Delphi method

From Systems Analysis Wiki
Jump to navigation Jump to search

The Delphi method (also known as the Delphi technique) is a structured communication technique and a systematic, interactive forecasting method that relies on the opinions of a panel of independent experts[1]. The essence of the method is to conduct a series of anonymous surveys (questionnaires) in several rounds to obtain a consensus and a statistically aggregated group opinion on complex issues, especially when precise empirical data is lacking.

The fundamental principle of the method is that forecasts or decisions from a structured group of individuals are more accurate than those from unstructured groups or individuals[1]. This approach effectively combines the benefits of in-depth expert analysis with elements of the so-called "wisdom of crowds"[2]. The method was specifically designed to neutralize cognitive biases and social pressures inherent in traditional group work methods, such as meetings or focus groups[3].

History

The Delphi method was developed in the 1950s at the RAND Corporation, an American military research organization, as part of a top-secret defense project funded by the U.S. Air Force during the height of the Cold War[4]. Olaf Helmer and Norman Dalkey are considered key figures in its development[2].

The initial objective of the method was to forecast the potential impact of technological advancements on warfare. One of the first practical tasks for which it was used was to achieve consensus among military experts on selecting targets for atomic bombing[5]. Due to the project's secrecy, the first publicly available publication describing its application did not appear until 1962–1963[4].

The name "Delphi" was borrowed from the Oracle of Delphi from ancient Greek mythology, symbolizing the aspiration to obtain wise predictions. The authors themselves treated the name with irony, believing it had an "occultish" feel[1].

Key characteristics

The Delphi method is defined by four interconnected characteristics that collectively distinguish it from other group work methods[1].

Participant anonymity

Experts answer questions anonymously. This allows them to express their opinions freely, without fear of criticism, and minimizes the negative effects of group dynamics, such as groupthink, the halo effect, and pressure from authority figures[6].

Iteration and controlled feedback

The process consists of two or more sequential rounds of questioning. After each round, a facilitator (the study organizer) collects, analyzes, and anonymously summarizes the responses. This summarized information, including statistical summaries (e.g., median, quartiles) and key arguments, is provided to the entire group at the beginning of the next round. This allows experts to reconsider their positions in light of the group's opinion and their colleagues' arguments[1].

Structuring of the information flow

Interaction between experts is not direct but occurs exclusively through the facilitator. They play a central role: designing questionnaires, collecting and processing responses, filtering out irrelevant content, and preparing summary reports for feedback. This approach ensures the process remains objective and focused[1].

Statistical group response

The group's final opinion is expressed not through voting but through the statistical aggregation of individual estimates. This allows for an objective and quantitative representation of the group response, making the analysis process more rigorous.

Procedure

Preparatory stage

  1. Defining the problem and research objective. The research question is clearly formulated.
  2. Forming the working groups. Two groups are created:
    • Expert group (panel): Experts (usually 10 to 50 people) with deep knowledge in the study area are selected[7].
    • Analytical group (facilitator): A neutral facilitator is appointed, responsible for organizing the entire process.

Main stage (survey rounds)

  1. First round. The first questionnaire, often with open-ended, unstructured questions, is sent to the experts. This stage functions as a remote, anonymous "brainstorming" session[8].
  2. Analysis of the first round and preparation for the second. The facilitator systematizes the responses and uses them to develop a more structured questionnaire for the second round.
  3. Subsequent rounds. The second-round questionnaire is sent to the experts along with a summary report of the first round's results. Experts answer the questions, having the opportunity to revise their initial judgments. If an expert's opinion deviates significantly from the majority view, they may be asked to provide additional reasoning. The iterative cycle is repeated several times (typically 2–4 rounds) until a predetermined criterion is met (e.g., stabilization of responses)[1].

Analytical stage

After the final round, the facilitator conducts a final analysis and prepares the final report. The report describes the methodology and presents the results, including areas of consensus and any remaining disagreements with the arguments from each side[9].

Variations and modifications

  • Classical (forecasting) Delphi: The primary goal is to obtain the most accurate forecast. It assumes that a "correct" answer exists, and the group's task is to approach it. It involves a homogeneous group of specialized experts[10].
  • Policy Delphi: The goal is not to achieve consensus but to structure and clarify the full spectrum of opinions and arguments on a complex issue that has no single correct solution. The outcome is a "map of opinions" for the decision-maker. It requires the participation of a highly diverse group of experts and stakeholders[11].
  • Real-time Delphi: An accelerated version that uses specialized software to process responses and provide feedback in real time, significantly reducing the study's duration[1].

Applications

The Delphi method has found wide application in various fields:

  • Strategic analysis and business forecasting: Forecasting market trends, assessing demand, identifying risks[2].
  • Technological forecasting (Foresight): Identifying promising directions for scientific and technological development.
  • Public policy development: Analyzing complex socio-economic problems and building consensus among interest groups.
  • Healthcare: Developing clinical guidelines, setting priorities in medical research.
  • Education and information systems (IS): Developing curricula, forecasting trends in technology development.

See also

References

  1. ↑ 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 "Delphi method - Wikipedia". en.wikipedia.org. [1]
  2. ↑ 2.0 2.1 2.2 "What Is the Delphi Method, and How Is It Useful in Forecasting?". Investopedia. [2]
  3. ↑ Gordon, T. J. (1994). "THE DELPHI METHOD". Eu-Med AgPol. [3]
  4. ↑ 4.0 4.1 "Generating Evidence Using the Delphi Method". RAND Corporation. [4]
  5. ↑ "Delphi method | Research Starters". EBSCO. [5]
  6. ↑ "What is the Delphi Method - Pros, Cons, and Examples [2025]". monday.com Blog. [6]
  7. ↑ "How To Conduct a Delphi Study". Cardiff University. [7]
  8. ↑ "Delphi Technique: Master Expert Consensus Methods". SSDSI. [8]
  9. ↑ "Delphi - MSP Guide". mspguide.org. [9]
  10. ↑ Avella, J. R. (2016). "Delphi Panels: Research Design, Procedures, Advantages, and Challenges". International Journal of Doctoral Studies, 11, 305-321. [10]
  11. ↑ "The Policy Delphi". ResearchGate. [11]
Retrieved from "https://systems-analysis.ru/eng/index.php?title=Delphi_method&oldid=2050"