I'm developing a command based application where the user asks an agent to do stuff using a Verb--Object--Parameters pattern. e.g.
- SCHEDULE a meeting with bob for Tuesday
- BOOK meeting room at 9pm
- ADD "Buy milk" to todo list
So you can imagine it a bit like:
- Launchers like Quicksilver / Alfred
- Siri
- Unix command line
- Chatting with a person
- Text based adventure games
A key problem with all of these interfaces is that there are no affordances. That is the user does not know the 'actionable possibilities' which are available. One has to learn what actions are possible.
I think we can improve things a lot by using Quicksilver/ Alfred style autocomplete with visual icons to assist the user in finding commands and in filling in the object and parameters correctly. But this does not really help people discover what possibilities exist - per say. Indeed most non power users never use launchers like Quicksilver and Alfred for anything other than basic search and application launching so I wonder if there is a better way.
Is there any existing research/experience on ways to make a users aware of the 'actionable possibilities' within "command based" systems?