Is exploration good for our wellbeing?

In this project, I used multilevel regression to examine whether everyday exploration relates to wellbeing.

It may not seem like it, but everyday life offers many opportunities to explore: We can take a different route during our daily jog, try a new coffee shop, or finally visit that museum we’ve been meaning to check out.

Exploring our environments allows us to engage with novel people and situations — rewarding experiences that psychologists believe may be beneficial to our wellbeing.

In collaboration with researchers at NYU and the University of Miami, I led a study that examined whether everyday exploration relates to emotional and social wellbeing.

The data and R scripts to replicate these findings can be found on this project's GitHub repository.

Designing the study: How can we measure everyday exploration?

Before jumping into the analyses, let me provide some background for how we measured “exploration” — a pretty abstract idea. How do we distill the complexity of everyday human behavior into a single measure?

Although this study was in humans, we drew from a study in mice that defined exploration as the randomness of one’s movement over a given space during a given period of time. The mouse study attached little antennas to mice to capture their movement in a constructed maze. For every day of data collection, the researchers used a probabilistic model that used all of the locations visited by each mouse to compute that mouse’s exploration value for a given day.

Fortunately for us, we didn’t have to attach antennas to our participants or construct a maze for them to explore. Instead, we happened to have two things working in our favor: (1) most humans already have an antenna practically attached to them — a smartphone with GPS tracking — and (2) our participants happened to be located in the closest thing to a maze for humans: New York City!

We translated this design for use in humans. First, participants downloaded an app that let us record their continuous GPS location. They kept this app on their phones for three months while we recorded their location approximately every minute of the day (with their consent, of course!). These data were collected prior to the quarantine, so no lockdowns or restrictions were affecting people's daily routines.

We drew from existing research in mice in order to distill thousands of GPS data points into a composite measure of daily exploration.

During these three months, we also sent participants text messages approximately every 48 hours asking them how they were feeling.

By the end of the study, we had collected tens of thousands of GPS coordinates. Using the same probabilistic model used in the original mouse study, I computed participant-specific exploration values for every day of data collection.

The fun stuff: Data structure and analysis

In this design, the data were nested within person: Each participant had ~90 days of GPS data (with a single exploration value per day), and on about half of those days, they rated how good they were feeling using a 0 to 100 scale. Once I had taken the raw GPS coordinates and computed daily exploration values, the "tidy" data looked something like this:

Participant ID	Date	​Exploration value	​Mood rating (0 to 100)
P001	07/01	7.1	75
P001	07/02	5.2	55
P001	07/03	4.3	40
...	...	...	...
P999	07/01	8.2	86
P999	07/02	10.1	98

(This is just an example. If you want to see the actual data, check out the GitHub repository for this study.)

As you can see, each observation (row) is a single day's data, nested within participant. To account for the hierarchical structure of the data, I used multilevel regression to model the relationship between daily exploration and mood ratings. I used the lme4 package in R to fit this lmer model:

mood rating ~ exploration value + (1 + exploration value | participant)

I decided to allow for person-specific intercepts and slopes in this model. It's reasonable to assume that people vary from one another in their average exploration levels (intercepts). It's also possible that for some people, daily exploration is very rewarding (i.e., linked to much higher mood), while for others, the association isn't quite as strong (slopes).

What we found

Using this model, we found that greater daily exploration was associated with better mood. Below, I've plotted each person's specific effect (colorful lines), as well as the group-level, fixed effect (black dashed line). As you can see, as daily exploration values increase, mood ratings increase as well.

Presumably, other factors could affect *both* your level of exploration and your mood on a given day. For instance, a sticky, humid summer day in New York could cause you to stay in your air conditioned apartment, lowering your exploration, and also put you in a bad mood (because, let's be honest, it smells like trash outside). This poses a challenge known as the third variable problem, where a third variable (in this case, bad weather) could be explaining the relationship between our two variables of interest.

To test this possibility, I scraped weather data from the web (darksky.net). Using the raw GPS coordinates and dates, I downloaded the temperature and precipitation for a given location on a given day. I was then able to incorporate these values into my multilevel model:

mood rating ~ exploration value + precipitation levels + temperature + (1 + exploration value | participant)

My results remain unchanged: Even after controlling for weather patterns, people reported better moods in days in which they explored more. Although we can’t say with certainty that exploring more causes someone to feel good, these findings definitely suggest that exploration and daily mood are linked.

Bonus question: Do people who explore more have larger social networks?

Many scientists believe that, evolutionarily, exploration serves several adaptive functions. One such function is helping us grow our social networks, a task that is fundamental to human survival given that we live in highly social environments.

Based on this reasoning, I wanted to know whether people who explored their environments more also had larger social networks. Halfway through the study, I asked people how many people they had been in touch with via phone calls, text messaging, or other messaging platforms in the past two weeks.

I used these data to test whether people who explored more on average also had larger social networks. Because each person had a single observation for the predictor (average daily exploration value) and the outcome (number of people in their social network) variable in this model, I used a good old-fashioned linear regression using the 'lm' function.

As expected, I found that people who had higher average exploration levels also had larger social networks!

Key takeaways

Based on these results, the main takeaways are:

1. Day-to-day variations in exploration are linked to emotional wellbeing: People reported better moods on days in which they explored more.

2. This finding is true even after accounting for potential confounding variables, including weather patterns.

3. Individuals who explore more (on average) also have larger social networks.

4. Together, these results suggest that exploration is positively linked to emotional and social wellbeing.

Interested in learning more?

The peer-reviewed manuscript that details this study is published in Psychological Science. You can read it here!