One of the most common questions I get from students and early researchers is: should my study use a survey or interviews? Often the better answer is: why not both? A recent conversation with a startup working on emotion data gave me a perfect example of why.
The "what" and the "why"
At a recent networking event, I met the team behind HiStranger, a Korean deep tech startup building Emotion Intelligence Infrastructure for AI. Their product, Insight Flow, looks at how people respond emotionally to content, moment by moment.
Cammie Kim, COO of HiStranger, explained it simply:
"A survey tells you what people say after watching. Emotion data shows how they actually responded, moment by moment. Put the two together, and you get both the 'what' and the 'why.'"
— Cammie Kim, COO, HiStranger
In research methods, we have a name for this idea: mixed methods research. It means combining two kinds of data in one study, usually what people say (surveys, interviews) with what they do or experience (observation, behavioural or physiological data), so that each one fills the gaps of the other.
Why one method is often not enough
A survey is excellent for reaching many people and measuring patterns. But it depends on memory and self report. As Cammie puts it:
"Surveys capture what people remember. Emotion data captures what they felt, while they were feeling it."
— Cammie Kim, COO, HiStranger
People may not remember accurately, may give the answer they think is expected, or may simply not know why they felt something. Adding a second source of data helps you check, explain or deepen what the first source shows.
Three common mixed methods designs
- Convergent design: collect quantitative and qualitative data at the same time, then compare them. Example: run a survey and record viewers' responses during the same session.
- Explanatory sequential design: start with quantitative data, then use qualitative data to explain the results. Example: a survey shows low satisfaction, and follow up interviews find out why.
- Exploratory sequential design: start with qualitative data to explore a topic, then build a survey to test what you found on a larger group.
Tips for your own study
- Start with your research question. Use mixed methods only when one type of data cannot answer it fully.
- Plan how the two data sets will connect. Decide in advance whether you will compare them, use one to explain the other, or use one to build the other.
- Be realistic about time. Two data sets mean two rounds of collection and analysis.
- In your methodology chapter, clearly name your design and justify why you chose it.
Final thought
Good research is about understanding people, not just collecting numbers. As the HiStranger team says, "The future of AI is not just data. It is people." The same is true for research.
Thanks to Cammie Kim and the team at HiStranger for sharing their perspective. You can follow them on LinkedIn.
— Dr Afsana Salam