Auto-detecting vs. manually tracking activities

In order to keep tabs on both the iOS and Android ecosystems, I split my time between a Galaxy Note 5 + Gear S3 and an iPhone 7 Plus (yes, I like larger phones) + Apple Watch. I also bike to work daily and like to keep track of my rides (why is somewhat of an open question; it’s not like I’m trying to set speed records or anything).

The Gear S3 and Apple Watch have different philosophies for how to track my activities. The Apple Watch expects me to track all my workouts manually. Want to track a ride to or from work? You’ve got to manually start tracking the workout before you leave. And you’ve got to remember to stop it once you arrive at your destination. The failure modes are obvious: get thinking about something else (maybe a technical challenge you’re working on) and forget to start tracking, and your workout doesn’t exist. Or forget to stop tracking, and suddenly your ride home is twice its actual time.

The Gear S3 has a different philosophy. You can still start and stop tracking a workout manually, but the Gear S3 will also detect when you’ve started a workout and initiate tracking for you. And it’s pretty reliable: I can’t think of a single time that the watch has failed to detect when I start cycling, or when it’s thought I was cycling when I wasn’t (false positives and negatives, for those of you playing along at home). No more need to remember to start tracking your workout. The Gear S3 will also detect when the workout has ended and stop tracking for you. And that’s where it gets a little sticky. The watch doesn’t want to end the workout just because you’ve stopped moving; you might be stopped at a traffic light on a ride. So it typically takes it a minute or two before it decides you’ve really stopped your workout. It does try to then calculate back to determine when the workout really ended, but it’s still often off by a minute or so.

What I really want is a hybrid behavior. Autodetect the start of a workout; that’s really useful. But provide a manual method to stop a detected workout (as well as an autodetecting the end; that’s useful in case I forget). Strangely the Gear S3 doesn’t currently provide a way to manually end an autodetected workout, even though you can manually stop a workout you start yourself. With that one simple addition, the Gear S3 would provide workout tracking that I’d strongly prefer to Apple’s purely manual approach.

Of course, that’s for my commute ride. For running, I like the Apple Watch more than the Gear S3. That’s largely a size / weight issue, though. I helped out with some user tests of the Gear Fit 2, and it was actually better than both for comfort while running. But for general use I prefer the watch form factor.

So there you have it: human + AI would beat human alone and AI alone for tracking workouts. But then, I’ve long argued that AI + HCI can be better than either alone.