Language is Context

By Arundati Dandapani

Remember the last time you were made fun of for being too sensitive to errors in language, grammar or the imagination? This goes much beyond but does not exclude the times when people misquote your title, mispronounce your name or distort some demographic or other fact about you. Not to say that anyone is mistake-proof in those realms, but I have learned across sectors and especially through work in survey research, that having a strong ear for language elicits better data quality and stronger accurate insights. This becomes even more apparent in traditionally taboo categories like “cannabis”. However, problematic questions aren’t always problematic only because of inaccuracies, or because there is an element of social acceptability/desirability involved in respondents holding/expressing certain views/non-views. Think of the plight of respondents who are answering surveys that talk down to them or dealing with interviews that just don’t resonate? How often have you walked in their shoes? All the time?

What are some language considerations to make in survey research in single population or multiple population studies? What about when research is conducted in hyper-diverse markets and in markets where the mix is constantly shifting?

When Mandy Sha first treated me to a gelato in Chicago post an insights conference to discuss including my chapter “Nura Knows You Better: Redesigning Conversations with Artificial Intelligence” in The Essential Role of Language in Survey Research along with Tim Gabel of RTI International, I was elated and humbled by the company I kept among all the acclaimed linguists and researchers who co-authored this textbook.

A COVID-19 book release is every self-isolating author’s dream come true! Thank you Mandy Sha and Tim Gabel for this ambitious undertaking.

Published by RTI International Press and edited by Mandy Sha and Tim Gabel The Essential Role of Language in Survey Research really expands on the opportunities and challenges in designing research for multicultural, multiregional, and multilingual (3MC) comparative surveys that are designed to collect data and compare findings from two or more populations. Improved translation and pretesting procedures allow for a wholistic and accurate measurement of audiences in specific/different markets.

All the coauthors’ names are etched in grey on the cover with elan

There are some excellent pointers for designing such research. Translation models like TRAPD offer best practices in survey translation. Each chapter dives into how language creates context for different groups and populations. Some examples of the challenges of cultural variability arising from different backgrounds and experiences of survey respondents (specifically in the context of the book) like sociolinguistic and regional differences make comparing survey data incongruent in scale, tonality or equivalence. Words like ‘Civil Disobedience’ or ‘National Pride’ or ‘Social Security’ are perceived differently across societies, making equivalencies across populations hard to measure. The absence of equivalent concepts across languages is another important consideration (“prostate” has no translation in the Hmong language, for example) and the new meanings of a word upon its translation to another language also changes its context in research. What is the “right number” of translators to have on a study, given there is so much reading of texture and context involved among cross-cultural respondents ?

“H2: Among respondents from recent immigrant groups (e.g., Mexican and Korean Americans), those who are more acculturated to American culture will be more likely to request clarification when confronted with problematic survey questions.”

Chapter 2: Seeking Clarifications for Problematic Questions

Three themes struck me as clear from all the chapters.

Mixed methods are powering the future of research innovation: Mixed methods are powering the future of research innovation: Methods like vignettes (hypothetical scenarios) combined with cognitive interviews, online probing and behaviour coding are just few examples of how mixed methods allow researchers to pretest when measuring public opinion or data across cultures. Beholden to budgets and scope, there are limitations with research design in such studies that can or must be prefaced in each study.

Emotional measurement has never been harder: Language is complex and honesty and specificity of detail are always bad expectations to have of respondents without being well-versed with their cultural contexts. Equivalence in language translation extends to interpretive equivalence (or the subjective translation) and procedural equivalence (or the objective and comparable measures of survey translation). However true equivalence, can never be fully achieved, and remains what some might describe as an ongoing negotiation (or even battleground) for those who navigate more than one language or culture.

Technology is driving the next frontier of cultural understanding: New methods are constantly challenging the status quo and how we conduct research. As regions, markets, societies and cultures adopt technology with varied degrees of trust and fluency, the use of AI-powered research, text-mining technologies and online qualitative research methods continue to power the next realm of cultural awareness across diverse markets in the US but also globally and across demographics breaking new ground.

Hear more about the book from its editors in this podcast where Tim also gives a shoutout to (my chapter on) chatbots among the several innovations in this book.

Leave a Reply