For my second project, I explored my interactions with strangers over a 2-week period. My interest in this topic stems from a fascination with “small” talk. While others tend to see this as fluff or filler, i.e., without much substance, I noticed during the pandemic — a period of intense isolation — I benefited immensely from seemingly minor interactions with neighbors, strangers on the street, baristas, etc. Even when we didn’t talk about anything important, the fact of engaging at all was what made the difference.

My research question for this project was: What is the quality of my conversations with strangers? How often do I speak with strangers, who do I tend to have conversations with, and do these conversations have an impact on my mood?

These questions have implications for anyone interested in the quality and patterns of their own experience. Because whether we see lots of strangers in our day-to-day or none at all, their presence (or absence) likely affects us in ways we might not even be aware of. To this extent, this research area would be of interest to researchers in fields like social sciences, anthropology, and digital humanities.

Before diving into my visualizations, I’ll share more detail about how I collected the data. To start, I only logged interactions with strangers that went beyond “hello.” I didn’t count these because a simple nod or acknowledgement, I found, typically did not alter the flow of my thinking or movement. Instead, I tracked any interaction that included at least 2 active engagements. These ranged from brief exchanges with dog owners (e.g., Person 1: “Hi.” Person 2: “Hi, what’s your dog’s name?” Person 1: “It’s Jeff, how about yours?” Person 2: “Bark Twain.”) to lengthy customer support calls to deal with health insurance billing issues.

Immediately after each interaction, I would log a series of variables: the category of interaction (was this a worker with whom I was transacting? A fellow dog owner? A completely random stranger on the street?); the duration; the location (latitude and longitude); my mood immediately before; and my mood immediately after.

Mood was measured on a scale from 0 to 10, with 0 being the saddest, worst feeling ever, and 10 being complete elation. In reality, my mood ranged from 5.0 to 8.3 throughout the entire data collection period, and my mood would change anywhere from -.03 to +.05 after a given interaction.

OK, onto the visualizations.

You can see all 4 visualizations in a single dashboard here.

I chose a tree map, below, to represent the distribution of conversation lengths across the dataset. I went with a tree map to show how parts add up to a whole (the entire dataset) while giving more real estate to the dominant duration bin — which, in this case, was interactions of 15 seconds or less. I chose blue as a neutral color that differs from green or red, which are predominant in the subsequent visuals.

The next chart is a pie chart. I know they have a poor reputation — but I felt this was the simplest way to show the distribution of the effect on my mood, a categorical variable with 3 categories. Readers will likely understand intuitively the meaning behind the colors — green meaning positive, red meaning negative. I also accentuated the green by changing the font color to green for the title word “boost.”

I next dove deeper into the emotional valence of these interactions. I was especially excited to figure out how to create this visual. I knew I wanted to represent the distribution of data along a quantitative measure (positive or negative change in mood), broken down into three categories of interaction (dog owner, worker, and miscellaneous).

So I plotted the distribution for each category, and then added a “jitter” (a random small increase or decrease) to the individual points so that they would not completely overlap and obscure one another — this way, you can see where data points tend to cluster.

I also further encoded the emotional valence using red, gray, and green — with the hope this would help viewers understand the chart more quickly and intuitively.

Finally, I kept up with the color scheme to look just at the average mood boost by category of interaction. The average across categories was consistently positive, so you only see shades of green here, with a deeper shade representing a stronger emotional effect. I add a label to offer concrete examples of what I’ve tagged as miscellaneous interactions.

In terms of next steps, I think this research would benefit from a more substantive dataset. For one, tracking this over a longer period of time would increase the n and make any subsequent analysis more solid. I decided against doing any sort of geographical mapping of the data points, because there weren’t enough to make for a compelling narrative. I bet that would change with more data.

I would also like to expand this to more individuals during the data collection stage. This research is inherently subjective, not only because it measures something as fickle as mood, but also because individuals log the data themselves and decide what to represent. So doing it across more people would get a wider range of experiences and expand the the possibilities of the analysis.