Data Audiation

Illustration of a folder with papers meant to represent sheets of paper containing worker ownership resources

Whether through pie charts and bar charts or infographics, visualizing data is common in journalistic publications and annual reports alike, helping us to interpret data that may be challenging to understand if presented only in numbers on a spreadsheet. While visual processing is a human specialty and easy to drop into other visual mediums, we have other senses we can use to understand data. What form could those take?

At Data Driven Detroit the result of our efforts is often a presentation of visual data—after gathering, processing, and analyzing data, it at least lives in the tables of a spreadsheet, but usually we pass it on to our data design specialist who will make it into a static chart or graph, or an interactive visualization such as those on State of the Detroit Child or the Housing Information Portal. In her off time, our senior research analyst Stephanie even crochets data visualizations, like this daily high temperature blanket that covers the first year of life of one of her children.

The numbers are the purest representation of the data, and putting them into a visualization adds a layer of abstraction that when done well can help viewers understand the data or reveal patterns that might otherwise not be as obvious. For example, in our Statewide Nonprofit Leadership Census, we learned that there are 2,271 funder-organization connections between 920 funders and 543 nonprofit organizations, with an average of 4.2 funders per organization, but 223 of these organizations had no known funders. While there’s enough information there to pique interest, seeing the collection of the connections on an interactive network map, with the distribution of these organizations by region, and the option to filter by the funder, location, leadership type, and more, nonprofit leadership and funding across the entire state of Michigan becomes more digestible.

Preview image of our Michigan Statewide Nonprofit Leadership Census network map

When done poorly, the visualization can confuse understanding and require so much effort to properly interpret that the most likely conclusion is the wrong one. Take, for instance, this chart produced by the St. Louis Federal Reserve Bank that compares military spending over time between countries. At first glance, it looks like the USA has significantly less spending that has also sharply decreased compared to China’s, which has rapidly increased. But a closer inspection would show the visualization was created with a different axis for the USA, the abstraction hinders understanding that the US still spends significantly more money than China on military spending.

Poorly made line chart by the Federal Reserve that shows military spending by top six spending countries, with the USA on a different axis that misrepresents the relative amount spent

Besides the charts and graphs that we are familiar with, there are other ways to portray data that don’t involve visualization. What form would the data take if it was created for our other senses instead? We can differentiate between sweet and salty flavors, a literal pie chart of cloudier days represented by slices of salted caramel and slices of key lime to represent sunny days. A mist of floral perfume with greater intensity to represent days of blooming flowers and a mist of an earthy perfume to represent days of days without blooms. A blood pressure cuff tightening and loosening relative to a dataset of housing insecurity of populations in different census tracts.

We can also use western musical notation to assign notes to data points. As a simple example, using the Two Tone app, here are the high temperatures for each day of 2022 in Detroit, with higher and lower temperatures correlating to higher and lower notes played on a piano in the key of C major:

Detroit Daily High Temperature in C Major

January - December 2022

We can hear the lower notes in the beginning of the year gradually get higher, and stay at a higher register until dropping down again at the end of the year. With eight notes in an octave and these data represented by eight octaves, that gives us 64 notes to represent the 80.1 degrees of high temperature measurement range from 16°F to 96.1°F in 2022. This creates an imperfect connection between the audio abstraction to the dataset, but besides those with perfect pitch, it is still possible to interpret the data through sound instead of visually.

Many pop songs are written in major keys, which tend to evoke more positive feelings, and minor keys are often utilized for evoking more negative emotions. As our high temperatures rise over time with climate change and we create more complex audiations, the key can contribute to the interpretation of the audiation. 

By utilizing data with multiple sets of the same form, we can apply more notes, octaves, and scales to better represent the sound and feel of the data. In this example, we’re using a jazzy Dorian scale in D sharp to represent counts of pedestrian- and cyclist-involved crashes in Detroit by age from 2017 to 2021. The piano represents the count of pedestrian-involved crashes, the trumpet cyclist-involved crashes, and the upright bass the combination of both:

Detroit Pedestrian- and Cyclist-Involved Crashes by Age

2017 – 2021

With far more pedestrian-involved crashes than cyclist-involved crashes, the piano is more distinct across all ages, and the higher notes of the trumpet help to distinguish the age ranges of higher counts that swell around the ages of 20 and 55. But to maintain a sense of musicality, each instrument has unique attributes that increase abstraction: four notes played per pedestrian age starting at age 1 and ending at age 99 with a descending arpeggio; two notes played per cyclist age with a descending arpeggio, and 1 note played per age of the pedestrians and cyclists combined with no arpeggio. The piano gets three octaves to account for the broader range of pedestrian-involved count compared to the trumpet and base which are restricted to two octaves. All of this mixed together helps convey the chaos of the events represented by the data while sounding at least somewhat musical.

Focusing more on the sad emotions that reflect the nature of the pedestrian- and cyclist-involved crashes, this is a smaller dataset of the crash count per hour. Inspired by Arvo Pärt’s Spiegel im Spiegel, this version has piano representing pedestrian-involved crashes and the violin cyclist-involved crashes:

Detroit Pedestrian- and Cyclist-Involved Crashes by Hour

2017 – 2021

This stretches out the notes representing 24 individual hours into two morose minutes of music that probably wouldn’t pass even an avant garde composition class, but conveys the sadness of pedestrians and cyclists being hit by drivers.

To make a comparison of nonprofit organizations across Michigan regions, in this audiation we have six regions all represented by the marimba. Each of the 16 notes represents the percentage of nonprofit organizations in that region that focus their efforts on the following 16 equity issues from our Statewide Nonprofit Leadership Census:

  • Aging and Older Adults
  • Arts and Culture
  • Carceral System
  • Class, Labor, and Economy
  • Disability and Accessibility
  • Education
  • Environment and Climate
  • Faith and Religion
  • Food
  • Gender and LGBTQIA2S+ Identity
  • Healthcare
  • Housing
  • Immigration, Migration, or Refugees
  • Language
  • Race and Ethnicity
  • Transportation

Michigan Statewide Nonprofit Survey – Equity Issue

Upper Peninsula Comparison – 2021

All of the notes hitting at once gives an auditory average of the percentage, and to highlight the Upper Peninsula region as a comparison it gets three hits for every one of the other regions. By listening to the pitch of the extra two notes, one can compare whether Upper Peninsula nonprofits have a higher or lower than average focus on an equity issue compared to nonprofits statewide, or if no nonprofits in the UP are addressing that equity issue.

Population in the Detroit tri-county area of Wayne-Oakland-Macomb goes back to 1840 in the decennial census. In this audiation the pipe organ in the background is Wayne County that rises before Oakland County (piano) and Macomb County (glockenspiel), but we can also hear the pitch of the organ start declining in 1980 while the other two continue to rise:

These initial exploratory audiations have been limited by utilizing only the basic functions of the TwoTone app, there are certainly more possibilities for incorporating different presentations of sound and instrumentation, pairing them with interactive or animated visualizations, and possibly combining them with our other senses as well.

If you have an idea for a data audiation but need data, you can fill out our AskD3 form and we can complete a data request for you for free!

Copyright © 2022 Data Driven Detroit. All Rights Reserved.