The “World Café” as a Participatory Method for Collecting Qualitative Data

Abstract

We introduce and discuss “World Café” (WC), a participatory assessment tool widely used in community development and organizational change processes, as additional qualitative research method. We propose WC as a participatory method of data collection for a large group of participants, discussing its strengths and weaknesses in comparison to semistructured interviews and focus groups, two well-established methods in qualitative research. As a research method, we find that WC complements other methods in important ways. When there are many participants, it helps guide the exploration and verification of themes. Integrating the method into the research design may help increase both the reference sample and the level of participation. Furthermore, as a participatory method, it not only produces data for the researchers but also has the potential to benefit the participants, as it facilitates dialogue and mutual learning, thus motivating their participation and responses.

Introduction

The World Café (WC) method is a tool widely used as a participatory method for citizen participation and organizational change processes. It has also gained some traction as methodological approach to collect qualitative data. Following up on this idea, in this article, we analyze its potential as a research method. We compare it not only to focus groups but also to individual interviews; two well-established qualitative research methods broadly used in the social sciences. As a participatory method, it not only has the potential to benefit participants, as it facilitates dialogue and mutual learning, but also produces valuable data for researchers to address their scientific questions.

Designed to be as inclusionary as possible, WC facilitates an open, yet intimate, discussion that accesses the views and knowledge present within a larger group of people. First introduced in 1995 by J. Brown and D. Isaacs, it is inherently designed to be a participatory tool that facilitates community change by hearing the ideas and opinions of as many community members as possible. Its name draws upon the café ambience that is created in order to facilitate informal conversation, representing a neutral public space where people feel free to engage with each other (Brown & Isaacs, 2005).

Initial studies examining WC as a method focus on comparing it with other research approaches, in particular appreciated inquiry and participatory action research, thereby treating it more like a methodological approach than like a research method (Fouché & Light, 2010). However, WC is increasingly being used as a research method, resulting in publications analyzing its strengths and weaknesses as a method as well as some comparisons to other research methods, in particular focus groups. Geographically, there appears to be a concentration of application and publications in Canada and the United States (for a collection of publications using WC and supporting materials, see http://www.theworldcafe.com).

It is within this notion that this article reports on the experience of using the WC as a research method in a German–Tanzanian project, the Trans-SEC food security project. The method’s potential is compared with focus group discussions (FGDs) and individual interviews, two other well-established methods. We are interested in using the comparison to provide information that will help researchers trying to decide which method might be best for them in any given project.

Our study posits the WC as a method that is well suited to obtain qualitative data from a large group of participants. It complements other research methods well, in that it helps to guide the exploration and verification of themes. Integrating the method into research designs may help to increase both the size of the reference sample and the range and scope of individual views on a certain topic. This article also provides recommendations on how to improve the use of WC as a method to increase control over the data collection process and to enable a greater differentiation of group results. The aim of this article is also to provide researchers with practical advice and recommendations on how to use the WC as a method in their research.

The WC

The WC (Brown & Isaacs, 2005) is said to enable the exploration and discussion of topics within large and heterogeneous groups. The WC focuses on intimate exchange, disciplined inquiry, cross-pollination of ideas, and possibility thinking. It is a conversational process that helps groups to engage in constructive dialogue around critical questions, to build personal relationships, and to foster mutual learning (Fouché & Light, 2010). The originators of the WC format describe seven integrated design principles (used as headings below) that also guided this article’s WC: (1) set the context; (2) create hospitable space; (3) explore questions that matter; (4) encourage everyone’s contribution; (5) cross-pollinate and connect diverse perspectives; (6) listen together for patterns, insights, and deeper questions; and (7) harvest and share collective discoveries (Brown & Isaacs, 2005, p. 40).

Since its inception in 1995, WC has been used in a wide variety of settings. For example, it is used by community-based organizations, small nonprofits, educational institutions, large multinational corporations, and government offices (Aldred, 2011; Brown & Isaacs, 2005; Fouché & Light, 2010) in fields as diverse as community development (Aldred, 2011; Brown & Isaacs, 2005), organizational development and strategic planning (Chang & Chen, 2015; Jorgenson & Steier, 2013), the medical field (Broom et al., 2013; du Plessis et al., 2013; MacFarlane et al., 2017; Roos & du Toit, 2014), the educational sector (Fullarton & Palermo, 2008; Partridge, n.d.), and in consumer studies (Ritch & Brennan, 2010).

To achieve the goal of a structured but conversational process, artificial cafés are set up in which participants can sit, meet, and talk. Enclosed spaces are created temporarily, possibly with drinks and music available to facilitate the café atmosphere. Participants are seated at café-style tables with a total of four or five participants. At least three consecutive rounds of conversation, approximately 20 min each, take place. At each table, the same question is discussed by the participants. After each round, the participants move from their table to a new one, where they continue their discussion with a new group of people. At each table, one host is selected by the group to remain at the table in order to convey the preceding discussion to subsequent discussants. The same question can be used for one or more rounds of conversation. It is also possible to pose different questions each round to build on and help deepen the exploration. Throughout each discussion round, participants record results in the form of text, sketches, or symbols on the paper tablecloth (or note key ideas on large index cards or place mats at the center of the table). At the end, after at least three rounds of conversation, a period of sharing discoveries and insights in a whole group conversation should take place. This way, common patterns can be identified, collective knowledge grows, and possibilities for action emerge (Brown & Isaacs, 2005).

While the WC method is primarily used outside academia, academics are starting to incorporate it into their research (Broom et al., 2013; du Plessis et al., 2013; Ruppert-Winkel et al., 2014). Fouché and Light (2010) claim to have been the first to apply the WC as a research method, with others following, including Chang & Chen (2015), du Plessis et al. (2013), Roos & du Toit (2014), Broom et al. (2013), Stöckigt et al. (2013), and Takahashi et al. (2014). The strength of WC for data collection is its “cross-pollination of ideas” through evolving rounds of information exchange and the use of a café-style social context that facilitates the sharing of information in an equitable and nonthreatening manner. Bertotti et al. (2012) describe the WC as “a novel qualitative method that attempts to highlight and capture the lived and varied perceptions and experiences of people living within.” Aldred (2011, p. 5) describes the approach in methodological terms as “a self-facilitating focus group investigation.”

The WC and Participatory Research

Being able to include as many stakeholders as possible in the data collection process is increasingly acknowledged as an important way to ground results. Within scientific research, an ongoing paradigm shift is moving from top-down approaches and knowledge produced mainly for the scientific community to bottom-up participatory research approaches. To find sustainable solutions to global challenges, such as food security, it is well acknowledged that including as many different actors as possible in the research process is necessary for cocreating knowledge that benefits both science and society (Jahn et al., 2012; Lang et al., 2012).

This shift in the epistemology of the adopted scientific research practices links to the turn in qualitative research toward narrative approaches. Such approaches focus on the lives of individuals as told through their own stories. The emphasis in such approaches is on the story, typically both what and how it is narrated. In this notion, the WC is a particularly suitable data collection method. As a method relying on collaborative action, one of its strengths is its creation of an opportunity for a group of diverse participants on any given issue to be able to share their reflections on their lived experiences and to use these reflections to create a deeper dialogue on the topics of the WC (Carson, 2011; Prewitt, 2011). WC participants are not recipients of knowledge created elsewhere, rather they are societal stakeholders engaged in the knowledge production process who also benefit from it (Aldred, 2011; Fouché & Light, 2010; Weitzenegger, 2010). WC facilitates collective knowledge-sharing by creating dialogue between diverse groups (Bushe, 2013; Bushe & Marshak, 2014). This may also allow investigating how people construct new understandings socially and how they engage in collective approaches to meaning-making, topics which have received a stronger emphasis in recent discussion of qualitative inquiry (Flick et al., 2004). It encourages mutual and reflective learning, supporting the development of new webs of personal relationships. In particular, the development of new ideas is especially well supported by the WC methodology (Stöckigt et al., 2013).

In this notion, studies find that WC is a process that helps groups engage in constructive dialogue, contributing to cognitive reframing and individual sensemaking, allowing a different perception of knowledge generation and moving participants beyond information transfer to information exchange (Lee & Garvin, 2003; Stöckigt et al., 2013). The focus on collective discoveries enables the harnessing of resourcefulness and a sense of hope, while the drive toward collaborative learning allows for equitable and collective participation. However, while this highlights the advantages of WC in participatory research settings, the question of whether it is useful in more traditional research settings remains.

While the increasing use of WC as a research method is apparent, research investigating its appropriateness as a qualitative research method is still limited, including comparisons with other, more established qualitative methods of data collection. An exception is Stöckigt et al. (2013), who compare WC with focus groups. Both are described as discursive methods that facilitate “collaborative dialogues wherein knowledge is gathered and shared” (Stöckigt et al., 2013, p. 1). However, while focus groups are only suitable for small groups, WC is “a simple, effective, and flexible format for hosting large group dialogue” (Stöckigt et al., 2013, p. 2). Unlike WC, focus groups are led by a moderator, without breaks and without changes to the group setting. In focus groups, the moderator stimulates discussion among a small number of participants (Linhorst, 2002). WC enables groups of all sizes to participate in evolving rounds of dialogue with a few others while remaining part of a single, larger, connected conversation. This allows for small, intimate conversations to link and build on others: The WC is inherently designed to have people move between small groups, thus discovering new insights into issues. In this article, we expand on these findings, comparing the WC method not only to FGDs but also to individual semistructured interviews (ISIs).

Differences Between Different Data Collection Methods

This section briefly introduces individual interviews and focus groups, two well-established research methods, and relates them to each other. In doing so, we derive criteria and dimensions by which to compare research methods and to judge their different disadvantages and advantages.

ISIs

ISIs are a flexible tool for collecting data that is capable of identifying important, often hidden, facets of human and organizational behavior. Individual interviews offer the opportunity to establish interpersonal trust between the interviewer and the interviewee, thus creating an atmosphere of confidentiality and comfort that facilitates the reporting of socially undesirable attitudes, beliefs, and behaviors (Holbrook et al., 2003).

The method is well suited for the deep exploration of the perceptions and attitudes of respondents regarding complex, if not sensitive, topics (Bariball & While, 1994). Interviewees can freely express their thoughts using their own everyday language. Different interview styles exist that can impact the kind and depth of responses. Approaching the interview using a (semi)structured approach, the interviewer can elicit rich and detailed responses from interviewees by varying the style, pacing, and ordering of questions depending upon the situation and by probing for more information and clarification of answers. Furthermore, interviewers can ensure completeness and comparability of the data collected, ensuring that all questions are answered by each respondent (Bailey, 1982; Qu & Dumay, 2011). When a narrative approach is chosen, emphasis is put on a free and spontaneous expression of thoughts, emotions, and recollections, stimulated by an opening, “narrative-generating” question (e.g., Clandinin & Connelly, 2000). This approach seeks to trigger in interviewees those kind of considerations and memories that reflect personal meanings and relevance and that they would not express in response to direct questioning (Flick et al., 2004).

However, the fact that researchers only speak with individuals, quite likely not reaching all individuals in the population of interest, makes sample selection critical; otherwise, the sample may not be representative. No matter which sampling strategy is used, it will always result in a somewhat arbitrary, albeit purposeful, selection that may result in the material collected being biased. Interviewer bias may also play a role, as interviewees might be intimidated by the researcher’s presence or might apply socially desirable response behavior, the tendency to admit to socially desirable traits and behaviors while denying socially undesirable ones (Krumpal, 2013; Tourangeau & Yan, 2007). Response behavior is also influenced by power structures, economic positions, and culturally specific communication behaviors (Hofstede et al., 1990; Johnson et al., 2005; Tellis & Chandrasekaran, 2010).

FGDs

The method was developed by Merton and Fiske in the 1940s to study how different audiences perceive media and is generally used to collect data on a specific topic (Cohen & Crabtree, 2006). In FGDs, data are collected through a semistructured group interview process where several people are interviewed together. FGDs make use of interactional exchange in groups to observe data less accessible through individual interviews. Researchers take on the role of a moderator, trying to stimulate a flexible and exploratory discussion that triggers lively interactions between participants rather than just a dialogue between the interviewer and interviewees (Kitzinger, 1995; Linhorst, 2002). FGDs are especially valuable for the comparisons that participants draw between their experiences, through which one can assess the (non)existing consensus of a certain topic within a social group or network.

There are different kinds of focus groups. Morgan (2018) differentiates between high and low moderation focus groups. High moderation groups are more similar to group interviews, comprising serial interactions between the moderator and individual focus group participants. Low moderation groups come much closer to generating socially constructed collective meaning as individuals’ contributions merge through the shared conversations (which are then transcribed and used as data).

With FGDs, it is possible to collect a large amount of data on a specific topic in a relatively short amount of time. While the optimum number of participants per focus group may vary, 6–10 participants per group are recommended (Krueger & Casey, 2000; Rabiee, 2004). Each focus group usually lasts between 1 and 2 hr, depending on the complexity of the research topic, the number of questions, and the number of participants (Rabiee, 2004). Opinions differ on the composition of groups. While some advocate homogenous groups in order to facilitate trustful and open exchange for the research of sensitive topics, others argue for heterogeneous groups to foster more spontaneous and honest response behavior (Kitzinger, 1994; Rabiee, 2004). FGDs generate large amounts of data. One hour of an FGD can take up to 5 or 6 hr to fully transcribe, leading to transcripts of 30–40 pages. While the recorded spoken language derived from the interview is the main source for data analysis, it is also possible to integrate the nonverbal communication expressed by the FGD members in the data analysis (Rabiee, 2004). When integrating nonverbal communication into the data analysis process, the use of a reflective diary by the interviewer is advisable.

Comparison of ISIs and FGD

There are a variety of ways to compare these interviewing methods. One simple and straightforward criterion for comparison is the number of unique ideas “generated” by each of the methods. Another measure is the richness of the data generated, that is, the depth of the exploration of a topic that becomes possible. Two criteria that link to the style of a focus group or interview and also the types of questions asked.

For both criteria, the social nature of FGDs (Krueger, 1994) may be a critical advantage. While narrative interviews allow for a very deep and broad exploration of information, output remains individually constructed. By interacting with other group members, FGD participants engage in a social process where ideas are heard, developed, and challenged (Heary & Hennessy, 2012). This social dynamic can deliver far more thoughts and ideas than would a series of individual interviews (Lederman, 1990). Therefore, through the interaction between group members, the understanding of a research topic may be far greater using FGDs.

Another important issue when comparing the methods that is also linked to the interviewer’s level of involvement is the effect researchers and other participants can have on the expression of experiences and opinions. Some of the unique advantages claimed for ISIs over FGDs focus on the ability to gain more in-depth information from individuals as the greater control of the interviewer over the interview process (Morgan, 1997). Following this reasoning, ISIs allow the researcher to locate specific ideas with specific individuals more easily (Denscombe, 2003). However, as already stated, the one-on-one situation between an interviewer and an interviewee may result in a power imbalance, with interviewees trying to satisfy the interviewer by providing answers they assume are expected (interviewer effects).

In FGDs, the researchers typically function as a moderator. Different moderation styles exist (Morgan, 2018), with high moderation FGDs consisting of serial interactions between the moderator and focus group participants. This allows researchers to guide participants and to keep the discussion focused on the research topic. A lower moderation style, by researchers who intervene as little as possible, may foster the creation of an own group dynamic within the group. This may lead to more “natural” interactions between session members, potentially including joking, arguing, and teasing (Kitzinger, 1995), and the generation of socially constructive collective meaning. Researchers thus may observe “how accounts are articulated, censured, opposed, and changed through social interaction and how this relates to peer communication and group norms” (Kitzinger, 2005, p. 58). Consequently, FGDs may be more effective at capturing the knowledge, experiences, and attitudes of individuals than ISIs with their direct face-to-face interview setting (Kitzinger, 2005). They may also lead to smaller interviewer effects compared to individual interviews, as the interviewer is less directly involved in the data collection process (Roulston, 2010). Therefore, FGDs potentially lead to a deeper exploration of the world of the interviewees than other data collection methods. This also means that FGDs are not well suited to trace and explore attitudes at the individual level but rather to explore themes and topics by the group as a whole, even though it is important to see how people differ in their underlying views on an issue.

An often-cited advantage of ISIs is their provision of a private and confidential setting, facilitating a more open response behavior. Following this notion, some authors suggest that FGDs are not suitable for studying sensitive topics that people are reluctant to discuss in public, for example, drug consumption, because they have concerns regarding disclosure of information to their peers. However, others argue that there may be safety in numbers in bigger groups and that the peer context may actually facilitate self-disclosure (Kennedy et al., 2001; Kitzinger, 1994, 1995). Thus, while FGDs try to overcome the disadvantage of ISIs by putting people together and facilitating more dialogue between them as group and interviewer influence is likely reduced, the group situation may hinder the free expression of thoughts and opinions, as individual participants may dominate the discussion with their views, preventing other participants from speaking their mind freely.

Method

This section briefly outlines the experience of using WC as a data collection method before discussing it in relation to individual interviews and focus groups in the Results section.

Case Study

The Trans-SEC project, which serves as the case study here, aimed to improve the food situation for the most vulnerable, rural population of four villages in Tanzania (Graef et al., 2014). The project represents a common project structure in the field of sustainability sciences working on global challenges such as food security. Within sustainability sciences, it is well acknowledged that sustainable solutions can only be found if actors across countries and institutions collaborate. Hence, it is common to integrate project members from different countries, organizations, and disciplines, as well as stakeholders (Brandt et al., 2013; König et al., 2013; vom Brocke & Lippe, 2015). Such complex settings are inherently risky as destructive conflicts between project members or between project members and stakeholders can arise (Löhr et al., 2018; Stokols et al., 2008; Zscheischler et al., 2014). To prevent conflict and facilitate its management, Trans-SEC management integrated a conflict management system.

ISI Data Collection

For the design and implementation of the system, a participatory research design was applied. Narratives of project members were collected and exchanged in multiple steps using different methods and formats over the course of more than 1 year (Löhr et al., 2017). A key activity was the collection of qualitative data on organizational conflict experiences and system expectations of project members at the project’s kickoff meeting. For this purpose, individual interviews, a FGD, and a WC were conducted. This article’s comparison of research methods is based on this research activity. In addition, measures such as team-building workshops, workshops on intercultural sensitization and communication, and strategic planning sessions with selected internal conflict contact points served as further platforms for personal reflective dialogue, mutual learning, and joined system development.

For this research, we analyze data from the conduct of (1) one FGD, (2) 13 individual interviews, and (3) one WC session. These qualitative methods were used to analyze project structures and needs in terms of conflict prevention and management. Data collection took place during the Trans-SEC project kickoff conference in September 2013 in Morogoro, Tanzania. Following an introductory workshop, which introduced the idea of integrating conflict management measures into the project, and a team-building session with the project consortium, data collection began. Ethical clearance for the research was provided by the German Ministry of Education and Research, which funded the project.

FGD was chosen as a method to facilitate dialogue among project members and to ensure representation of each institute working in the project. Each institute was asked to send one representative (16 participants in total), with a two-person team (one German, one Tanzanian) moderating the session. The FGD took place in a seminar room of the conference hotel, with chairs put in a circle. A question guide was designed and followed by the moderators in a semistructured approach to allow for group dialogue and narratives but to also ensure that research objectives on recommendations for conflict management were met. Questions focused on (1) conflict experiences in prior research projects, (2) existing strategies to overcome challenges, and (3) needs and recommendations for a conflict management system. The FGD took about 90 min.

ISIs: ISIs were selected to complement the FGD, comprising 13 ISIs. Members were selected to represent different nationalities, genders, and institutes. Interviews took place at the conference hotel, in its garden, the dining hall, or empty seminar rooms at the availability of the interview partners. The interview guide included questions on (1) (potential) conflict drivers in a research project, (2) prior experience with conflict prevention and management, (3) recommendations for a conflict management system, and (4) implementation obstacles. The interviews were conducted by one person, with each interview lasting between 12 and 54 min.

WC: The method was chosen to enable all 45 project members present at the kickoff conference to provide their input on the research topic (whole-in-the-system-approach). A seminar room in the conference hotel was prepared to create a café-like, hospitable space. Nine tables of five seats each were set up. On each table, writing material and markers as well as some snacks (nuts) were provided. Posters on the “Café Etiquette,” “Café Assumptions,” “Change of tables,” and “the four discussion questions” were put up and introduced to the participants in the beginning. Tea was served during the session by the co-moderators who also helped to change the paper formats at the completion of each round. All participants met in front of the room’s door where they were greeted by the moderating team and invited to enter the Café and to find a seat at any of the tables provided. Once all participants were seated, the method with its underlying principles and assumptions was introduced, and also the procedure of change of tables got explained.

The four questions discussed were similar to those used in the FGD and the ISIs. The first three questions were: (1) What kind of challenges could occur in the project? (2) What kind of measures and tool do we need to address the challenges? and (3) Which guidelines do you regard as important for a code of conduct in the project? These were discussed in three discussion rounds each, while the fourth and last question of “Which vision do you have for the overall outcomes of the project?” was only discussed in one round. The WC was moderated by one person, with two additional researchers providing assistance. The session lasted 90 min. No audio recording was used due to the sensitive nature of the topic and the fear that such a device would negatively affect the livelihood of discussion.

For documentation purposes and to allow for sharing of results, different paper formats were chosen. For the first question, paper tableclothes were provided at each table to note points of discussion by the table host and participants. In subsequent rounds, papers cut into different forms related to the participant’s working context of food security were distributed for note-taking (e.g., leaves, blossoms, beans; see Figure 1). After each round, the table hosts provided small-group table reports to the consecutive table participants. Plenary small-group report-outs did not happen due to time limitations.

Figure 1.

World Café result: The Trans-SEC tree; Trans-SEC consortium, Morogoro, Tanzania; September 2013.

A plenary harvest round took place after all question rounds were completed. Responses were clustered and arranged in the form of a big tree, symbolizing the project (Figure 1): blossoms and leaves (containing suggestions on the code of conduct incl. tools and measures to foster good collaboration), bugs as threat (challenges) to the project, and beans (symbolizing the vision and what project members aim to harvest). The roots and trunk as fundament of the tree were filled by results from a previous day’s team-building exercise in which people discussed joined strengths and weaknesses and points of joined interest. Results were summarized in plenary and a brief discussion took place.

Data Produced by the Methods

The FGD and individual interviews were audio recorded and later fully transcribed. Individual interviews resulted in a total of 333 min of audio file material plus an additional 90-min audio recording of the FGD. The transcription process was time-consuming and resulted in the decision to outsource transcription activities to a professional transcription service. However, this was also problematic as the external staff was unfamiliar with the project terminology and its abbreviations; further challenges came from the wide variety of spoken English, which incorporated dialects that the service was not familiar with. Hence, externally produced transcripts required time-intensive revisions by the interviewers. Such aspects are also important to consider when choosing research methods.

The WC session resulted in a variety of handwritten materials. For each of the four questions, a new set of writing material was distributed. For the first question, large paper sheets were put on the table as tablecloths, resulting in a total of nine different paper sheets handwritten by the participants (the host of each table). In subsequent rounds, different paper formats cut in forms that fit the respective research environment were used. Figures 1 – 3 show the different formats of WC output. The figures are inserted here to exemplify the usability of data output from the WC method.

Figure 2.

World Café result sheet example 1.

Figure 3.

World Café result sheet example 2.

Overall, the WC method generated a wide range of ideas that were collected and typed up for analysis. Data comparison is based on the WC data of the first round (Appendix), while method comparison looks at the full WC session. All data were analyzed using qualitative content analysis, a method to identify manifest and latent structures in texts and other qualitative material (Kohlbacher, 2006; Mayring, 2000, 2014). Coding was done using f4analyse coding software (Version Beta 4.0). In the first round, broad categories (codes) aligned with the research questions were applied for data analysis (e.g., project challenges, existing conflict management practices) and, in a second step, categories were refined, resulting in further categories and subcategories (e.g., challenges between project members, challenges with stakeholders, challenges of finances).

Data analysis was done consecutively, that is, method by method: First, codes were derived from individual interviews as they provided most data (defined by length of interviews in total, length of transcripts, and amount and depth of information). Second, the FGD was analyzed. Data could be integrated in the coding system of the individual interviews; no additional code emerged from the FGD. In a third step, the data from the WC were analyzed, also using the same coding system.

Results

In brief, this study finds WC is a method that is well suited to complement other methods in order to either help explore a research topic or verify findings. Integrating the method in the research design helped to increase the reference sample and to increase participation levels. Table 1 summarizes this article’s findings (adapted from Straus, 2010). The table is structured in three different phases of data collection: (1) planning, (2) implementation, that is, conduct, and (3) results. Each phase lists multiple criteria, derived from the previous literature study, that need consideration in the research process. This should help researchers who are choosing between qualitative research methods.

Table 1.

Strengths and Weaknesses of ISI, FGD, and WC Methods.

Phases of Research	Criteria	ISI	FGD	WC
Planning	Researching individuals	++	+	−
	Researching groups	−	+	++
	Number of participants (per session)	1	4–12	>12
	Duration (per session in minutes)	15–60	60–120	>90
	Number of research questions (per hour)	6–15	4–6	1–3
Implementation	Exploration of new topics and concepts	+	++	++
	Obtaining an overview	+	++	++
	Depth of exploration	++	+	O
	Researching delicate/sensitive topics	++	O	−
	Level of researcher–participant interaction	++	+	O
	Interviewer influence	++	+	−
	Spontaneous adaption of research question	++	++	−
	Impact of power relations	O	+	O
	Interaction with third parties	−	++	++
	Degree of participation	−	+	++
	Input of researchers’ time and effort	+	++	O
	Confidentiality	++	+	++
	Degree of reality of communicative setting	O	+	++
Results	Analysis of interaction patterns	−	++	++
	Data usability	++	+	O
	Differentiation of individual responses	++	+	−

Note. Own conclusions based on experience from our case study. Signs express the strengths and weaknesses of each method for the different criteria of comparison ranging from great strength (++) to weakness (−). ISI = individual semistructured interviews; FGD = focus group discussions; WC = World Café.

Planning

In the case study, the WC enabled all project participants (45) present at the kickoff meeting to participate in the data collection activity, thus providing input on the research topic. Thus, looking at participation levels from a purely numeric perspective, WC allows for the maximum degree of participation. Four research questions were discussed in the session, with participants changing tables for a cross-pollination of ideas on the first three questions. Due to the method’s procedure, not only was the order of research questions predefined, but fewer questions were asked than during the FGD and ISIs. With respect to ISIs, while individual interviews took less time for the participants, the interviewer had to invest more time in this data collection activity. As the WC session was an integral part of the kickoff conference program, no parallel meetings were scheduled and no extra time was requested of the participants. This proved an advantage over ISIs and the FGD for which it was challenging to allocate slots during the conference as the agenda was already tight. Interviews and the FGD were scheduled in parallel to work sessions, during breaks, or in the evenings; times often characterized by fatigue or reduced motivation to participate.

Implementation

WC provided insight on the group’s opinion(s) on the topic of study. Multiple themes (ideas) were established and discussed by the participants during the session (see Appendix). During the discussion rounds, no direct interaction between the interviewer and the participants existed. Participants discussed among themselves; the interviewer only facilitated the process. Throughout the discussion, notes were taken by table hosts on paper sheets. In contrast, during the interviews and focus group, direct interaction between the interviewer and the interviewee existed. A set of question was prepared, but the order varied depending upon how the interview progressed. It was possible to explore questions in more depth and for the interviewer to ask further questions seeking clarification or to further explore the issues raised. Respondents were able to provide examples or highlight issues that were important to them outside the question guide. During the focus group, different people responded to various questions, and dialogue between the participants took place. An unbalanced contribution by participants was observed, with some being more active than others.

Results

Within a relatively short amount of time, much data from a relatively large number of people were obtained using the WC method. The Appendix shows the multitude of answers (ideas) that resulted from the discussion on the first question on conflict potential. Results are grouped in categories that were developed as part of the data analysis process. Answers (ideas) tend to be broad categories, rather than in-depth exploration of individual ideas. The data depth was limited under the WC method, particularly when compared to individual interviews and also focus groups. This limitation is mainly due to the limited number of questions that were discussed during the single session. We are unable to identify differences between subgroups or individuals in WC data, which are based on handwritten notes by the participants during the discussion process, as Figures 2 and 3 exemplify. Thus, data usability depends on the legibility of the notes and their comprehensiveness. There was no opportunity to ask clarifying questions, to obtain examples if not given, or to get any other additional information.

For ISI and FGD, the transcription required a very high time investment and very careful attention due to the different accents of participants coming from various countries. As noted earlier, outsourcing transcription proved challenging for a variety of reasons. In addition, anonymization of transcripts was necessary. Even so, most material cannot be made public in order to safeguard confidentiality. In contrast, except for illegible notes, WC results were easily typed up and completely anonymous, thus easily made accessible to third parties without risk of identification.

Discussion

Exploration of Research Questions and Collection of New Ideas

The WC process facilitated a broad exploration of the research topic, resulting in a large number of ideas. It is a good entry method in research, as it can bring out major themes and topics important for the group. For this study, the method served its objectives well, to allow for maximum level of participation by all project members, to initiate dialogue on conflicts and conflict management among project members, and to obtain recommendations for the design of a conflict management system. As the research was embedded in an annual project conference with a tight agenda, a given group of people, and also clearly goal-driven, that is, to obtain data for the design of a conflict management system, rather focused and specific questions were asked to get very specific answers in a relatively short amount of time. This shows the adaptability of WC to different contexts and research needs.

Despite the focused scope of our research questions, participants’ feedback suggests that the WC initiated dialogue beyond the session. Also, WC was successful in creating opportunity for building relationships and sharing stories as the method enabled communication on a more personal level beyond the usual work context. In addition, the WC was also only one of the first steps in a participatory design process for conflict management. Further narratives were obtained and dialogues facilitated over the course of 1 year (Löhr et al., 2017).

This shows the importance of always considering the context of data collection and the way each question is framed and worded, as this impacts the kind and depths of responses. Other settings could allow the opportunity to apply a more open and narrative-oriented approach. In such settings, an appreciative, more personal first round WC question might be more advisable, asking participants something along the lines of, “Please share a story of a conflict in the workplace that you were able to successfully resolve. Where were you in the organization, and what did the organization have in place to help you address the conflict?” Such a question could create an opportunity for participants to respond on a more personally reflective level, thus allowing for the potential capture of narrative stories that can emerge. Further, it is possible that such an introductory question deepens the way participants respond to subsequent questions.

Considering the sensitive nature of “conflict” as a research topic and the complex nature of a research project with project members originating from different institutions, disciplines, and countries, beginning with a question on personal conflict can be difficult. Also, in our case, such a question was not chosen to account for power relations between participants. Colleagues from same institutions of different status (e.g., professors and PhDs) participated in the WC and met at tables for joined discussion rounds. Considering the sensitive nature of conflict, such a question might have hampered narratives. Instead, questions on personal conflict experience and management mechanisms were asked in the accompanying individual interviews and also the focus group.

To verify group results, the different result sheets from various tables can be compared to assess convergence. Subsequently, these themes can inform further interview questions and FGDs, thus providing more in-depth insight on specific issues. Alternatively, and as done in this case study, the emerging ideas can be compared to those extrapolated from interview and focus group data analysis. In this way, WC data help to cross-check the validity of individual and group level. Convergence of results, as found in our data analysis, is supported by Stöckigt et al. (2013) who employed FGD and WC on research exploring the use of alternative medicine among German senior citizens. This implies that WC should be used as a complementary—not the only—research method, thus reinforcing Roos and du Toit (2014) who advocate using a variety of data gathering methods to account for limitations.

However, as in our case, in a WC setting it is not possible to differentiate responses between different (sub)groups of participants. Further, we were unable to establish if all participants contributed to discussion equally or if certain groups or individuals dominated the discussion, thus resulting in response bias. This experience coincides with Aldred (2011) who highlights that WC results could portray a feeling of consensus in the group, with potentially problematic implications.

Sampling and Representativeness

When aiming for the participation of as many people as possible, WC is well suited as a method. It allows for a large number of participants to participate, up to 1,000 people (Brown & Isaacs, 2005). This stands in contrast to individual interviews and FGDs, as these two methods require drawing on relatively small samples of participants. With its inclusivity and ability to integrate a large pool of people, WC may also reduce selection bias, that is, the tendency that only people with certain characteristics take part in the study while others are systematically left out (an issue underexplored and often ignored in qualitative research).

However, there is debate as to what extent the “whole-in-the-system-approach” (Brown & Isaacs, 2005) is advisable. Originally, the WC principle was aimed at broader inquiry in organizational or community settings and was not conceived to be a part of formal research, nor did it have to deal with ethical review boards. Considering power relations, as in our case, for example, between students and professors, scientists and nonscientists, and participants from the Global North and Global South, other research suggests that for data gathering it is better to work with peer groups and only apply the whole system approach in community consultation around determining research questions (Fallon & Connaughton, 2016; Sheridan et al., 2010).

Data Collection and Response Process

As direct interaction between the participants and the interviewer is rare when applying the WC method, it can be argued that it reduces risk of interviewer bias and socially desirable response behavior. Further, a high degree of confidentiality is possible, as individual responses cannot be traced. However, it is not possible to clarify responses or obtain additional information.

In contrast, individual interviews and focus groups are led by at least one interviewer. Thus, direct interaction between the interviewer and the interviewee exists. It is possible to explore topics in depth and for the interviewer to ask clarifying questions or further explore the issues raised. Respondents are able to provide examples or highlight issues that are important to them (MacFarlane et al., 2017).

Under the WC method, therefore, interviewers cannot control the response process (Prewitt, 2011; Weitzenegger, 2010). It is possible that participants put a different thematic emphasis in their discussion than intended by the interviewer. Therefore, it is important to choose the WC as a method carefully and to pose good questions, with Prewitt (2011) arguing that question selection is critical.

To some extent, there is convergence with focus groups, as both are discursive methods that facilitate dialogues and knowledge exchange between participants (Stöckigt et al., 2013). However, group size and process dynamics differ. Focus groups are only suitable for small groups, while WC allows for hosting a large group dialogue. Further, knowledge creation in WC depends completely on the participants, as the interviewer falls into the role of a process facilitator, with no direct interaction in the discussion process itself. While FGDs can also exhibit low levels of moderation and ISIs may be purely narrative, the interviewer still remains in direct interaction with the participants and thus has the option of taking back control if needed.

In contrast to WC, FGDs are led by an interviewer, without breaks and a consistent group composition. Unlike in FGDs where a moderator uses the group process with a small number of participants to stimulate discussion (Linhorst, 2002), WC enables groups of all sizes to participate in evolving rounds of dialogue with a few others while remaining part of a single, larger, connected conversation. This allows for small, intimate conversations to link with and build upon others as people to move between groups and discover new insights into issues (Fouché & Light, 2010).

To circumvent power inequalities, the WC method has participants change tables, thus meeting other participants with whom they can interact differently. By switching tables, group dynamics are affected, with new group formations occurring every round. This strategy can balance group imbalances and unequal contributions. Still, we do not know if participants dominated discussions or remained passive. Thus, depending on research interests, using a variety of data gathering methods will account for this limitation (Roos & du Toit, 2014).

Possible Adaptations to Fit a Qualitative Research Design

To address the shortcomings of the WC method in future and to take discussions to a deeper level, introducing a set of subquestions every round is one strategy. Further, researchers could be assigned as observers, seated at each discussion table to observe the process. Another possibility is to place recording devices at each table as done by other researchers (Carson, 2011; Takahashi et al., 2014; Wood, 2014); however, the impact of such measures needs testing across different research topics. In our case, no recording devices were placed at the tables as impact on the discussion was feared. Researching on work conflicts is a sensitive topic and even more so when participants are either colleagues or intend to collaborate for years to come; in such cases, recording devices might have negative impact on the sharing of narratives.

Alternatively, it is also possible to select table hosts in advance, from the group of participants, preparing them for their role with regard to moderation and note-taking. During the design process of a new neonatal intensive care unit in Australia, 10 people were asked to host during a WC. To assist these hosts with their task, they were emailed information about WC methodology in advance and personally briefed about the methodology and an outline of the workshop (Broom et al., 2013). In other cases, hosts were preselected in order to provide process guidance, moving the discussion in the right direction and helping attendees to stick to the original discussion topic (Chang & Chen, 2015; MacFarlane et al., 2017).

Alternatively, if the group size is large enough, more than one WC session could be conducted, thus providing the opportunity to compare group outcomes and increase variability. This could also be a strategy in order to account for power inequalities or other ethical issues. Further, for comparative studies, WCs could be repeated in one or across case study sites, with results correlated. For example, during the restructuring process of a hospital, WCs were facilitated at four different hospital campuses and other sites on the same day at different times to increase participation rates. In another case, two WC sessions were conducted at two different hospitals to engage faculty and direct care nurses in a dialogue on leadership experiences (Kempnich & Costanzo, 2014; MacFarlane et al, 2017).

If it is crucial for a research question to obtain information on subgroup opinions or knowledge, adapting the WC process is possible: To gain insight on (sub) groups, WC discussion could start with an internal group discussion in the first round, before starting to move tables and meeting participants who are members of other (sub) groups. However, this is difficult if, for example, some groups are underrepresented and, thus, unable to have individual tables in the first round. Further, it could strengthen group dynamics and power structures, thus negatively impacting the open sharing of opinions and creation of new knowledge.

Conclusion

This study analyzes the WC as a method for qualitative research. We introduce the method and analyze its potential and constraints in comparison to individual interviews and FGDs. We find that the WC method can considerably enrich the toolbox of qualitative researchers. It is well suited to collect the views and perceptions of a relatively large group of people over a relatively short period of time and is resource efficient. It helps in the exploration and verification of themes over a large number of participants. It also benefits the participants as it facilitates dialogue and mutual learning. In brief, this study finds that the differences in method application and data output make WC a method that is well suited to complement other methods in order to either help explore a research topic or verify findings. For future research, we suggest to run similar studies on the different methods applying varying moderation styles (low to high) and question formats (focused to narrative). Establishing a matrix with moderation styles, question format, and data output might be very interesting and helpful for methodological decision-making in future research.