Yes and no. I've tried this sort of thing before, and obviously the machine can't know when you are actually reading a paragraph slowly or when you've just been distracted. It would have to be clever about weighting an overall average with a short term average. I'll be curious to see how they do it, definitely.
Good points. Given enough data it could do a decent job at normalizing lows & highs, enough to be useful anyway.
It would be interesting to see a proximity sensor in the iPad. I'm not sure what the range of those is. The accelerometer might be sensitive enough to detect when it's being held up as it's going to move a little bit in your hand.
Yeah, that's where I went from "this could be useful" to "that's stupid". It's like doing an internet speed test to determine how long it will take to download a file, ignoring how fast the file has actually been downloading.
It's one thing to calibrate, it's another to rely on context-sensitive data and pretend it is universal, when instead you can just rely on the context-specific data for any specific context and refine the estimates as you go. To my knowledge, web browsers do not "calibrate" like this.
