hello, computer!

A quick first posting, which will also lay out a little of what I'll try to do in this blog...

Before I even start, let me detour. In the Star Trek 4 movie, the crew goes back in time and has to use a computer from the 1980s. Scotty, the Star Trek chief engineer, says "computer!" to it a couple times with no response. Someone hands him the mouse to use. He says "ah", and speaks into the mouse, "hello, computer!"

(Unfortunately the clip isn't available on Youtube.)

The point is not to rouse Trekkies or mock speech recognition (which has come a long way since the mid-80s, even), but to show a "naive" interaction with technology. He interacted with it the way he always has -- indeed, in what is sometimes* the most intuitive way possible, talking to it the way he'd talk to a person.

We need to design products that are immediately, intuitively usable. A person uses it without thinking about the fact that they're using it, without wondering how to use it.

This is a simple but deep concept. All the time we build products with interfaces determined by the way the features are set up in the system, not by the way people will use them. We talk about "users" and "interaction", for that matter.

I feel very strongly that technology should adapt to people, not the other way around.

A common example of the wrong way of doing things -- the other day I called a utility and was told "say 'one' to talk to maintenance..."

1. Why should I say "one" instead of e.g. "maintenance"? What the system designer did was simply port over the touchtone system to a speech-rec system, changing the input resulting from "press one" to that from "say 'one'." No thinking involved by the system designer.
   
   This asks the caller to make that mental connection, 1 = maintenance, I want maintenance so I say 1. When touchtone was all we had, that was a necessary step. Now, it shouldn't be.
   
   
2. Very simple (but common) problem with the order of what was said to the caller: "say 'one' to talk to maintenance". This makes sense from a flow-chart perspective (the caller says "one" -> they get put through to maintenance). But it also requires the caller to listen to the whole phrase, then remember the beginning of it. OK, that's right, I want maintenance... what was that number again? And remember, this was one of 7 or 8 options that the caller was hearing. The system designer could have easily avoided this by phrasing it "to talk to maintenance, say 'one'." With this wording, the caller hears and thinks, right, I want maintenance, and then hears what they're supposed to do.
   
   In other words, when we're listening to instructions, they need to be in the order "to accomplish this, say this" rather than "say this to accomplish this". The second way requires the user to remember what it is they're supposed to say until they hear whether it leads to what they want to accomplish.

(A sidebar re order-of-listening issues: I hope the traffic-radio folks pick up on this. It's not so useful to hear 60 seconds of traffic with "there's a four-car accident and a major back-up on Route 880 northbound in Fremont" when the driver who's half-listening hears only Route 880 northbound in Fremont -- oh no, what was that again? Instead, the traffic-radio announcers should give the location first, allowing their listeners to pay close attention when it matters.)

'Nuff said.

- Bruce

* I'll discuss in a later post where speech recognition should work and where it shouldn't -- not because of technology but because of how our minds work.