Calvin has a crush on Amazon Polly. It’s the invisible voice of Pollexy that speaks and he listens–to him it’s a real person. But most importantly, he follows her instructions . . . mostly.
We are not computers.
Underneath Calvin’s body of an 18-year old boy is a now-adult brain, battling the maturity of a 3-year old. Let’s face it, kids are kids–18 or 3. They don’t always want to do what they’re told.
At his age, boys are falling in and out of live, experimenting with new hobbies and generally just finding out who they are. Calvin is different, but he’s not.
On the other hand, if there’s one thing a computer can do better than anything, it’s doing the same thing, every time. That’s good, right? That gives us stability, right? Yes and no. Humans are not computers–and this is the drum my wife continues to beat.
“Did you know that the first Matrix was designed to be a perfect human world? Where none suffered, where everyone would be happy. It was a disaster. No one would accept the program. Entire crops were lost. Some believed we lacked the programming language to describe your perfect world. But I believe that, as a species, human beings define their reality through suffering and misery. The perfect world was a dream that your primitive cerebrum kept trying to wake up from.”
— Agent Smith, from The Matrix
In other words, human brains cannot deal with . . . Pure Perfection. Absolute consistency. 100% Discipline. While Olympic athletes have incredibly strict regimented diets and schedules, in reality, their food quantity, quality, their timings all have micro-variations that can subtly break up the monotony. And that brings up one of my wife’s common issues with care automation–Calvin will quickly get used to the voice and the schedule and start tuning out.
Care automation systems must adapt to human nature.
I’m currently working on adding imperfections to my care automation system to help constantly keep the system fresh for Calvin in a few different ways:
Imperfection #1: Changing voice and syntax
Thankfully, Amazon Polly has dozens of voices to choose from, so that’s the easiest piece to implement. The real challenge is dynamically changing sentence structure on-the-fly so he doesn’t hear the same repetitive phrases over and over again. I’m currently looking at an approach using NTLK . . . let’s just say that this is a lot more difficult than you can imagine.
Imperfection #2: Ever-changing visual cues
Adding Philips Hue Bulbs routines to flash, fade, change color, I believe, will really help in this area. I’m building a standard set of different visual cues and then capturing results to see how he responds.
Imperfection #3: Randomizing audio cues
There’s a lot of potential here to keep the experience different–from musical introductions, to background music during speech, animal sound effects, sirens . . .
I’m just starting work in this area, but I think this is crucial to a successful care automation system. There’s so much potential here to use machine learning to understand what works for him and automatically adapt the system to his likes and preferences. I hope to post much more about my progress on these fronts over the next year.
Leave a Reply