As part of the pilot test for the project, which ultimately involved a survey administered to a rigorously stratified sample distributed across a major world city, the researcher conducted interviews with a number of city residents selected from the same sample. The idea was to refine the survey instrument to capture more subtle gradations in travel behavior.
As the researcher described it, the interview subjects seemed eager to tell their own stories of living in and moving through the city and, according to the researcher's account, some became quite animated in the telling.
But the researcher had taken a particular attitude toward the subjects, and the theory to which the researcher appealed made very specific, very stringent demands as to the kind of data that would be acceptable.
The model was a variant of welfare economics, a kind of cost-benefit analysis of travel modes in particular contexts; the researcher's innovation is a refinement on the benefit side, a different way of accounting for the various ways in which travel generates utility.
The model itself takes the form of a series of equations, into which raw stories cannot be fed. Somehow, those stories must be translated into numbers.
I've struggled with this process in my own research, in the past, particularly in an attempt to understand the likely response of Atlanta-area residents to the ongoing BeltLine project. There are various ways of coding responses, reading through the raw text of an interview transcript and identifying statements that can be captured or summed up or placed under the heading of a pre-selected label, from a finite set of such labels that are convenient for the purposes of modeling.
I've always been uncomfortable with this process of deriving quantitative data from qualitative interview responses, and I've given a name to what concerns me most: value compression.
Just like data compression, the more tightly values are compressed, the more fidelity to the original is lost. Imagine listening to a symphony orchestra live, then on vinyl, then on CD, then on a series of MP3s of greater and greater compression. I suppose there is a threshold beyond which it would not even sound like music any longer.
So, in the research about which I have been reading, the researcher took the full richness of the subjects' stories, picked them apart, and started sticking the bits into different pigeon holes: this one is about the utility of scenery, this one is about the utility of convenience, and so on. Numbers can be attached to those bland labels by various techniques, and the model is ready to roll.
It seems to me that this does considerable violence to the meanings and values embedded in the interview subjects' stories. The full richness of their lived experience is compressed down to a single dimension, the satisfaction of one or another of a finite set of pre-defined preferences.
Take a hypothetical example, a bit of fiction that occurred to me as I thought about the act of value compression, the violence of it.
"I'll always have a special fondness for sardines," the man said. "You see, my wife always adored them; her favorite breakfast was sardines on toast.
"When she was bed-ridden, in her final illness, I'd greet her every morning with a cup of tea and a plate of sardines on toast. I brought enough for us to share, so we could sit, and talk, and eat sardines for half an hour.
"Then, when she moved to the home, she wasn't supposed to have sardines any more. So, I smuggled them in to her. It became our secret, our conspiracy. I brought her her last illicit snack the day before she passed.
"Now, I still eat sardines for breakfast once every week or so. When I open the can and smell the oily little things, I think of her, of our years together, of our Great Sardine Conspiracy."
The researcher nodded sagely, and jotted in his notes: "Subject prefers fish."