Yesterday my wife received a call from John Kerry, the American presidential candidate. She was pretty excited, but realized fairly quickly that it was a recording. When she was done listening, she tried to delete the message.

‘No honey,‘ I said, ‘it’s a phone call.‘

‘Which number do I use to delete again?‘

‘Honey, it’s not a message, it’s a phone call.‘

‘What do you mean? I’m listening to a recording.‘

Fascinating. To her, listening to a recording (plus the reaction to the notoriety of the caller) was equal to listening to voice mail, a natural response. She was thrust into the mental model of listening to a voice mail by the characteristics, the stimuli, of the call.

Examining the stimuli that activate certain mental models would be a nice complement to Indi Young’s mental model thingy. So we not only react to what people already understand but also stimulate them into using information in a certain way.

PeterV’s story of automated phone system vs. web (why is the phone system, which forces one to navigate a tree, easier to use?) reminded me of a story from Chris Fahey of Behavior. Chris was designing an interface for a mobile screen, I forget whether it was a PDA or a mobile phone. Of course he was forced to reduce the interface to its fundamental elements. He subsequently would keep designing for a screen that small even when working on websites, pretending he had only that much space and concentrating on the fundamentals of the interface.

I think that’s why the automated phone interface is easier, there’s no auditory ‘room’ to introduce clutter, indirectly leveraging Hick’s Law.

Whereas websites are like our homes: the clutter expands to fill the available space.

New Architect’s By Design: Wisdom from the Industry gives off lots of sparks, especially this one from Henry Lieberman, Agents Research Lead at MIT Media Lab:

Applications and menu operations are like hammers and screwdrivers. Each is specialized to do a particular task. There’s nothing wrong with tools, but having a different tool for each task means that if you want to do too many things, you get too many tools, and it takes too many steps to do anything.

The alternative is to cast the computer in the role of an assistant or agent, like a travel agent, a secretary, or a real estate agent. Computer agents don’t have to be as smart as a person, but they do have to be proactive. They do have to learn from interaction with the user, and they do have to be sensitive to context.

I think another advantage might be in the form of directly replacing tools with agents, even if the number of tasks doesn’t decrease, the amount of repetitive effort will via automation. As Lieberman says, agents can be like ‘macros on steriods‘.