Delphi Method Report

In this paper, Dave Ziffer, one of the original founders of the Illinois Loop, described his first encounter with a facilitated school meeting using Delphi/Alinsky techniques to cultivate a desired conclusion.

He discusses the structure of the meeting, and tells how this led him to try to research the history of the technique. He also raises questions about the appropriateness of referring to the technique as "Delphi."

Dave wrote a follow-up article in 2000, adding more information.



---



Delphi Method Report
by Dave Ziffer
September 4, 1996

Over the course of my discussions with all you fine folks and on one of the videotapes that Elyssa lent me, I have occasionally heard people mention something called the "Delphi Method", which is said to be used by the administrators of school systems when they get the local citizenry together for a pep rally.

Basically the idea goes something like this: An organization (such as a school district) has already decided what it is going to do, but it wants to avoid the appearance that it has acted without public approval. So it schedules a public meeting which is advertised as being held for the purpose of soliciting community input. In fact the organization has no desire to solicit opinions; rather the real intent of the meeting is to give the community the impression that input was solicited.

The meeting goes like this: Everyone arrives and sits down. After an introductory talk, audience members are told that they are going to be divided into N groups (N might be any number depending upon the audience size). So everyone is asked to count off from 1 to N going up and down the seated rows. Each person is then directed into a group whose name is the same as the number that he or she counted off. So for example, if you counted "2" when the count came around to you, then you go to Group 2. There are, of course, N group leaders who then carefully direct the discussion of each group. Each group leader controls the format and, to a great degree, the content of each group's discussion.

Toward the end, group members are asked for inputs, which are then listed on a big sheet of paper by the group leader. At the end of the meeting, everyone reassembles. The sheets from all the groups are posted around the room and each group leader reads the list of suggestions that are on his list. Unpopular or unusual inputs are glossed over and downplayed, while the majority of attention is focused on those ideas that are generic enough that almost everyone would agree to them. At the end a summary is delivered and everyone goes home.

The purpose of the countoff is to split up anyone who came in together, so that each of the N groups will most likely consist entirely of people who have never met, and almost certainly will not contain any two or more like-minded individuals who may have come in together. Once everyone is split up in this manner, the group leaders can then easily control the group conversations because dissenting or outspoken individuals are generally alone in each group, and such people quickly discover that they have little or no support from other group members.

The purpose of the sheets of paper is to graphically display to everyone, at the end, that everyone's opinion has indeed been registered. And finally the whole meeting format is designed specifically to avoid community input; obviously a bunch of strangers divided up into groups and chatting for 2 hours are not going to accomplish anything of importance.

I know about this technique because I have seen it used in community meetings put on by the schools here in Batavia. Thinking that this was the "Delphi Method", I decided to do some reading and find out more about it. What I discovered was a total surprise.

First of all, it was very hard to find anything at all about the Delphi Method. It does not appear in ordinary library databases; I had to work with a librarian and look in some special academic databases to find anything. What we eventually discovered was about 10 or 12 books on the subject. The earliest and seemingly seminal book on the subject was written in 1969 by a guy named Norman C. Dalkey for the Rand Corporation (a division of the Air Force) and it is entitled, "The Delphi Method: An Experimental Study of Group Opinion". Certainly this seemed to be the right subject so I ordered the book and it arrived from inter-library loan within a week.

This book is actually a bound research paper and surprisingly, it seems to have nothing to do with the technique that I described above, except that it deals with the subject of group opinion. In fact the "Delphi Method" described in this book is radically different and the intent of using this method is precisely the opposite of the community meeting method described above.

Briefly, the Air Force was justifiably concerned about making the best judgements about long-range issues where no concrete answers could be obtained. For example, they might want to know how quickly non-nuclear nations might develop the capacity to develop their own nuclear weapons, or they might want predictions of global climate 50 years from now. Dalkey refers to all these sorts of judgements as "opinions". Let me quote from the paper:

"There is a kind of technology for dealing with opinion that has been applied throughout historical times and probably in more ancient times as well. The technology is based on the adage, 'two heads are better than one,' or more generally, 'N heads are better than one'. Committees, councils, panels, commissions, juries, boards, the voting public, legislatures ... the list is long, and illustrates the extent to which the device of pooling many minds has permeated society.

"The basis for the N-heads rule is not difficult to find. It is a tautology that, on any given question, there is at least as much relevant information in N heads as there is in any one of them. On the other hand, it is equally a tautology that there is at least as much misinformation in N heads as there is in one. And it is certainly not a tautology that there exists a technique of extracting the information in N heads and putting it together to form a more reliable opinion. With a given procedure, it may be the misinformation that is being aggregated into a less reliable opinion."

So anyway, the research paper is about a procedure called the "Delphi Method" which goes like this:

You have a group of experts in a field but they disagree to some extent (as experts always do) and you want to get a more reliable opinion than any one of them could deliver alone. If you simply stick them in a room together, you might get a good answer but you might not, because the dynamics of personal human interaction can get in the way of the facts. Dominant personalities, for example, can lead a group to ignore the inputs of weaker personalities. Or people who need group acceptance may be willing to forego their possibly correct opinions to avoid being excluded by other group members. So instead of having a normal meeting, you split everyone up - totally. They never meet. You solicit input from each member of the group entirely separately, by mail or E-mail. You then collect all the written inputs, summarize them, and send the summary back to all the group members. The group members then formulate a second set of opinions based on this controlled feedback and submit a second opinion. This goes on until the group reaches a point of relative stasis, i.e. peoples' answers aren't changing significantly on successive iterations.

If you think about it, the intent of this technique is precisely the opposite of the community meeting method I described. The obvious intent is to accurately gauge the opinion of a group of people whose opinions are obviously desired and respected by the people conducting the "meeting". The irony of the name was not lost on Dalkey (who did not coin the name) because in the paper he makes a note about it, saying that "Delphi" is "a somewhat misleading appelation, since there is little that is oracular about the methods."

According to the research paper the technique actually works; by experimenting using almanac-type questions whose answers were actually known ahead of time, but were of such a nature that most members of any group would be unlikely to know the true answer, the researchers discovered that groups that were managed using this method fared better than ordinary groups using face-to-face interaction. They also found that Delphi "groups" tended to converge toward correct answers with successive iterations.

So anyway I just wanted everyone to know about this. There may very well be a method out there called "Dephi" that has something to do with the stuff that goes on at our school meetings, but if that is the case then it is something entirely different from the Delphi Method that was developed at the Rand Corporation in the 1960s. I don't know about you, but until I find out more about "Delphi", I'm going to be careful about using the name.

MORE: Dave wrote this follow-up article in 2000, adding more information.