Use of Internet Relay Chat (IRC) in Distributed Consensus Forming Billy Vaughn Koen Mechanical Engineering Department The University of Texas at Austin Austin, Texas 78712 ABSTRACT The growing popularity of the internet has profound implications for engineering education. This paper describes a one and a half year study of the potential for using one aspect of the internet, Internet Relay Chat (IRC), for merging the opinions of people distributed over the United States. In principle a similar strategy could be used for self-paced instruction, cooperative problem solving, and distributed engineering design. Internet Relay Chat (IRC) with multiple chat sessions configured as a tree in conjunction with a suitable consensus forcing strategy such as the Nominal Group Technique offers a new, original, state-ofthe-art implementation of an Electronic Town Hall. This strategy is effective because it is interactive, realtime, flexible, and offers simultaneous collection of opinion on a state, regional, and nation-wide basis. Although it relies heavily on modern technology, an important characteristic of this approach is that 90 per cent of the participants do not have to own a computer or be computer literate, and the computers that are required may be standard, privately-owned PC's instead of expensive workstations. Twelve pilot studies have been completed to demonstrate the potential of this strategy. In the largest of these, twenty-six computer sites, from twenty states, and including approximately 200 people throughout the United States were interconnected. In some of these pilot studies, the computer videoconferencing software, CU-SeeMe, was used to allow the participants to see each other in real-time.

Introduction The rapidly increasing availability and popularity of the Internet has provided many exciting, new capabilities for politics, education, and society, but is has also introduced problems worthy of research because of its lack of central management, wide geographical distribution, and reliability. This paper describes 12 highly successful pilot studies in distributed consensus forming over the last year and a half in what we will call an Electronic Town Hall(ETH). In principle a similar strategy could be used for selfpaced instruction, cooperative problem solving, and distributed engineering design. The Electronic Town Halls in these studies were constructed using: (1) internet Relay Chat(IRC), (2) IRC channels organized as a tree, and (3) a consensus forcing technique called Nominal Group Technique. The largest of them was held on August 12, 1995, in conjunction with the National Issues Conference of United We Stand America. Twenty-six computer sites, in twenty states, and including approximately 200 people throughout the United States were interconnected. During this session, specific questions were generated at a central site in Dallas and passed down the tree, discussions were then held in living rooms across America, votes were taken, and a consensus was achieved. As part of this ETH, U.S. Rep. Marcy Kaptur (D-Ohio) answered a large number of questions from the people attending the ETH interactively and in real-time. The videoconferencing software for Macintosh and PC computers, CU-SeeMe, allowed anyone with access to the Internet to view the ETH and Rep. Kaptur live. This experiment has been widely reported in magazines such as Inter@ctive Week,1 and Media Bypass,2 reported in newspapers,3, 4 carried on the news wires,5 and mentioned on radio and television. These preliminary pilot studies have demonstrated the feasibility of creating consensus among large numbers of people distributed throughout the United States over Internet using this basic configuration..

Background, Nomenclature, and Representative Scenario Our baseline strategy for studying distributive decision making over Internet to force consensus in a large number of people spread throughout the United States is to interconnect them using Internet Relay Chat with IRC channels configured as a tree. Before we can understand this strategy, its three technical components: Internet Relay Chat, the organization of IRC channels in the form of a tree, and the consensus forcing technique called Nominal Group Technique must be understood. Each of them will be developed briefly in the next three sections followed by a description of the specific scenario that was used in the largest Electronic Town Hall held to date to tie them together into a meaningful whole. Internet Relay Chat

Internet Relay Chat (IRC) is a multi-user, multi-channel chatting network implemented on top of Internet where people from all over the world convene on "channels" to talk in groups or privately in realtime.6 At any time of the day, typically 16,000 people, from over 60 countries are chatting in over 5000 channels maintained by 100 operators. IRC gained international fame during the 1991 Persian Gulf War, when IRC users gathered on a single channel to hear and discuss updates from around the world. IRC was also used as the major, if not only, immediate source of information during the coup against Boris Yeltsin in 1993, during the earthquake in Los Angeles, and the earthquake in Japan. Figure 1 will make clear the configuration of IRC. A number of Internet nodes Ohio around the world are tightly connected and maintained as IRC servers. They are indicated in Figure 1 by the black dots interconnected on the ellipse. Anyone with access to Internet may connect his or her computer to one of Massachusetts these servers as a client as has been done by individuals in Alaska, Texas, Alaska Virginia Massachusetts, and Virginia in the figure. Typically, up to 250 people can connect to any one server at a time. Texas A message typed at one of these client computers is flashed around the world and may be accessed by other client Figure 1 programs almost instantaneously. IRC Configuration Once connected to the IRC network, a person may join a channel that has already been established or establish a new channel for himself. The person who establishes a channel has complete control over it. He can make the channel an open, private, or secret channel. He can limit who can change the topic for discussion on the channel and who may talk, etc. This is done by setting the mode of the channel. Tree Structure Although the PI has participated in a discussion on an IRC channel with over 300 participants, preliminary experiments indicate that newcomers to the electronic world will feel more comfortable if fewer than 30 or so people are talking at the same time. A group of 30 people is essentially equivalent to a typical focus group. It is uninteresting to use complex technical strategies to interconnect this small number of participants in an ETH. To increase the number of people in the Town Hall, a tree structure such as the one shown in Figure 2 must be established. ROOT

Conn1

Alaska1

Virginia1

Penn1

Mass2

SoCarol1

Ohio1

NwYork1

Wash1

Indiana1

Delaw1

Caucus#2

Oregon1

Minn1

Texas2

Kansas1

Texas3

Calif1

Wyoming1

Caucus#3

Mass1

Caucus#1

Texas1

Missouri1

Caucus#4

CONF

Caucus#5

Nevada1

NwMex1

Calif2

Caucus#6

Figure 2 Actual Tree Used This figure shows the actual tree that was used in the largest pilot study on August 12. It will be discussed in more detail later. It consists of a root, a number of interior nodes (6 in this case), and a number of leaf

nodes (19 in the figure). At least ten participants may be at each of these 25 sites (nodes plus leaves). In two of the pilot studies, an overhead LCD device to project the computer screen demonstrated that the number of participants at a site could be significantly increased. Information about the topic for discussion originates at the root, is quickly passed down the tree to the leaves, and the results are passed back up the tree. Note that the interior nodes are on two IRC channels at the same time to maintain the path for the information down the tree. The preliminary studies indicate that this is a feasible strategy. An important aspect of this configuration is that only a minority of the people who participate in the ETH needs to be computer literate. Each site needs only one computer and only one person needs to know how to use it. Therefore, at least 90 percent of the participants in the overall ETH do not need to be computer literate. A second important aspect of the tree structure outlined is that it permits information to be captured at any level of detail desired. With IRC a complete transcript of the discussions may be captured electronically. Therefore, we can evaluate the response to a question on a local, state-wide, regional, and national basis simultaneously. A final importance reason for using a tree structure for the Electronic Town Hall is demonstrated by the multiplier effect it allows. For example, if only ten interior nodes are established, if ten leaf nodes are established for each interior node, and if ten people are present at each interior and leaf node site, the number of people participating in the Electronic Town Hall would be approximately 1,100 people using a tree of only three levels such as the one depicted in Figure 2. This number is significant because it is approximately the number of people who are contacted by the poll takers when they report a poll with 3% accuracy. We have done pilot studies with, admittedly, a much smaller number of people participating, but with a tree of this "three level" complexity with no problems. Nominal Group Theory Random people making random comments will not produce worthwhile results. To obtain interesting data on the likes and dislikes of the participants in the Town Hall and to establish a representative decision of a group separated over a large distance, a consensus forcing technique is desirable. In the pilot studies we used an approximation of Nominal Group Technique(NGT)7, 8 with great success. For example, the root can propose a question such as: "Prepare a priority ordered list of what the national government should be doing at the present time." This question is passed down the tree. In small groups of ten or more in living rooms all over America the participants discuss, debate, lobby, and vote to create an ordered list of perhaps ten items. Now we must merge this list at each level going up the tree. To demonstrate this procedure, extract the left most interior node and leaves of Figure 2 as shown in Figure 3. NGT begins by asking one of the leaf nodes, say Alaska1, to report the entry that is first on their own priority list. Each of the other nodes (Mass1, Mass2, and Conn1) write it down as item 1 Conn1 on a combined list for the complete caucus. Then a second leaf node, say Mass1, gives the first entry on their list that has not previously been mentioned and this is made item 2 on the combined list. This process Alaska1 Mass2 continues in a round-robin fashion until six or seven items have been provided. Conn1 is included in this process, of course. Now each participant Mass1 (not each node) is given a fixed number of votes, say four, that can be distributed as desired. A participant can put all of the four votes on one choice Figure 3 or put some votes on one choice and the remaining on other choices. These Merging Results votes were tallied to produce the rank ordered list for these four sites. In parallel, of course, the other caucuses are preparing their lists in a similar fashion. Now we must merge the results of each of these caucuses to produce the overall list for the tree, again using NGT. This process was successfully completed in about 30 minutes in the pilot studies. Obviously the time needed for a specific merging operation depends on the number of individual sites that must be merged, but since much of the work is done in parallel, the time undoubtedly scales less than linearly with the number of participants. Typical Scenario The three important components of the ETH: Internet Relay Chat, Tree Structure, and Nominal Group Theory were discussed in the preceding sections. This section demonstrates how they are combined by describing one pilot study in some detail. At 9:45 a.m. CDT August 12, 1995 six computer sites (one each from Connecticut, Pennsylvania, Virginia, Texas, Wyoming, and California as indicated in Figure 2) joined an IRC channel called #guth on an IRC server that had been severed or guillotined from the general IRC

network. Each site established a second channel named #caucus1, #caucus2, etc. These channels formed the interior nodes of the final tree. As computers from other states including Alaska, Ohio, Massachusetts, New Mexico and others logged on to IRC at 10:00, they immediately joined their previously assigned caucuses. The home addresses of each computer site and the names of all people present are typed in for the caucus record and to ensure integrity of the voting. Approximately 10 people, many not computer literate, were at each of the interior and leaf node sites. Present were students, doctors, and housewives with ages from 20 to 90 years old. The news media including CNN were at some of these local sites. At the conference site, a special node of 24 people was attached to the tree in Caucus#5 with a large projection system to demonstrate that it is feasible to include large numbers of people at one location. This site simulated a typical chapter meeting. See Figure 4 for the geographic distribution of the computer sites interconnected for this study. (New sites since August have also been marked as "new" and indicated by a square.)

Figure 4

Status of IRC Electronic Town Hall

Key: root node leaf new

At approximately 10:10 a question that could be answered with a "yes" or "no" was passed down the tree from the root to the leaves. After ten minutes of consideration, discussion, and analysis, a simple vote was taken at each site (NGT was not used, of course, at this stage), merged by caucus, and passed to the root node for final tally. This overall result was distributed to all levels of the tree. At about 10: 25, a more complex question requiring the construction of a list of options ranked in priority order was passed down the tree to the individual sites. Each site, including the one at the conference, was required to develop a priority list representing their best response to the question which was then merged to produce a priority list representing their caucus and finally merged again to represent the best response for the nation. This entire process took about 30-40 minutes. Since separate computer transcripts were keep by state, region, and nation, a detailed analysis of the opinions by political entity were available. All states were invited back to the root channel, #guth, to watch the final merging of the separate lists, but IRC was configured so they could not interrupt by talking. Once the final results were in, the channel was opened and U.S. Rep. Marcy Kaptur agreed to answer questions posed by the 200 people throughout the nation in real-time. Throughout the entire Electronic Town Hall, a video camera connected to a computer using the video-conferencing software, CU-SeeMe, was focused on the computers that were managing the root node and people throughout the United States who were equipped with the appropriate software could watch the proceedings on their own computers. An important conclusion of this experiment is that it engaged all who participated and produced sound, mature, credible answers. Representative comments from the participants are

included in a later section to document the success of this pilot study. This conclusion replicates the results of the previous 10 pilot studies. One reporter commenting on the originality of this concept likened it to a combination of a focus group and a poll.9

Pilot Studies A series of 12 pilot studies were performed over a one and a half year period to evaluate the feasibility of using Internet Relay Chat for an Electronic Town Hall. Table 1 summarizes the results of this activity. Each of them had a specific research objective. Some of these pilot studies had the objective of testing an increase in size; others evaluated the effect of more people at a site; others investigated the different computer modes that could be used; and still others tried different tree configurations. Complete transcripts from the perspective of the root node are available for each of these examples. All of the studies No. Date Sites States were very successful with the exception of number five. In 1 9/22/94 4 1 this case the tree completely collapsed due to an unfortunate choice of IRC modes. Actually this experience proves the 2 10/ 06/94 6 1 complexity of correctly setting the appropriate channel 3 10/21/94 8 1 modes and the importance of research in this area. A 4 11/17/94 8 2 complete debriefing of this experiment is available. In the 5 12/15/94 6 3 last pilot study reported here, Mr. Russ Verney, Executive 6 2/09/95 12 6 Director of UWSA participated and answered questions 7 3/08/95 12 5 from the participants. One interesting aspect of this 8 3/30/95 4 3 experiment was that he was not actually at one of the 9 4/28/95 7 6 computer sites, but joined us by telephone. This opens up 10 5/08/95 15 8 the possibility of inviting busy politicians to participate in 11 8/12/95 26 20 this project in the future with a minimum of effort on their 12 10/29/95 20 14 part. Table 1 does not include numerous practice sessions with small trees that were used to train the people Table 1 Pilot Studies as they became involved in the project throughout the year and a half. The value of this ETH is not limited to the raw results produced in the voting. In addition, the attention of a large number of people is focused on a specific political issue and they feel they can make a difference. The individual discussions at the local, state-wide, and regional levels are as beneficial as the final vote. These pilot studies have proven that an Electronic Town Hall using IRC can achieve group decisions over Internet.

Typical Comments from participants Of equal, or perhaps of even more, importance than the technical aspects of an ETH to its success are the human aspects of it. It will only be successful and expand if those who participate in it feel it is worthwhile. Once again the advanced studies provide confidence that the human aspects have been carefully designed. Since the last study, the number of potential sites has arisen to 49 and the number of states represented, to 24 distributed as shown in Figure 4. This section contains unsolicited, representative comments from people who have participated in the past to document this point: "I must say, I was very impressed.. Both parties will 'likely look at [UWSA's weekend] as a benchmark' during the 1996 conventions." —Roger Schneider, Democrat's technological guru for the 1988 and 1992 conventions. Quoted in: Inter@ctive Week, August 28, 1995. "I got a surprise on Fri. the 11th, when CNN called to say they wanted to send a camera crew to our site...They were fascinated by the process. Quoting their reporter 'This is the wave of the future!" In our local living room discussion, the idea of 'free media time;' came up on our list. At that point the CNN people put their camera down and started participating! (although they did not vote)." —John Blare, El Cerrito, California "We had the greatest time last night! You knocked the socks off our chapter members. We had a total of eleven here, though some had to leave early and some came late. We'll probably see sales of modems skyrocket here in Lafayette as a result of last night! They had NO IDEA what

the Internet looked like and were stunned to realize the power of what you guys are doing with real time networking." —Char Roberts, Lafayette, California "All in all our group really enjoyed [it]. I think we helped re-energize some ETH people here. I think things went very smoothly and we all want to thank you for all the time and effort you put into this project. You did great! Everyone did great! We look forward to the next ETH." — Bob and Karen Powell, Springfield, Massachusetts "This could be a great advance over the one-way communications of the past (TV, radio, newsletters) for influencing not only UWSA directions, but political directions as well." —Glenn Busbin, Inman, South Carolina.

Political Disclaimer The pilot studies used members of the United We Stand America and their friends. This group was chosen because they were available, knowledgeable in general about Electronic Town Halls, and willing to help. Obviously the results of this study are not intended to be of use only to UWSA, but to improve the political dialogue throughout the United States. According to the Dallas Morning News,10 Democrats and Republicans are watching the results of this experiment.

References 1

Meeks, Brock, Inter@ctive Week, August 28, 1995, Vol. 2, No. 16, pg. 38. Watkins, Janet, Media Bypass, October, 1995, Vol. 3, No. 10, pg. 58. 3 Weiss, Jeffrey, "Perot hosts independents' day; Computer town hall satisfies participants," The Dallas Morning News, August 13, 1995. 4 Brazaitis, Tom, "Table is set for Perot's political tea party," Plain Dealer, August 11, 1995. 5 "United We Stand America National Issues Conference to feature interactive real-time on-line technology," U.S.Newswire, August 2, 1995. 6 Pioch, Nicloas, A Short IRC Primer," Edition 1.1b, February 28, 1993 (available on the net). 7 Scholtes, Peter, The Team Handbook, Madison, Wi: Joiner, 1988. 8 Van de Van, A. H., and Delbecq, A. F., "Nominal Versus Interacting Group Processes for committee Decision-Making Effectiveness.: Academy of Management Journal, Vol. 14, No. 2 (1971), pp. 203-212. 9 ob cit. Weiss, August 13. 10 ob cit. Weiss, August 9. 2

Biographical Data of Author Dr. Koen received B.A., Chemistry. and B.S., Chemical Engineering degrees from the University of Texas at Austin. He received a Diplôme d'Ingénieur en Génie Atomique from L'Institut National des Sciences et Techniques Nucléaires, Saclay, France, in 1963 and M.S. and Sc.D. degrees in Nuclear Engineering in 1962 and 1968 from the Massachusetts Institute of Technology. He has been a professor at the University of Texas at Austin since 1968. Dr. Koen was a visiting professor at the Tokyo Institute of Technology, June-August, 1994 and at École Centrale, Paris, France, January, 1983. In 1971-72 and 197677, he was a research collaborator in the Reactor Safety Division, Saclay, France. Dr. Koen produced a widely acclaimed definition of the engineering method. He, also, introduced to engineering education and developed a method of self-paced instruction (Personalized System of Instruction: PSI). Since its introduction in 1969, it has become an accepted alternate means of teaching. Dr. Koen has consulted for 10 companies, organizations and institutes and has been or is a member of 10 professional societies and nine honor societies including Tau Beta Pi and Phi Beta Kappa. He is listed in 18 directories including Who's Who in America and International Who's Who in Engineering. Dr. Koen has written or contributed to six books. His professional publications include 108 papers and he has presented 105 invited lectures. Professor Koen served for six years as Vice-President for Public Affairs for the American Society of Engineering Education (ASEE). He has held 25 different Society posts and he is a Fellow of the Society. Dr. Koen has received 15 local, state and national teaching-based awards. The most recent are: The 1992 W. Leighton Collins Award which is the highest American Society for Engineering Education award for service to education in engineering, engineering technology and allied fields; The Centennial Medallion, 1993 (ASEE) and The Helen Plants Award, 1995 (ASEE).