The technology we will be using for the Global Assembly Dialog is called InterMix Group Dialog Software.  A single instance of the software, refered to as an InterMix hub, runs on a windows platform with SQL server as the database.  It uses dot-net technology, written in C#.

As terms are introduced, they are shown in bold dark red. Hopefully this will be helpful rather than disconcerting!  To begin, here are a couple of diagrams to help with understanding.

An InterMix hub can have many dialogs and many groups.  The hub has an administrator, each dialog has one or more coordinators and each group has one or more moderators.

A group has members.  A group can join a dialog by applying to the dialog coordinator.  When a group joins a dialog, its members become participants of that dialog as members of the group.  The technology is designed to be intuitive for the participant, but the administrator, coordinators and moderators have to understand how a dialog works.

Networks are groups, with the extra capability of being joinable by other groups.  A network cannot, however, join another network, so we only have one level of networking; there cannot be networks of networks.  Having joined a dialog, a group can join a network within that dialog by applying to the network moderator.  Both the group moderator and the network moderator must OK the join of the group to the network.  When a group joins a network, its members participate in the dialog as if they were members of the network.  The members of the group that joins the network do not actually become members of the network.  However, because the network is itself a group, it can have its own direct members.  The direct members of the network and the members of the joined groups participate together in the dialog as if they were all members of that network; there is no distinction between direct and indirect participation.

Participants may be a member of more than one group, but for a particular dialog, they must choose one group for participation in that dialog.  Participants can change from one of their groups to another at the beginning of a new round but not within a round.  See directly below for more about rounds.  A "Change Perspective" function allows a participant to swap groups for the next round, and even to join another group for the purpose of swapping to it.  While participants may only participate in a particular dialog as a member of one group at a time, they may participate in multiple dialogs within the same hub simultaneously.

A dialog also has rounds.  Rounds occur sequentially, following each other one by one.  Rounds are created by the dialog coordinator.  There are two types of rounds, unity rounds and diversity rounds.  These are also known as together rounds and apart rounds.

Every round has two main phases, a writing phase and a rating phase.  The writing phase must precede the rating phase in each round.  Later we will be adding an optional discussion phase between the writing and rating phase.

In a unity round, all participants are treated as members of a single overarching group, which may be thought of as the common humanity of the participating groups.  Participants in a unity round write messages knowing that they will be rated by all participants from all groups.  In a diversity round, participants write messages to be rated by members of their own group only.  Thus in a unity round a single message is elected to represent the common humanity of all the participating groups.  In a diversity round, each participating network or group that has not joined a network elects a message to represent its own diverse perspective.

Candidate messages are written by participants in the writing phase of a round.  Participants are not required to write a message, but every participant has the opportunity to write just one message for that round.  Each message has a subject and a body.  Participants can come back over and over to edit their message until the writing phase ends.

An important feature of the process is that many messages can share the same subject.  Moreover, the highest rated message for each particular subject is specially distinguished at the end of a round as a subject winner.  This feature therefore allows for many winners each round instead of just one.

When a participant first begins to write a message in a round, that participant is presented with a list of subjects already used by other participants, with the subject that has the most messages at the top.  The participant may either choose a subject from the list or start a new subject.  Depending on whether it is a unity or diversity round, the list of subjects presented are from all groups or only from the particpant's own group.  Again importantly, the dialog coordinator in unity rounds, and the group moderators in diversity rounds may seed the process with pre-selected subjects thus kindling the imagination of the participants and if not exactly steering the discussion, certainly influencing its direction.

The rating phase is divided into sections.  Each section winnows the list of messages until at the end of the final rating section, only one message is left.  Having a single winner provides the drama and closure that the human imagination revels in, but actually the number of final winners can be set to more than one if that is what is wanted.  The number or percentage of surviving messages can be set by the dialog coordinator for each section.

Because group sizes may differ dramatically, in a diversity round, one group may have more rating sections than another group.  For instance, a group with a thousand participants may have 300 messages in a round and require three rating sections, with the first winnowing the list down to 60, the second winnowing the 60 down to 8 and the third finally producing a winner.  While in the same diversity round, another group might have only 30 participants and only 10 messages, requiring just one rating section to produce a winner.

In each rating section, each participant is doled out a preset number of messages to rate, normally somewhere between 5 and 10.  Care is taken in unity rounds that messages are evenly distributed over the groups.  Participants do not have to rate all their messages in one go.  They may come back over and over.  They are not required to rate their messages at all, and it is OK if they rate only some of the messages that were distributed to them.

Each message is rated on two scales, from 0 to 4 for interest and from -3 to 3 for approval.  A total score for each message, called the value of the message, is calculated as the average approval rating times the sum of the Interest ratings.  The winning messages are those with the highest positive value.  A message may have high interest, yet get low approval.  A good example, from a previous version of InterMix, was a message suggesting convicts be giving the option of being executed in return for $20,000 for their families.  This message got a high negative value because it got high interest ratings and very low approval ratings.  By separating interest and approval, we greatly increase the general interest of winning messages and correspondingly reduce the possibility of electing platitudes.