CASCADE researches have analyzed processes required for collaborative authoring. In those cases where a process consists of a set of task of which only a small sets are important to the author, the process been redesigned to offload the ``housekeeping'' tasks to the computer. For example, we found that making a hypertext comment required 10-15 actions (depending on how the process was defined). Only one of these tasks, the comment itself, actually required the commenter -- all the rest could be done by the system. We call this kind of redesign ``augmentation'' in the sense that the term was first used by Douglas Engelbart in his early work in the 1960's.[2]
There are many auxiliary processes in the collaborative authoring process that take up a great deal of time. For example, mailing a revision to a group of co-authors can take up a great deal of someone's time. It shouldn't however, because somewhere, the system knows who the people involved are, and how they can be contacted, and it knows what document was just worked on and what review means. Shoshanna Zuboff[11] uses the term ``informating'' to refer to a process parallel to automating, but where the process involves the application of an information stream generated by a computer to some human endeavor in contrast to the power applied in the automation process. Thus, in CASCADE, opening the mail function begins with the ``Subject'',``From'', and ``To'' fields filled in -- the ``Subject'' is the document you were working on, the ``From'' is the author, and the ``To'' is the members of the group responsible for the document. CASCADE knows the email addresses, and how they like to receive their mail. Clicking one button will attach the document. The user simply types the note and the document is off. We call this kind of redesign ``information'' in the Zuboffian sense of the term as a direct parallel to ``automation''.
There are many things that we would like others to do for us. With advances in computing power and techniques, it is increasingly possible to design simple software programs that appear to be intelligent -- and in some cases the learning algorithms and control logic merit that title. In CASCADE, the current plan for agent design focuses on simple ``contribution'' agents that do little tasks on behalf of the user. For example, the communication agent keeps track of where people ``live'' when they are using CASCADE. Knowing that, the agent can go check those places when someone wants to contact them. The CASCADE communication agent does just this. It keeps track of who a user wants to talk to and watches for them. When found, it let's the user know, and if the timing is still right, sets up an interactive talk session. We call this kind of redesign ``substitution'' in the sense that the agent undertakes some well defined task on behalf of the users and substitutes its effort for the users.