Library

5.2 Skin Tone Representation in Magazines

By Nicholas Bruns
Last updated over 1 year ago
32 Questions
In the project we will survey magazines and study the representation of people with different skin tones in articles and advertisements.

Recall the message of the video from last class:


Research Question

Let's begin with defining our research question. We want to look at the representation of people with different skin tones in magazine articles and advertisements.
1.

In the form of a question, what are we trying to find with our analysis?

2.

How might we go about answering this question? What could we measure or quantify?


Make a Plan: Describe Data Collection Method


Goals:
  1. To determine a way to collect and store images as a group from a selected online magazine.
  2. To create a database, collect and store these images.

Instructions:

Your goal is to collect images from an online magazine. You will use these images in your analysis of skin tone representation.

Working with a team of at most four people, start by choosing a magazine from the ones in the library linked below.

Magazine Library

Once you have selected a magazine, you will take screenshots of the people in them in order to get a sense of which skin tones are represented and which are not.

  • You will be comparing skin tone representation in advertisements vs articles. So it is important to save images in two different folders, eg, one called 'GQ_ads' and one called 'GQ_articles'.

Before digging into the data collection, answer the questions on the right...

Find a group


... and consider the following questions with your team. Write your decisions in the space provided
3.

Browse the Magazine Library link to the left and choose a magazine to study. Write the name of the magazine below:

4.

How many images from articles would be a good sample? How many from advertisements? (Note: there tend to be many more from articles than ads)


5.

What counts as a person for this data collection? For example, if we only see an arm does it count? Will we count each person in the picture? What if there are multiple pictures of the same person? How will you count it?


6.

Should we take a screenshot of the whole page or just of each person?

7.

Should we use a particular file-naming guide to ease our later analysis?

8.

Any other questions or decisions?

9.

Describe, in detail, your data collection method. Be thorough in recording the decisions that you made as this will be an important requirement in your deliverable.

Collect Your Data

Decide on a divide and conquer method and aim to collect as many screenshots as possible.
10.
Before starting collecting data, create two folders in Google Drive, one for screenshots from ads and one from screenshots from articles. Share among teammates and Mr. Bruns.

Now you are ready to start gathering data! This will take most of the period but make sure to leave time to answer the last two questions below.


Don't forget to share your links and to make them available to anyone in the district.
  • Link to the folder with images from ads:
_______
  • Link to the folder with images from articles
_______
11.

Decide on a divide-and-conquer strategy. How will you split the work between group mates to collect as much data as possible in the time allowed?


12.

Collect data!

Take Screenshots and add to your folders.

Note below any questions or issues you encountered along the way.


13.

Reflect on your data collection process with your group members:
  • What went well?
  • What was the most challenging part?
  • Did any questions come up during the process?
  • What do you think are the next steps?


14.

Create a Hypothesis

Based on your limited sample from collecting your data, what did you notice? Discuss with your group members.


Day 2 - Collect Data & Model

Today we will be categorizing the images we collected last class and modeling them with some simple statistical tools that we are already familiar with.

This is a precursor to a more precise method we will get into later using machine learning.
15.

Data collection review

Before we begin...
  • Describe the data that you have collected so far.
  • Describe the data collection method.


16.

Categories to sort data

Before you create a visual of skin tone representation for your data the class will need to decide which categories to sort the data. This will allow you to compare and compile data across magazines.

Consider these questions with your partner:
  • How many categories of skin tone do we want to include?
  • How do we define the skin tones in the categories?  Is there a range of skin tones in each category?
  • What are the possible names of the categories?

Here are some examples:

  • How the Fenty beauty brand labels and organizes their products.




We will have a class discussion about this topic.

Describe the categories from our class discussion that we will use to categorize the data in the space below.

Find some examples from your magazine or elsewhere and create a visual guide for the categories. Paste / create your guide in the Show Your Work section.

Data Collection 2: Categorize Skin Tones from your Screenshots

For our screenshots from the previous class, we need to categorize them to record the frequency of each skin tone. One simple way we can do this is to use a Frequency Table.

Remember: We will keep advertisements and articles separate!

You can use this example as an idea of how to record your results:



Using Frequency Tables:
17.

Divide and conquer OR Cooperate with your partner to record the frequency of each skin tone from your screenshots.

Describe your strategy below


18.

Record your data

Categorize each of the screenshots using a frequency table to tally.

Describe any complications that arose and how you adjusted your strategy.


Modeling And Analyzing Categorical Data with Two-Way Tables


Now you will turn the images you saved into categorical data modeled using two-way tables comparing skin tone categories to the images from ads or articles. One table will include image counts and the other percentages of skin tone representation in magazines.

Model skin tone representation data

Start by creating a two-way table representing the number of images in a skin tone and if the image is from an advertisement or article. If working in a team, make agreements about the category of the image before recording the count in the table.

Create a two-way table representing the percentage of images in a skin tone from advertisements or articles. Consider applying formulas in sheets to the count two-way table to calculate the percentage in each cell of the percentage two-way table.
19.

Record your results

Make a copy of this spreadsheet template and record your results. Paste your link below.

20.

Model

Create a chart in Google Sheets that displays the results of your two way tables visually. (Which kind of chart best displays results here?)

Add images of your two way tables and graphs.


21.

Analysis of the Model

What do you notice? What do you wonder?

What does the data tell you about skin tone representation in the magazine?

What connections can you make to what you read about skin tone representation at the start of the unit?

Who is the target market of your magazine? How could this affect the representation of people with different skin tones in your magazine?


22.

Be skeptical:

  1. How could this data be biased or misinterpreted?
  2. What are some blind spots of this analysis?
  3. How could this analysis be improved?


23.

In the next class, we will share our results with the rest of the class.

Make a prediction about how your results compare to the rest of the class. What do you wonder?


Day 3

Add your data to the class spreadsheet (by period)


Add your count data

Period 7 Class Data
Period 4 Class Data
Period 6 Class Data

We'll analyze our data next class period

Numerical Analysis

Our next step is to do a new kind of analysis where we represent colors with numbers. This will allow us to do a much more complex analysis.

Representing Colors with Numbers


Computers represent colors as a combination of values for red, blue and green light light (not paint). with a value between 0 and 255 for each color.

Note: with the light model of color mixing, higher values are brighter, with white being (255, 255, 255) and black being (0, 0, 0).

Here's a video explanation:

Furthermore, when computers store this combination, they combine them into one big number and use base 16 or Hexadecimal where A = 10, B=11 C = 12 ... F = 15.


This makes our color codes something like this:




Here's a video explanation:

24.

Download an eyedropper app
  • Colorzilla Extension for Chrome (good for Chromebooks and Chrome browser. Pictures must be in Chrome browser): https://chrome.google.com/webstore/detail/colorzilla/bhlhnicpbhignbdhedgjhgdocnmhomnp/related?hl=en
  • Macs have a built in Digital Color Meter (search for it). Hit Command + Shift + C to copy the current color
Try it out, what does it do? What options are available? How could we use this tool?

25.

Special Considerations for Numerical Color Data for Skin Tones

Next we will go back through each of your screenshot folders and use the Eyedropper tool to get a numerical color value for each person in your data set.

Try with an image or two. Does it seem accurate?

Are there any special considerations we shoud take when using this tool?


26.

Record Numerical Color Data

The Colorzilla allows you to copy the color value into your keyboard when you click on a part of your screen. Paste the value below. What does it look like? By default, it should print something like #1B37A4

Record a color code for each of your screenshots in your Google Sheets doc in either Hex # or as a Red, Green, Blue combo. You only need one for now. Since they are two different versions of the same number, we can translate aka clean the data later.

You can make a copy of this Skin Tone Data Collection Template

Ultimately we want something like this:


Paste a screenshot of your spreadsheet data below


Clean the Numerical Data

We ultimately want 3 columns of values for each person: Red, Green, Blue. If you have a hex value, you can extract each color value and translate it from hex (00 - FF) to decimal (0 - 255).

Here's a detailed explanation:
27.

Start by removing the #

Use the =Substitute() function to search your cell for the character "#" and replace it with nothing or ""

Do this for each Hex value you recorded and paste your results below

28.

Next use the =MID() function to only get 2 numbers at a time (first 2 for red, second 2 for green, third 2 for blue)


You can refer to the cell with your hex value (without the #) for the original string.

You should end up with something like the following:

29.

You'll notice from above, the values for red, blue and green are still in Hex. We can covert those to decimal with the =hex2dec() function.

Wrap your MID() functions from above in the hex2dec() function to convert those values to decimal.

Something like this:

This should give you something like this:


30.

Extension

It would be cool to have a visual representation of these different colors. There is a way to color cells based on the values in other cells (you will need to write (copy / adjust / hAcK) some code in order to do this: https://webapps.stackexchange.com/questions/91219/how-do-i-change-a-cell-to-the-color-of-the-hexadecimal-value-of-a-cell-in-google


Day 4 - Graphing in 3D with Python

We can graph points with 3 values on a 3D scatterplot.


In this example python code , they generate random points and put them into a 3D plot. You can copy and paste the code into replit.com project to make it work. Then import your data from your clean RGB values from your spreadsheet and plot those on a 3D graph
31.

Copy and paste the code from the first example on this tutorial into a replit.com (name it "3D Scatterplot") .

This just generates random data.

Try the second example, what does it do? What do you notice is different?


32.

Import your data and graph it in a 3D scatterplot with one axis for red, one for green, one for blue.

This is the code you can use to import data from a spreadsheet into Python:

Remember you must "Publish [your spreadsheet] to Web" as a .csv file in order to import it.

Paste your graph in the show your work section.

What do you notice about your plot? What do you wonder? What story does this tell you?