate when I was less than a day away from sliding off this mortal coil.
And this is a far bigger issue than simply the smartwatch market; rather it impacts the entire AI-infused tech world. Garbage in, garbage out is going to be the greatest challenge for AI-based solutions moving forward - especially as AI becomes more tailored a an individual - and between a scale that generously ignores my “padding” and a smartwatch that misreads my sleep patterns, that’s two rather significant chunks of garbage heading into whatever AI coach I could use. Further, this is not likely to be an issue isolated to health and fitness: while there is so much focus on the “magic” of AI solutions, there appears to be far less import placed on the quality of data that is pulled into the calculations. And that could give us all some pretty quirky results for a wide range of AI advice. Buckle up and enjoy the ride, but perhaps, keep your hands on the steering wheel occasionally.