CHI 97 Electronic Publications: Late-Breaking/Interactive Posters
Mind Maps and Causal Models: Using Graphical Representations of Field Research Data
David R. Millen, AT&T Labs
Labs (now known as Anika Savage)
Diane Z. Lehder AT&T Labs
Dray Dray & Associates
We recently completed a series of field visits to understand how workers use the Internet in their daily work activities. At each site, the team used traditional field research methods such as work observations, artifact walk-throughs, and contextual inquiry. An innovative debrief process was developed to understand, summarize and document each visit. In addition to a structured debrief questionnaire, the team created graphical summary notes using "mind maps." These mind maps efficiently captured a nonlinear, graphical clustering of key ideas. A "causal loop diagram" was also developed to document the team's understanding of the internal and external driving forces for each organization. Taken together, the debrief questionnaire, the mind maps, and the causal loop diagrams provided a rich multimedia representation of the field data.
research methods, ethnography, qualitative data analysis
© 1997 Copyright on this material is held by the authors.
In the spring of 1996, a project named "Thinking Spaces" was initiated at AT&T Labs to study how technology, particularly the use of the Internet, was changing the way people work and the way they would conduct their business in the future. The goal was to develop an understanding of real customers solving real problems - around the world. An ethnographic approach was selected to discover and understand the nature of these changes.
The research plan included visits to 31 organizations during a three month period. At each location, the survey team interviewed a principal manager, toured the facility and interviewed and observed various workers. The data
collected for each location included interview notes, work process flowcharts, floor plans, and photographs of the work environment and work processes. Paper artifacts, such as forms, brochures, and other company publications, were also collected and electronic artifacts, such as web documents, were indexed.
Given the number of visits, the short duration of the project, and the large volume of field data, a significant problem the team faced was to summarize the findings from each site in a timely fashion. The team needed a framework to easily document, share and archive a staggering volume of research notes for future analysis.
After each visit, a debrief session was held in which field observations were summarized. In addition to a structured debrief questionnaire, the research team prepared a mind map and a causal loop diagram. The intent was to use more than one analytical approach and representation scheme in order to better and more completely record the findings from each visit. Our use of mind mapping and causal modeling in the field is discussed here.
Mind maps are nonlinear graphical representations of information. The research team used mind maps to record key ideas about three research interests: the physical work environment, the tools used by the workers, and communication patterns using email, fax, pagers, telephone, etc. The mind mapping process allows for ideas to be generated in a loosely structured brainstorming session. As the ideas are generated they are informally categorized by their placement on one of the major idea branches. Simple illustrations were occasionally added for emphasis or to clarify an idea. At the end of the brainstorming session, the mind map was reviewed for completeness and color highlights were used to visually accentuate important ideas.
The development of mind maps is a relatively fast and efficient way to record important ideas about a field observation. The non-linear nature of the record prevents a formal and rigid analysis that may result in blind spots. Important patterns in the field data emerge as related ideas are grouped together. An example of a mind map from one of the field visits is presented in Figure 1.
Causal loop diagrams
Causal loop diagrams have been used extensively to analyze qualitative data (see, for example, ). In traditional causal modeling, a network of variables is developed and the causal relationships between variables are explicitly delineated. The model is typically developed after all of the field data have been collected and some cross-observation meta-analysis has been completed.
In recent years, there has been an enthusiastic use of casual modeling to understand organizations and businesses . In our research, we adopted this perspective, and decided to develop causal loop diagrams during the data collection phase of the research. At each debrief session, we generated a causal loop diagram to document our understanding of the internal and external driving forces that were important to the organization. An example of such a causal model can be found in Figure 2.
The causal models that were developed for each site were preliminary. For each site we captured important environmental variables such as political, social, economic and technology forces. We also recorded our understanding of internal variables such as organizational objectives, financial goals and human resource attitudes and skills. We often color-coded the model variables to help visualize the relationships between groups of people within the "system" (e.g., customers of the business, suppliers, or the customer's customer).
each of the causal models was
extremely difficult, we quickly saw the benefits of the process. As we
developed the models, we began to identify gaps in our learning. In
we realized that our understanding of the organization needed some
input. Conversely, there were several times when we were surprised at
of our understanding. As we developed a particular casual relationship,
could easily draw upon concrete examples from our visit to fill out the
Indeed, developing the causal model forced us to document some
understanding that might otherwise have been forgotten. And finally,
individual site models were available to the team when we began the
cross-observation meta-analysis. The individual models served as input
composite, more general models that we developed.
The field research for this project included site visits to 31 organizations around the world. The volume of data was every bit as large as we expected. At the beginning of the project, we believed that multiple representations of each field visit would help capture the richness of each site. We feel that we succeeded in that respect. The individual causal models have been helpful in constructing four more general models. While the mind maps have been less useful so far in our meta-analysis, they remain valuable representations of three particular aspects of each visit. Taken together, the debrief questionnaire, the mind maps, and the causal models provide a broad foundation for our continuing analysis.
 Miles, M. & Huberman, A. Qualitative Data Analysis. Qualitative Data Analysis. California: Sage Publications. 1994.
 Senge, P., Kleiner, A., Roberts, C., Ross, R., & Smith, B. The Fifth Discipline Fieldbook. New York: Doubleday, 1994.
CHI 97 Electronic Publications: Late-Breaking/Interactive Posters