The first round will produce a single joint message from the combined Jewish and Muslim participants.  In the first round and the other "together" rounds, one message is selected from all participants.  In the "apart" rounds, two messages are selected, one from each group.  "Together" and "apart" rounds alternate.

In every round after the first, the winning message(s) from the previous round will be prominently displayed in the email invitation to both groups to write a response for the next round.  For the results of an "apart" round, the message with the higher average interest rating will be displayed first.

The dialogue rounds will each have three cycles.  First cycle is the writing of the messages.  Messages will be due by Thursday at 8 am, no exceptions.  The dialogue is automated and there is no one to appeal to if you are late with your labored over and much loved missive, so please get your messages in by Wednesday evening, latest.  You will receive an email reminder on Wednesday morning.  Messages must not exceed 400 words.  Messages that exceed 400 words will be returned to the sender for editing.  

Second cycle is the rating of the messages.  Ratings will be due by Saturday at 8 am.  Each participant will be emailed a link to a list of 10 or 11 randomly selected messages for reading and rating on two scales: 0 to 4 for interest, and -3 to +3 for approval.  Each message will be distributed an equal number of times, and, for common rounds, equitably divided over the two groups.  The messages for common rounds will be anonymous until the conclusion of the round unless the author includes his or her name in the text of the message.  The score for each message will be the sum of its interest ratings times the average of its approval rating.  As you can see, it is possible for a message to be interesting, but disapproved, a combination that puts the message at the very bottom of the results.

The third cycle is an instant runoff vote on five top messages from the second cycle, provided they received a majority of positive or neutral approval ratings.  Each participant in the group will receive the same 5 messages to be ranked 1 through 5.  The result is reached by repeatedly tallying each participant's remaining high ranking and removing the one message that received the lowest tally.  In this method, if the message that you voted #1 gets the lowest tally, then your vote shifts to that message you voted #2 and so forth.  Third cycle rankings are due by 8 am, Monday morning.

Theoretically it is possible that no message will receive a majority of positive or neutral approval ratings in the second cycle.  In that case, the dialogue will be stopped if it is a "together" round.  In practice, this is very unlikely to occur.

A major concern is to encourage participation, yet have the two groups weigh equally in the process.  To this end, the exact time of the votes will be noted, and in the "together" rounds, only the first n votes of the majority voter group will be counted, where n is the total number of votes cast by the group with the minority of voters.  Thus if there are 120 votes cast altogether, 70 by Muslims and 50 by Jews, then only the first 50 votes cast by Muslims will count, and the other way about, of course.  In the "apart" rounds, all votes will count.  It would be too cruel, and there is really no need to apply the same principle to the messages, since it is the voting that needs to be equal.  Therefore, if the same 70 Muslims and 50 Jews write messages, all 120 messages will be rated and ranked to come to a single winner for the combined group.  However, at some point, an imbalance in participation could become too great.  Somewhat arbitrarily, we have set that point at two to one, so we will not allow either group to register more than two thirds of the total participants.  

A list of participants and observers will be posted on the web, but will not be made visible to the outside public.  Full results of the dialogue, including the identity of the authors of the winning messages, will be made known to the public after the final round unless those authors choose to have their identities withheld.  Full results of each round, including the text and authorship of messages that did not win, will be made available to participants at the conclusion of each round. On the principle of the secret ballot, individual ratings will not be made known.  The computer code used to conduct the dialogue is open source and is (or soon will be) available for download at http://groupdialog.org.