One thing we learned from our testing was, the absolute speed and accuracy weren’t actually the critical point. It was really the overall experience...
So with the [Apple] Newton [handwriting recognition] and other technologies, you would write something, and you would go one Mississippi two Mississippi, and boom it would show up. And if it’s wrong, oh crap. You erase it, and then you try again like one Mississippi two Mississippi. But with Graffiti you’re just going a-b-c-d-e, and if you make a mistake, like your keyboard on your computer, you just have this backspace stroke, and so it’s very fast to recover from errors, so you just keep going.
I think there's a parallel here with swiping gestures for words on smartphone keyboards, where I find that when the gesture is misinterpreted, it's frustrating because I'll have to delete the entire word and redraw it, or tap out individual letters. Which just feels bad.
As a side note related to that, in the same class there was also a guest lecture from Shumin Zhai, who worked on SHARK and subsequent systems which would eventually become ShapeWriter/Swype [2]. Very cool overview to how the system works conceptually.
[1] https://scs.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=...
[2] https://scs.hosted.panopto.com/Panopto/Pages/Viewer.aspx?id=...