There seems to be a new comedy trope around a misinformed character having an augmented reality filter turned on during a video conferencing session. r²dr (Simpsons reference).
Now, AR experts at Apple are racing to apply their facial recognition feature to just eyes. Iris unlock was a feature I used on the Samsung S8 but, at least in my experience, took a long time and often failed. I also use iris scanning whenever returning to Canada.
Facial recognition has benefits and drawbacks depending on the application. For unlocking a phone, it has the advantage of speed and convenience and the disadvantage of requiring additional verification because it can be hacked and that glasses, masks, and hats can create unlocking issues.
For AR implementations, just tracking movement through eyes is enough to create a believable effect. No one is looking to fool anyone on AR (yet). However for any refined movement or preventing the AR image from unlocking on an object, the system needs to be trained more with masks, glasses, and hats.
“Hey, it looks like you’re wearing a [mask/hat/glasses/eyepatch]… is it OK if we take 20 seconds to improve accuracy while you where [item]?”
It’s possible that with a large dataset of people wearing masks, AR systems might be better at understanding our queues while wearing these devices than us. Also, what’s a chin for?