I was filling out a procurement document for one of our financial clients the other day and the question was asked about whether we use synthetic data within our surveys. The simple answer is that we do not, but it sparked an interesting conversation in the office about the merits of synthetic data and AI in research. I think whatever your feelings about AI, it will become unavoidable, but the consensus was that speaking to actual people is important, particularly because they can come up with responses which may appear contradictory at first until the whole story is fully explored.
I had this presented to me recently when I had a rare day of face-to-face fieldwork for another of our clients. As a quantitative researcher, I hadn’t done any F2F work since the covid pandemic, so it made a nice change to chat to people, in this instance about their experience of using a bank branch. On the face of it, if comparing findings from a couple of key questions in the survey, you would have said that they just didn’t make sense, but having had the full immersive experience, it was totally clear to me why those discrepancies of logic had emerged.
We strive to ensure that the data from our projects is the best quality, regardless of methodology used. We only work with a small pool of trusted suppliers so for my area of quantitative research, we know that telephone surveys will be done sensitively with the interviewer making sure that participants fully understand all questions and ensuring that open-ended questions are fully explored. With the prevalence of bots filling out online surveys, we place numerous checks in place and carefully scrutinize all responses to make sure that we are happy with the data set before we start our analysis. We value our relationship with our clients so you can be confident that if you work with us, regardless of fieldwork methodology, figures and findings will be of the highest quality and any human contradictions explained in the analysis.
As an aside, I asked an AI generator to write an article on this theme to see if I would get any differences. It did a reasonable job although extolled the virtues of synthetic data far more than I have, so I present the human point of view in keeping with the rest of this piece.
By Martin Brock
Like this article? Follow us on LinkedIn.
