Showing Culture

MCL Backstory

By Dan Dohan

In seventh grade English, Ms. Wescott drilled into her Diamond Junior High School students the golden rule of effective writing: show don’t tell.

Showing culture is tricky. Culture includes values, beliefs, and practices shared by a group. It exists in informal communities, such as a neighborhood or ethnic group. It animates formal institutions, such as non-profit organizations or professions. Culture is playful, dynamic, and ever-changing, and even though everyone has a culture and every culture is shared, no two people use culture in the exact same way.

Showing culture requires creativity and interpretation. Standard tests and questions do a poor job. Culture shows up in qualitative interviews and focus groups and in observations, videos, and documents. Even there, it does not speak for itself. Showing culture is the special task of ethnographers. Effective ethnographers immerse themselves in field sites and forge relationships with informants so they can learn about values, beliefs, and practices. They craft narrative accounts, which typically fill 20-page journal articles or 300-page books, that illustrate their learning. Ms. Wescott would approve.

But scientists from other traditions are dubious. They see, measure, and show the world in a more structured and succinct fashion. They test pre-specified hypotheses using measures that are defined and validated in advance. They document each experimental procedure. They show their work in 5-page journal articles in IMRAD format: Introduction, Methods, Results, and Discussion .

Over the decades, IMRAD became standard in quantitative science publication, but it is ill-suited for showing culture. A tightly structured methods section does not allow an ethnographer to show how they immersed themselves in a field site and what they learned from informants. The results section is typically too short to interpret values, beliefs, and practices.

Wresting with IMRAD, many qualitative researchers present culture as a static list of themes, each illustrated by an interview quote. Cultural analysts hate this stifling approach. Even worse, quantitative scientists find it unconvincing. They can’t see how the ethnographer did their work. They wonder if the quotes are evidence or anecdote. IMRAD fails to show culture to scientists. But adapting scientific convention to cultural analysis makes sense. Ethnographers just need to cast a broader net to consider what types of scientific convention can help advance their work.

In “Beyond Text ,” we tried to imagine how scholars could show culture in some way other than through narrative. What if readers could visualize culture more immediately and directly. Imagine if we could take Ms. Wescott’s advice a step further by showing how culture shifts and varies among groups and institutions.

Quantitative approaches for visualizing rich, complex data provided a starting point for our cultural imagination. Biomedical and health scientists rely on creative data visualization. Their elegant figures summarize complex data accurately and beautifully. In “Beyond Text”, we outline how ethnographers might show culture by visualizing interpretative data.

We borrowed a technique from the biological sciences that uses a colored grid to show sequences of data. Each column of the grid includes data from a single person; each row shows a particular type of data. These data arrays depict astounding amounts of data. In genetics, for example, a single array may display thousands of genes from hundreds of people. Each tiny colored square, or cell, represents a single gene in a single person. Arrange thousands of colored cells and patterns emerge. Like a pointillist painting, the array makes no sense when viewed up close. But step back and view it holistically and the colored cells arrange themselves into a picture that tells a story.

The array could be an attractive visual approach to show culture. Each cell can depict an individual’s orientation towards a particular element of culture. In one ethnographic study , we used an array to depict cultural practices that shaped treatment decision-making among patients with advanced melanoma. For each patient, we included cells that indicated if they trusted their oncologists, read the research literature, or sought a second opinion. Other cells in the array captured values and beliefs, e.g. whether patients were hopeful, altruistic, or valued length versus quality of life.

Viewed as an holistic array, patients’ individual orientations cohered into recognizable patterns. One group of patients, reliant outsiders, left their care to the experts. They were happy to remain out of the loop in terms of the details of their condition and rely on their care team to guide treatment. Active insiders, in contrast, immersed themselves in the details of melanoma and its care; they sought to understand and act, themselves, rather than rely on experts for decision-making.

The story of the array was echoed in patients’ own words. Our narrative analysis of what we observed in the clinic and what we heard in in-depth interviews matched the patterns in the array. The array did not replace rich patient narratives. But it captured them better than an IMRAD results table of themes and quotes.

Visual displays make ideas accessible. Scientists use visuals to transform complex data into straightforward narratives. Ethnographers have always crafted narratives about culture. As ethnographers seek to connect with broader audiences – and as scientists, policymakers, and clinicians gain a deeper appreciation for the significance of culture – it is worth considering how qualitative data could be shared in more accessible visual ways.

“Beyond Text” suggests how to place ethnographic data in an array and why that might be helpful. The “ethnoarray” joins other visual innovations that show culture and enhance narrative. Developed fully, carefully, and responsibly, the ethnoarray may prove a capable tool to tell stories about culture with richness and subtlety. Scientific (and policymaker and clinician) audiences need to better understand how culture works. Tables of themes and quotes are inadequate substitutes for rich narrative.

As we strive to develop new tools to show culture, we recall that ethnography’s track record of analytical innovation can be discouraging. Quantifying culture has been done with malicious intentions — to prove one group superior to another, to justify oppression, and to excuse inequity. Even more troubling: well-meaning analysts have provoked unintended harm. Narrative meant to illustrate how community beliefs and practices are shaped by structural inequity have been used to “prove” that culture condemns groups to inter-generational misery[DD6] . Many ethnographers worry that quantifying culture opens the door to harmful misinterpretation. We worry if visualizing culture for health scientists and policymakers could open the door wider. We are proceeding with our work to show culture with caution. At present, our work focuses on showing how ethnographers analyze qualitative fieldnotes and interviews. 

For more than a century, ethnographers have analyzed their data by sorting and summarizing it. After completing fieldwork or interviews, analysts read and re-read all the data as they develop a set of codes that will be helpful in sorting it out. They then manually review the data, line by line, to apply each relevant code to each relevant chunk of data.

For anyone who does not personally participate in the activity, the coding process is a black box. Frankly, many qualitative analysts feel the process is a black box. Fieldnotes and transcripts go in; codes, themes, and illustrative quotes come out. The process is interpretative and mysterious. Many scientists don’t hesitate to call it suspect. Over the years many efforts have been made to explain and routinize qualitative analysis. Scholars in nursing and other health fields have embraced the step-by-step explanations of grounded theory analysis. In the medical sciences, COREQ  and SRQR checklists have attempted to routinize qualitative research, including procedures to ensure rigorous coding.

When we made the ethnoarray, we used standard approaches to code data, identify themes, and select quotes. This analysis guided how we decided the color — the disposition — for each array cell. The ethnoarray made the results of our analysis more accessible, but our analytical process remained opaque. Currently, we are working to make that process less opaque, more rigorous, and more reproducible. I feel ethnographic research should be held to a reasonable standard of rigor and reproducibility. I do not feel standardizing analysis is the right path to that goal.

To achieve rigor, ethnographers need not and should not use pre-set procedures. Culture’s fluidity, dynamism, and multi-vocality requires bespoke and flexible interpretative analysis. Rigorous coding occurs when analysts immerse themselves in data and lay out a codebook that reflects their interpretations. No grounded theory procedure nor COREQ/SRQR checklist can adequately capture the intellectual rigor of that creative process.

Greater reproducibility in qualitative research requires greater transparency. Qualitative researchers too often claim their research started with expectation-free exploration. Scientists greet such claims with skepticism. Ethnographers can do a better job describing the research puzzle that inspired their investigation. They can better link that puzzle to fieldsites and respondents. They can be more explicit about what they expected to find. Coding schemes always emerge from the kernel of the study question. Transparent research designs advance reproducibility.

Finally, our team is also working to more transparently show how codes are applied to qualitative data. Starting from a well-documented codebook, we have found that machine learning algorithms do an excellent job applying codes to data. These algorithms do exceptionally well when supplemented with selective manual (human) review. The algorithms are technically complex; many ethnographers may find they make analysis less transparent rather than more so. But machine learning is ubiquitous in data science and we feel it is worth exploring. Its use can enhance the transparency of qualitative analysis for scientific audiences.

Our work developing machine learning for qualitative data analysis is in late infancy. There remains much to do to rise to Ms. Wescott’s challenge to show, not tell, how culture shapes our lives.