Desiree Sevelius

The Delphi Method – Theoretical Base for the Use of Delphi Method

The epistemological bases for futures studies are the least developed aspects of futures studies in general, and this also concerns the Delphi method which is probably the most often used method in technological foresight studies made by futurists (Bell,1997; 167 & Kuusi 1999;sum.). This limited theoretical study concentrates on the background, techniques and expertise in Delphi method, as a background study for a research project examining the possible consequences of using waste materials in future community building.

1. Introduction

A defense research project in the USA called “Project Delphi” started the development of a futures research method called the Delphi method. The project, concerning the use of expert opinion in future possible developments, was sponsored by the US Air force and carried out by the Rand Corporation in early 1950´s. During the following decades Delphi technique became a fundamental tool for technological forecasting, used especially in many technologically oriented corporations. The Delphi method spread to Europe and the Far East and applications of this method have been used also in research projects concerning environment, health and transportation. (Linstone &Turoff 1975; 10-11)

In the Delphi method chosen persons are usually recognized as experts in a certain field, and they are asked to give their future oriented estimations about the development of a certain matter. Delphi is often also characterized as a method for structuring a group communication process (Linstone & Turoff 1975;3), and as  a communication process it has a wide number of applications in different areas. However, it is not the different applications that determine the use of Delphi, but the circumstances that are surrounding the need of a group communication process.

In The Delphi Method, Techniques and Applications written by Harold A. Linstone and Murray Turoff in 1975, the writers describe Delphi almost thirty years ago: (1975;3)

“For, if anything is “true” about Delphi today, it is that in its design and use Delphi is more of an art than a science”

With this the authors wanted to underline the fact that the numerous applications of Delphi differ from each other both in theory and practice and together sum up a wide range of expert methods under the name of Delphi forecasting. Nicholas Rescher describes the Delphi method some twenty five years later in his book Predicting the Future, as a “ A widely used predictive process that proceeds by way of a structured interaction among a group of predictors” (1998; 92). Wendell Bell writes about the Delphi method in the widely recognized book Foundations of Futures Studies (1997:261): “ Fundamentally, the Delphi method is a version of survey analysis, particulary that form of survey research that involves repetitive questioning of respondents”.

2. Characteristics of Delphi Method

Linstone and Turoff divide the Delphi in two main categories according to how they are carried out in their book The Delphi Method, Techniques and Applications published in 1975. The “conventional Delphi”, is a paper-and-pencil version where a small monitor group designs a questionnaire that is sent to a larger group of respondents. The monitor group will present the answers to the group, witch will usually have at least one possibility to reevaluate the original answers based on the group answers. In the other version, in the “real-time-Delphi” the monitor group is to a large degree replaced with a computer. The computer doing the compilation of the group results eliminates the delay caused of summarizing of the results of each round before the next questionnaire (1975;5).

2.1. Definition

Osmo Kuusi summarizes that Delphi method means the research methods that distinguish the following features (1993;135):

  1.  For the research a group or panel of experts has been formed
  2. The experts will inform the research manager separately about their view on the matter. The first round of the questionnaire can be exact or be based on wider questions that will be more exact in the fort coming rounds.
  3. The experts will be given information and feedback about the other panelists answers, though anonymous, before the next round of the questionnaire.
  4. The experts have a possibility to change their answers, after getting the feedback, at least once during the process. In Delphi projects one of the main goals used to be consistency, but in Delphi inquiries today this is often not seen necessary or even desirable.

In what sort of situations is the applying of Delphi meaningful? There is a list of situations under which conditions the use of Delphi would be desirable (Linstone & Turoff, 1975;4 and Kuusi 1993;136):

  • The problem cannot be examined with precise analytical methods
  • Collective subjective estimations are probably useful in the judgement of the problem
  • The problem is wide, complex or interdisciplinary, and the participants lack a “common language” or method for tackling the problem
  • Time and cost make frequent group meetings infeasible
  • In the process a bigger amount of experts are needed, a group that will be to big for to work as a committee or working group
  • The possible disagreements between the experts supports the use of a transmitter
  • There is a need to work anonymous, to avoid the majority influence and the influence of very strong personalities
  • Additional opportunity has been added by the possibility of using computers as the media for collecting information, giving the feedback to the panel and reporting the results. The Computer Mediated Communication Systems can support the communicating system in different ways, the technique was developed and refined in the evolution of the Delphi method. (Turoff &Hills 2001:2)

2.2. The Policy Delphi

The Delphi method was originally used in dealing with technical topics with the goal of seeking a consensus within a homogenous group of experts. The Policy Delphi, developed in late 60´s in the US, seeks just the opposite, to generate the strongest possible opposing views of the possible resolutions of a remarkable policy issue. The expert role in a Policy Delphi is different; there are no experts, only informed advocates and referees. The Policy Delphi is usually seen as a tool for the analysis of policy issues, and not as a decisions making mechanism. The process has been developed partly to replace committee work that has developed its own problems[1] as the companies using it grow bigger. (Turoff 1975;84)

From the policy Delphi other applications of Delphi methods have been developed, for example the Argument Delphi developed by Osmo Kuusi, will be briefly presented later in this study. Other application of the Delphi method like The Trend Method[2] have been developed to be usable methods for specific topics of inquiry.

Some common reasons for failure in a Delphi process, according to Linstone and Turoff are following (1975;6):

  • imposing monitor views and preconceptions of a of a problem upon the respondent, and not allowing other specifications of the problem for discussion
  • to expect Delphi to be a surrogate for all human communication in a given situation
  • the summarizing and presenting techniques for the group response are poor and ensuring common interpretations of the evaluation scales used in the questionnaire
  • ignoring disagreements, so that discouraged panelist drop out and an artificial agreement on the topic is reached
  • underestimating the panelists time and need for some sort of compensation for their work, if the Delphi is not an integral part of their job

3. Delphi Critique

In his book Delphi Critique, published in 1975, Harold Sackman widely criticizes the traditional Delphi method and discusses the principles as well as the applications of Delphi using more than hundred Delphi inquiries as material for his study. The impact of Sackmans crushing critique led to the decreasing use of the technique both in Europe and the US for almost two decades until new interest in Delphi methods and applications could be noticed in late 1980´s. The technique was during all this time used in Japan in many wide inquiries mainly concerning future technology applications.

Sackmans critique clarifies and expresses his opinion that Delphi is an unreliable and scientifically invalidated technique and that the results of most Delphi experiments are probably unreliable and invalid. The reason for this critique is that Delphi method does not meet the necessary criteria of generally accepted standards of strict empirical experimentation in social sciences, and Sackman suggest that “conventional Delphi should be dropped from institutional, corporate and government use until its principals, methods and fundamental applications can be established experimentally as scientifically tenable”.(Sackman, 1975:3-4). To examine Delphi as a discipline that is expected to meet with certain experimental social sciences methods is to my understanding not a valid criteria for the method. In Delphi the aim is to get reliable information on future oriented issues by experts or persons that might be in a position to make decisions as experts concerning processed matter.

In his critique Sackman sums up the evaluation of conventional Delphi, both method and applications, in sixteen clauses where he criticizes the method for  lacking minimal  standards both in opinion-item analyses and social science methodology, also neglecting reliability measurement and scientific validations on findings. He also claims that no serious literature on the subject has been published.  Sackman suggests further that the method requires forced consensus based on group suggestion. Moreover he claims that the Delphi method gives an exaggerated illusion of precision, misleading uninformed users of results.

According to Kuusi (1999;77) Sackman ignores in his critique the special point of Delphi studies, because the critique is based on the comparison of the scientific criteria of opinion polls with Delphi studies. Kuusi points out that there is fundamental epistemic differences between opinion polls and Delphi studies: in opinion research the idea is to is to find opinions or behavioral dispositions of the object of the study, but the main idea of a Delphi study is to find relevant arguments concerning future developments. As Kuusi also points out, literature concerning the methodology of Delphi had by the time of the critique indeed been published, but was not noticed by Sackman.

Sackman criticizes repeatedly the use of experts and questions their ability to participate to the questionnaire with any information that would be better or more valuable than a non-expert could present. According to Sackman anybody could play the Delphi game.(Sackman 1975;40-41). To my understanding the conclusive difference is that ordinary people will give reliable answers to questions asked according to their knowledge, but they are usually not in a position to make decisions concerning the matter in the future, as experts in a field can be expected to. Their answers and expertise is stressed with functionality and decision making and trough that with responsibility of decisions made.

Mika Mannermaa has also presented some critical points towards the Delphi method (Mannermaa, Kuusi 1993;139-140), which can be summarized as:

  1. The answers seems to spread according to the time distance of the forecast, the further in the future the matter of the forecast is the more spread the answers are.
  2. The exactness of the forecast is questionable. The experts are often pessimistic on questions concerning events on a long span, and optimistic concerning short coming events.
  3. The questionnaires reflect the knowledge and interests of the maker. Long questions can also be to difficult to understand similarly, and short questions again cause problems in interpretation.
  4. The more useful or economically profitable a project is the more probable and faster it is estimated to come true.
  5. The crisis happening in twenty years time does not seem as bad as if it was to happen in two months time. Also things happening in near past seem to have a bigger affect on the answers than incidents that happened a long time ago.
  6. People prefer certainty instead of uncertainty in their answers and plans. The future is seen to be happening in an environment very much like the present one, and the choices are made on those potions. Still the future will happen in different environment, nothing we can know now.
  7. The tendency to mix the probability and the preferability of a certain event. In this case the answers are likely to present the panelists own wishes instead of presumed occurrences.
  8. To make a plausible prognosis requires a broad expertise in the matter, and often the expertise used in the panels is not wide enough.

The answer to this critique can be seen as a general critique towards peoples ordinary, and to my understanding normal, way of behaving. Without doubt the first point (1) is true, in the sense of the further the forecast, the bigger the variances because the small variances will go in different directions and the differences will grow with time. To imagine a futures map, with the different scenarios as paths on the map illustrates to my understanding this as a normal situation. (2) The fact that one is pessimistic on long term and positive on short term question is a well known characteristic among people in general, also in the literature within futures studies. The experts are by no means an exception, and why should this especially be a burden to the Delphi method? (3) Difficulties in asking questions are always the same, how to make an other person understand exactly what you want to say. To be sure of the understanding one can instead of paper or computer questionnaires use personal interviews, where the problem of misunderstanding can be minimized.

Of course the more profitable a project is, the more likely it is to come true(4), there are many people that will work hard to get his or her own profits from it, this is the way the world works as long as it is directed by the market economy. Point five (5) is to my understanding similar to number two, and the same justification functions for it. The next two points (6&7) in Mannermaas critique are relevant and need to be adjusted so that misunderstandings and vagueness is minimized. The last comment on the possibility of narrow expertise can of course be attended to designing the panel as multidisciplinary as possible and to widen the concept of expertise.

4. The General Theory of Consistency

Osmo Kuusi uses the General Theory of Consistency (GTC), a theoretical framework developed by him and first published in 1974, as a base for a working epistemic foundation for futures studies in general, and especially for Delphi studies concerned with technological topics.

The core of GTC is a dual description of the reality using two kinds of beings: learning beings and not-learning beings. The not-learning beings cannot change their behavior, the truths concerning them now and in the future is the same. But because there are also the group of the learning beings, that can and will change their behavior, the future is partly indeterminate.(Kuusi 1999;14).

What kind of things can be seen as not-learning beings? The kinds that do not change their behavior in the future. Upon this can be obtained the criteria of sameness; a not-learning being cannot change its criteria of sameness[3], whereas the learning being can. This can be explained with the following example: A prediction, made by a researcher observing the behavior of a being, is based on sameness. Without the understanding of sameness there cannot be any relevant expectations of the beings behavior in the mind of the observer. The behavior of the not-learning being can be predicted according to the criteria of sameness, it is behaving the same when all the conditions for the observation are the same, in other words the scientific situation will have the same results always when all the conditions concerning the observed situation or object are the same.  The learning being on the other hand may have changed its criteria of sameness, because an experience of contradiction or inconsistency, and that is why the researcher observing the learning being cannot predict its behavior the same way as the behavior of the not-learning being.(Kuusi 1999;18).

The behavior of a learning being can also of course have stable elements. Because these invariances, patterns of behavior, traditions, expressed preferences and images of the future, can be changed by learning, Kuusi calls them transient invariances and in comparison with the (permanent) invariances of not learning beings. He also uses de concept of reliability as a common way to describe the stability of a transient invariance. (1999;19)

The following features define a learning being:  (Kuusi 1999;19)

  • the being can change its behavior as the result of its experiences
  • the being has interests, which direct its behavior
  • the being has an active memory, where its experiences are stored

According to Kuusi if any of these requirements are not met, the behavior of the observed being is possible to predict, and it can be classified as a not-learning being. In practice the limit between these two kinds of beings could be “ the possibility to predict the behaviour of the being so that the being cannot nullify the prediction by changing its behavior” (1999;19). Kuusi clarifies the epistemological meaning of learning and not-learning beings for future studies and Delphi methodology as following: “ The point is that the focus of scientific foresight is not to make outsider predictions based on prejudices of actors but to find what will happen in the future, if relevant actors act reasonably and use their leaning capacity as far as possible.” (1999;80).

5. Basic Types of  Futures Oriented Expert Knowledge

According to the General Theory of Consistency one can find four possible types of expertise on the future (Kuusi, 1999;35):

  • the expertise on the invariant behavior or invariant criteria of sameness of learning and not-learning beings
  • the expertise on the capacity limits of learning beings
  • the expertise on the interests of learning beings
  • the  expertise on the capability limits of learning beings

Kuusi refers to scientists or experts on invariances when he means knowledge about the invariant behavior of not-learning beings, and the prediction based on the assumption that some learning beings behave as they used to behave. Kuusi calls the group that have a large supply of relevant resources (wide capacity limits) and relevant interests (decisions about the generic technology belong to their capacity limits) decision makers or the makers of futures. The relevant difference between scientists and decision makers (that can be examined in commercial companies) is that the decision makers have been more interested in ways to widen capacity limits to achieve different targets, and the scientists have been more interested in finding invariances.

The third main group of  futures experts are called synthesizers, and they are masters of relevance. This means that because they are able to understand which invariances, capacites and decisions are most important, they are able to make good syntheses by anticipating the interplay of factors that shape futures. (Kuusi 1999;35-36).

Three types of experts about futures (Kuusi 1999;36):

1. Scientists
Knowledge on invariances: permanent invariances: criteria of sameness of not-learning beings transient invariances: habits, routines, equilibrium points of learning actors

2. Decision makers
Real and perceived capacity limits; perceived interests and routines;
Real and perceived capability limits

3. Synthesizers
Relevant invariances, relevant capacity limits, relevant interests and relevant capability limits

Kuusi uses different kind of scenarios to examine these three different types of expertise in practice. The use of futures research is usually connected with decision making in companies, and without this connection (where futures research is used as a tool for decision making), the futures research is refered to as a waste of time. Also decision making without futures research is refered to acting without freedom, because the one cannot be aware of the possible alternatives for action and then one has not true freedom of choice. (Kuusi 1999;37). Tacit or hidden knowledge is often mentioned as one way to have knowledge. This concerns also expert knowledge because often all the information a person possesses, cannot easily be expressed in words, but it still is an important director of our actions (Kuusi 1999;108-109).

6. Reality Construction as a Product of Delphi Interaction

The interaction between persons in a Delphi process can be seen as a way to construct reality. The reality is our own, but as a product of interaction also a shared reality. We call our tacit assumptions about things reality. We carry around several different realities, depending on the interactions and things we share with others as well as our personal convictions about how things are. We have parental realities, professional realities etc. These realities give meaning to our interactions, and these realities are confirmed and developed by our acts and interactions and conversations with others.(Scheele1975;37).

In a Delphi process the product of each panel is the reality that is defined trough the interaction of the panel. (Scheele 1975;37). Sometimes a group reality can be renegotiated, as can happen in a Delphi panel process. The important thing is to find ways to manage this process in a way that will be but intentional, also meaningful to the participants. The panelists are aware of the differences of these realities they presume, and as a result of an interaction they can produce a common reality of situation or question they have worked with. How this will happen depends on the work of the Delphi monitor[4], who will conduct the inquiry.

Sam Scheele refers to a theory of reality suggested by the phenomenologist Maurice Merleau-Ponty, where reality is a current prevailing shared assumption about a specific situation. This means that reality is a product of our experiences and not an external, objective state of things (1975;43). We collect situational definitions about how things are, change some opinions due to new situations, and reconstruct our reality within the society were we live.

According to Scheele, this means that when conducting a Delphi process one has to consider how reality is defined and processed. He lifts the following four points to discussion (1975;46-48)

  • The results of Delphi are produced by interaction, the reality constructed is  a common or shared reality of the participants of the panel.
  • If the reality is a negotiable construct it is important to look for information that can shape reality, identifying new ways to think and also introducing new options both in doing and thinking.
  • The reality that will be constructed will be different at least as much as the technology will develop during the time that is under discussion. The future development is not anything given for the people of the future, but something they themselves developed for their own needs.
  • We can expect reality to go on being negotiable, meaning that the point from where different things or possibilities will be looked upon, depends totally of the reality that will be negotiated at that time. This reality is shaping things so that it will wary from time to time what we consider plausible, feasible and desirable.

7. The Delphi as a Group Communication Process

7.1. The influence of organizations in a Delphi process

In projects where The Delphi method is used, usually include situations where the panelists represent different organizations. The society is built on organizations and most of us spend a big part of our time in these. The organizations are different and functioning on different levels in our society.  Kindergartens, schools, universities, organizations for hobbies sports and leisure, as well as companies, councils and administrative organizations are recognized as selfevident and necessary constructions of a modern society. The organization can be formal or informal and their purpose and role vary.

In conducting a Delphi inquiry it is important to recognize the organization the panelists represents and to evaluate the importance of the commitment to it, in order to estimate the influence of the organization on panelists and their answers. Even if the organizations work in dissimilar ways, they normally have a goal they work for or a concept as a base for their operation. The goals of the organizations represented by the panelists as well as the panelist personal goals can either be simultaneous or differ from each other. When an expert represents an organization, is he expected to answer as a member of the organization he represents or as a private person using his expert knowledge?  A third viewpoint is the imagined or outspoken goal that is set on the group working in the Delphi process. Therefore the question of how the different organizations and their goals affect both the group behavior and the personal standpoints and answers in a Delphi process, can, depending on the issue, be adequate.

7.2. Organizational conflict

Conflict is a term used in many situations, but concerning organizations it is usually linked with the situation when the individual or a group experiences difficulties in selecting an alternative for action. Conflict in decision occurs in individual decisions making, individual or group conflict within on organization or a conflict between organizations or groups. (March & Simon 1959;112). Conflict in individual decision making arises in different ways from which March and Simon in their book Organizations, published already in 1959, distinguishes three major ways; unacceptability of alternatives, incomparability of alternatives and uncertainty of the consequences of alternatives. How an organization reacts to conflict can on the other hand be divided in four major processes: problem-solving, persuasion, bargaining and “politics”.(March & Simon 1959;129).

Organizational conflict in a Delphi process is also dependent on perception and identification. The most significant identifications are according to Simon & March (1959;153) organizational and professional identifications. People could be expected to behave differently depending on how big the conflict between the organization or the profession they represent and the standpoint they could take as a member of the panel is. We are on the other hand expected to behave rationally, but rational again is a concept with no ready given implication, and people do behave rationally only relative to some given identified characteristics of a certain situation. These given things include: (March & Simon 1959;150)

  1. Assumptions about future events or conditions and the probability of them
  2. Knowledge of alternatives available for action
  3. Knowledge of consequences attached to alternatives
  4. Rules or principles for ordering consequences or alternatives according to preference

7.3. Decision making as a part of Delphi

When the purpose of a Delphi project is to gain information and to create a picture of the future, it is also necessary to consider the decision making process that will take place even if the construction of the future reality is hypothetical. Creating a future, even a hypothetical one, does include making choices and deciding what options and possibilities are primarily in order in carrying out a chosen vision.

To make decisions usually has a background in solving some kind of problems, or trying somehow to avoid them. Science and technology play a major part in the way we construct our reality today, it is said to be the most powerful religion of modern times. When we talk about problems in our technically constructed physical world, we usually assume that a solution to the problem exists (Linstone 1984;9). A complex problem usually has many possible solutions, that will create new questions. Instead of solving the original problem we tend to shift the problem into new questions or projects that need to be done.

When making decisions, the decision maker will make his choice for action from a whole set of alternatives. To each set of alternatives a set of consequences is attached, consequences that will come true if the particular alternative is chosen. Kuusi presents the alternatives as three categories (1999;86):

  1. Certainty: theories that assume the decision maker has compete and accurate knowledge of the consequences the will follow on each alternative.
  2. Risk: theories that assume accurate knowledge of a probability distribution of the consequences of each alternative.
  3. Uncertainty: theories that assume that the consequences of each alternative  belong to some subset of all possible consequences, but that the decision maker cannot assign definite probabilities to the occurrence of particular consequences.

When problems are being solved and decisions are being made, it is important to be aware of what we are viewing, and how we are looking at the problems. (Linstone 1985;39). What are we looking at? A system, or a subject of inquiry usually involves nature, man, society and technology in different possible combinations. The main interest in a Delphi inquiry is, due to its use in technical applications, often looking at sociotechnical systems, where the problems deal with both the technical and the social and human factors dealing with the subject. Being aware of what sort of system is the object of our interest, will of course influence the construction of the inquiry, choice of panel, questions and the utilization of the results.

The other interesting question is; how are we looking? Linstone describes (1984;44) three viewpoints from where and how we examine the systems of our interest; a technical perspective, an organizational perspective and a personal perspective. The different perspectives are used to view and understand different problems. A technical element needs to be understood by a technical perspective. A personal perspective to a technical question or problem would give an inadequate answer. Using multiple perspectives[5] is becoming the common way to examine all kinds of systems.

Looking at the possibility of making choices it is necessary to consider behavior, and what is rational behavior. Capacity and capacity limits are concepts that define the possibilities of learning, and real capacity limits are boundaries of rational behavior (Kuusi 1999;20-21). Furthermore it is reasonable to suppose that within the real capacity limits of a person there are more choices to be made than only one. From the actors point of view any of the choices can be rational within his or hers way of perceiving the reality, even if it might look different from an other actors point of view. This leads to an interesting conclusion: the rational future is not determinate, but instead there are many rational futures. (Kuusi, 1999;21-22) This will leads to the conclusion that the decisions made in a Delphi inquiry are only one of many rational possibilities, and what is asked and how the background information is given and how the questions also influence the answers collected in the inquiry.

7.3.1 Group influence on reality construction in a Delphi interaction process

The conception the group has about itself, affects the products of a Delphi interaction. In Delphi investigations group members are expected to view the group differently, partially because they are often anonymous. Being in the panel has according to Scheele (1975;56) no particular meaning for the panelist, but all the same a part of the energy and attention in the panel will go to trying to define the panelist internal relationships. This means that the panelists experience of the group and its identity, tends to shape the individual answers.

Here the role of the Delphi monitor or the manager is important. He or she will shape the conceptualization of the group, which is assumed by the different panelists. This concept of the group, usually implicit, dictates the way of interaction between the panelists. It also determines the kind of reality one uses as a frame of reference in participating in the panel and the shaping of the common reality of the group. Until each group member is comfortable with the conceptualization of the groups image or nature and believes that it also will affect the common reality in the agreed direction, it will be difficult to produce meaningful contributions. The conceptualization depends if the panelist interaction is of its nature tarncations, experiences, episodes, events, affairs or occurrences.(Scheele 19975;57).

Each Delphi interaction produces a shared reality that will be formed by the panelists from their own expectations and also how the material is presented in the beginning of the process. This reality is based much on how successful the interaction in the panel will turn out to be. The size and shape of the reality where the things are being discussed is more important than the specific descriptions produced by the panelists. (Scheele 1975;64).

8. Argument Delphi

Earlier the Policy Delphi was mentioned as different variant of the Delphi method, that has a goal to generate the strongest possible opposing views of the possible resolutions of a remarkable policy issue. The Policy Delphi is usually seen as a tool for the analysis of policy issues, and not as a mechanism for decision making. Osmo Kuusi has developed a variant he calls the Argument Delphi, which has focused on the production of relevant arguments. As a base for Argument Delphi  Kuusi uses three types of reasonability of argumentation. He makes a distinction between three types of the reasonability of a judgement: predictive reasonability, option reasonability and commitment reasonability.(1999;116) According to Kuusi the epistemic value of an argumentation process, or a Delphi exercise depends on what sort of reasonability can be expected.

The features of the Argument Delphi are following (Kuusi 1999;128):

  • The names of the panelists are informed to all the panelists, but the answers are given anonymously
  • The main purpose is to make to argue seriously (like is done in the Qualitative Controlled Feedback procedure). The panelists have direct contact only with the Delphi managers during the first round
  • The Argument Delphi is mostly focusing on option reasonability

The Argumentation Delphi is also based on rules on argumentation and on modern argumentation theory and on collusion on critical discussion (Kuusi 1999;128-134). Kuusi suggests that the evolution of the Argument Delphi in the future would be to concentrate more on commitment reasonability and identifying more specific options. Also because similar topics have been discussed in technology foresight studies in different countries, Kuusi proposes that it would be reasonable to use international panels in future Argument Delphi projects  (1999;207).

9. Designing a Delphi

Scheele points out that the conceptual productiveness[6] is often lost in a Delphi process because the contests and the subject are not concrete enough (1975;65). The mental processes in most people involve something concrete to try your thoughts and imagination on. The subjects must not be general propositions but concrete examples of the discussed subject. The questions should proceed from specific details and everyday circumstances to general propositions.

The results from a Delphi inquiry are often looked upon as “captured “ from the Delphi process. The results should instead be seen as created from the interaction process, the Delphi monitor playing an important part in creating the reality where the subject is being processed.(Scheele 1975;68)

9.1. The Panel

Creating a panel is an important part of designing the Delphi inquiry. To get a mixture of panelists who do as well as possible cover the accurate field the panel should consist of following groups (Scheele 1975;68):

  1. Stakeholders – those who will be directly affected
  2. Experts – those who have a suitable or relevant experience
  3. Facilitators – those who can and will organize, clarify and stimulate

Just how the panel should be constructed, like the amount of representatives from each group, is of course dependent on the application of the Delphi. There are no exact rules how a panel should be designed but the groups are usually dominant when following circumstances are acute:

1. Stakeholders
– options and interests are clear
– the direction of action is unclear or fractioned

2. Experts
– the actors are obvious and pointed out
– it is unclear how to act

3. Facilitators
– issues, relationships and values are unclear

According to Scheele (1975;69) in aggressive moments the experts will start acting as stakeholders and in passive moments the facilitators will start acting as stakeholders. Important issues when conducting a Delphi are motivating the panelists, directing the interaction between the panelists, interpreting the responses and sharing the results in a useful way.

In communicating the results it is important to:

  1. create involvement [7]
  2. generate interest [8]
  3. present the results interactively [9]

9.2. The Delphi method as a research and communication process

The Delphi method has numerous applications and several recipes about how to conduct a Delphi inquiry. Finally I will present three different views on the Delphi process. These are by no means opposite views but complementary viewpoints to the same method.

According to Linstone and Turoff a Delphi process usually has four distinct phases (1975:5-6):

  1. Exploration of the subject under discussion
  2. Reaching and understanding of how the group views the issues
  3. Significant disagreement is explored, and finding the reasons for it
  4. A final evaluation, when all information has been gathered and analyzed

According to Wendell Bell the Delphi method generally involves an eight steps  research and communication process (1997;262-263) and he presents it as following:

  1. The specification of some topic or subject whose possible, probable and preferable futures are to be investigated.
  2. The construction of a questionnaire as an instrument or data collection
  3. The selection of some individuals (respondents) whose opinions are to be studied, usually experts (to serve as “oracles”) on the topic being investigated
  4. The initial measurements of the opinions of the respondents by means of a questionnaire
  5. The preliminary organization and summary of the data resulting from the initial measurement
  6. The communication of the results of the initial measurement of opinions as feedback to all the respondents (reminiscent of the stimulus or “treatment” in standard pretest-posttest designs).
  7. A re-measurement of the opinions of the respondents as they have been informed and may have been changed by their knowledge of earlier results including of other respondents’ supporting comments for their opinions.
  8. An analysis, interpretation, and presentation of the data and the writing of a final report.
  9. Model for a group solving process

One of the advantages of groups is, that different persons with different perspectives and skills contribute to the processing of complex problem for which they have the appropriate knowledge as well as the ability to contribute to the solving of the problem in process. One typical model for a group problem solving process according to Turoff and Hilz is the following (1999;3):

  1. Recognition of the problem
  2. Defining the problem
  3. Changing the representation of the problem
  4. Developing the goals associated with solving the problem
  5. Determining the strategy for generating the possible solutions
  6. Choosing a strategy
  7. Generating the evaluation criteria to be applied to solutions
  8. Conclusions

This limited study concentrated on the theoretical aspects of Delphi method and the interesting applications and their results were deliberately outlined. I am however aware of the burden such a one-sided study carries, when the reader is not able to examine the theories trough the applications belonging to the subject. The study has nevertheless fulfilled one function; forcing the writer to try to understand the complicated theoretical side of a practical and widely used method in futures studies. Although the theoretical side of Delphi method is interesting and diverse, I am well aware that in this study I have hardly scratched the surface of one section of the theory of futures studies.


Bell, W. 2000 (Third printing).Foundations of Futures Studies. Human Science for a New Era, Vol 1. New Jersey: Transaction Publishers.

Mannermaa, M. 1991. Evolutionaarinen tulevaisuudentutkimus, Tulevaisuudentutkimuksen paradigmoja ja niiden metodologisten ominaisuuksien tarkastelua, Helsinki: Vapk-kustannus.

March, J.G. & Simon, H.A. 1959. Organizations. London: John Wiley & Sons, Inc.

Linstone, A.L. & Turoff, M. 1975. The Delphi Method. Techniques and Applications. Massachusetts: Addison-Wesley Publishing Company.

Linstone, H. 1984. Multiple Perspectives for Decision Making. Bridgining the Gap between Analysis and Action. Amsterdam: Elsevier Science Publishing Co.

Rescher, N. 1998. Predicting the Future. An Introduction to the theory of Forecasting. New York: State University of New York Press, Albany.

Kuusi, O. 1993. Delfoi-tekniikka tulevaisuuden tekemisen välineenä. In P. Vapaavuori, M. (ed.) Miten tutkimme tulevaisuutta. Kommunikatiivinen tulevaisuudentutkimus Suomessa. Helsinki:Painatuskeskus Oy.

Kuusi, O. 1999. Expertise in the Future Use if Generic Technologies. Epistemic and methodological considerations concerning Delphi studies. Acta Univesitas Oeconomicae Helsingiensis, A-159.Helsinki: Helsinki School of Business Administration.

Sackman, H. 1975. The Delphi Critique.The Rand Corporation. Toronto: Lexington Books.

Scheele, S. 1975. In P. The Delphi Method. Techniques and Applications. Massachusetts: Addison-Wesley Publishing Company.

Turoff, M. & Hilz S.R. 2001. Computer Based Delphi Process. In Söderlund, S. (ed.) Tulevaisuudentutkimuksen tieteelliset menetelmät. Turku: Tulevaisuudentutkimuksen VerkostoAkatemia.


[1] According to Turoff problems in committee work are many, see Turoff 1975:86.

[2] In the Trend Method, the panel will focus on one chosen trend or a set of related trend variables (Turoff & Hilz 1999:10)

[3] A being behaves in similar ways in similar situations, based on its own criteria of sameness. Sameness of behavior in Osmo Kuusis theory he also means the behavioral language of beings, the language might be different from an other beings perspective it might look different. (Kuusi 1999, p.248)

[4] Sam Scheele uses the term Delphi monitor for a person who conducts the Delphi study. The monitor will prepare the material, interprate the responses and prepare the results. S. Scheele; Reality construction as a product of Delphi Interaction in The Delphi Method by Linstone &Turoff, 1975.

[5] Multiple perspectives are viewed in Multiple Perspectives for Decision Making (chapter IV), Linstone, 1984.

[6] The conceptual productiveness Scheele means as the interplay between intentions an circumstances

[7] this usually means engaging users in other processes where they will deal with the same subjects

[8] this usually means sharing the awareness of the Delphi inquiry with colleagues and constituents

[9] For example to organize an exercise to find out which results are applicable and when and how they might be used.