This lecture, as the rest of the course, is adapted from the version Stephanie C. Hicks designed and maintained in 2021 and 2022. Check the recent changes to this file through the GitHub history.
Material for this lecture was borrowed and adopted from
At the end of this lesson you will:
Be able to state best practices for:
In philosophy departments, classes and modules centered around data ethics are widely discussed.
The ethical challenges around working with data are not fundamentally different from the ethical challenges philosophers have always faced.
However, putting an ethical framework around building data analyses in practice is indeed new for most data scientists, and for many of us, we are woefully under-prepared to teach so far outside our comfort zone.
That being said, we can provide some thoughts on how to approach a data science problem using a philosophical lens.
We start with a grounding in the definition of Ethics:
Ethics, also called moral philosophy, has three main branches:
While, unfortunately, there are myriad examples of ethical data science problems (see, for example, blog posts bookclub and data feminism), here I aim to connect some of the broader data science ethics issues with the existing philosophical literature.
Note, I am only scratching the surface and a deeper dive might involve education in related philosophical fields (epistemology, metaphysics, or philosophy of science), philosophical methodologies, and ethical schools of thought, but you can peruse all of these through, for example, a course or readings introducing the discipline of philosophy.
Below we provide some thoughts on how to approach a data science problem using a philosophical lens.
We begin by considering a case study around ethical data analyses.
Many ethics case studies provided in a classroom setting describe algorithms built on data which are meant to predict outcomes.
Large scale algorithmic decision making presents particular ethical predicaments because of both the scale of impact and the “black-box” sense of how the algorithm is generating predictions.
Consider the well-known issue of using facial recognition software in policing.
There are many questions surrounding the policing issue:
[Source: CNN]
The two main ethical concerns surrounding facial recognition software break down into
When thinking about the questions below, reflect on the good aspects and the bad aspects and how one might weight the good versus the bad.
To settle the questions above, we need to study various ethical theories, and it turns out that the different theories may lead us to different conclusions. As non-philosophers, we recognize that the suggested readings and ideas may come across as overwhelming. If you are overwhelmed, we suggest that you choose one ethical theory, think carefully about how it informs decision making, and help your students to connect the ethical framework to a data science case study.
This is a challenging topic, but as you analyze data, ask yourself the following broad questions to help you with ethical considerations around the data analysis.
For more on this line of thought read this book by @CCriadoPerez or other books 📚 https://t.co/ws3ctX3ijr pic.twitter.com/8IYmCpxeGx
— 🇲🇽 Leonardo Collado-Torres (@lcolladotor) April 2, 2024
Data sharing is an essential element of the scientific method, imperative to ensure transparency and reproducibility.
Different areas of research collect fundamentally different types of data, such as tabular data, time series data, image data, or genomic data. These types of data differ greatly in size and require different approaches for sharing.
In this section, I outline broad best practices to make your data publicly accessible and usable, generally and for several specific kinds of data.
Sharing data proves more useful when others can easily find and access, interpret, and reuse the data. To maximize the benefit of sharing your data, follow the findable, accessible, interoperable, and reusable (FAIR) guiding principles of data sharing, which optimize reuse of generated data.
The following material is from a LIBD rstats club presentation I gave in 2023-09-15.
I’ve adapted some of the content from the public notes.
Pros:
Cons:
Pros:
Cons:
styler as shown here: https://github.com/LieberInstitute/template_project/blob/devel/code/update_style.RPros:
Most future proof version.
Enables downstream use of the GitHub contributor graph instead of relying on pure memory.
Cons:
“The greatest value of a picture is when it forces us to notice what we never expected to see.” -John W. Tukey
Mistakes, biases, systematic errors and unexpected variability are commonly found in data regardless of applications. Failure to discover these problems often leads to flawed analyses and false discoveries.
As an example, consider that measurement devices sometimes fail and not all summarization procedures, such as the mean() function in R, are designed to detect these. Yet, these functions will still give you an answer.
Furthermore, it may be hard or impossible to notice an error was made just from the reported summaries.
Data visualization is a powerful approach to detecting these problems. We refer to this particular task as exploratory data analysis (EDA), coined by John Tukey.
On a more positive note, data visualization can also lead to discoveries which would otherwise be missed if we simply subject the data to a battery of statistical summaries or procedures.
When analyzing data, we often make use of exploratory plots to motivate the analyses we choose.
In this section, we will discuss some types of plots to avoid, better ways to visualize data, some principles to create good plots, and ways to use ggplot2 to create expository (intended to explain or describe something) graphs.
The following figure is from Lippmann et al. 2006:
The following figure is from Dominici et al. 2007, in response to the work by Lippmann et al. above.
Elevated levels of Ni and V PM2.5 chemical components in New York are likely attributed to oil-fired power plants and emissions from ships burning oil, as noted by Lippmann et al. (2006).
In order to determine the effectiveness or quality of a visualization, we need to first understand three things:
No plot (or any statistical tool, really) can be judged without knowing the answers to those questions. No plot or graphic exists in a vacuum. There is always context and other surrounding factors that play a role in determining a plot’s effectiveness.
Conversely, high-quality, well-made visualizations usually allow one to properly deduce what question is being asked and who the audience is meant to be. A good visualization tells a complete story in a single frame.
The act of visualizing data typically proceeds in two broad steps:
Again I highly recommend checking the content made by Christine Zhang!
Actually, on 2025-10-16 we were lucky enough to have Christine visit us in person and give a guest lecture. Below you can watch a recording of it and check her slides. The YouTube video has the links to all the resources she listed.
Slides:
Zoom AI summary:
Christine Zhang, a data journalist and graphics editor at the New York Times, discussed the importance of community in the R programming language and shared her experience with data visualization principles and election result mapping. She explored various visualization techniques, including the use of color scales, annotations, and uncertainty representation in election polls, while emphasizing the importance of accurate data representation and visual consistency. Christine also shared examples of creative data visualizations and discussed the collaboration between graphics editors and reporters in creating effective visual storytelling across different platforms.
Zoom AI next steps:
- Students to explore the GitHub repo Christine created for the Anne Arundel County election map project.
- Students to review the Visual Vocabulary poster at https://ft.com/vocabulary for guidance on chart selection.
Tools mentioned by Christine:
R: https://cran.r-project.org/ggplot2: https://ggplot2.tidyverse.org/skimr: https://docs.ropensci.org/skimr/Adobe Illustrator: https://www.adobe.com/products/illustrator.htmlrawgraphs: https://www.rawgraphs.io/Datawrapper: https://www.datawrapper.de/Layer Cake: https://layercake.graphics/Self practice links:
Other recommended resources by Christine:
Thank you very much @christinezhang.bsky.social for taking the time to share your experience with us! I’m sure that everyone who attended your talk or sees the recording will benefit from it when crafted their own data visualizations 🎨
Feel free to come visit us in Baltimore 🦀 whenever you want ^^
[image or embed] — 🇲🇽 Leonardo Collado-Torres (@lcolladotor.bsky.social) October 24, 2025 at 3:30 PM
Initially, one must decide what information should be presented. The following principles for developing analytic graphics come from Edward Tufte’s book Beautiful Evidence.
What’s sub-optimal about this plot?
d <- airquality %>%
mutate(Summer = ifelse(Month %in% c(7, 8, 9), 2, 3))
with(d, {
plot(Temp, Ozone, col = unclass(Summer), pch = 19, frame.plot = FALSE)
legend("topleft",
col = 2:3, pch = 19, bty = "n",
legend = c("Summer", "Non-Summer")
)
})
What’s sub-optimal about this plot?
airquality %>%
mutate(Summer = ifelse(Month %in% c(7, 8, 9),
"Summer", "Non-Summer"
)) %>%
ggplot(aes(Temp, Ozone)) +
geom_point(aes(color = Summer), size = 2) +
theme_minimal()
Some of these principles are taken from Edward Tufte’s Visual Display of Quantitative Information:
This section is based on a talk by Karl W. Broman titled “How to Display Data Badly”, in which he described how the default plots offered by Microsoft Excel “obscure your data and annoy your readers” (here is a link to a collection of Karl Broman’s talks).
Karl’s lecture was inspired by the 1984 paper by H. Wainer: How to display data badly. American Statistician 38(2): 137–147.
Dr. Wainer was the first to elucidate the principles of the bad display of data.
However, according to Karl Broman, “The now widespread use of Microsoft Excel has resulted in remarkable advances in the field.”
Here we show examples of “bad plots” and how to improve them in R.
Here are some examples of bad plots and suggestions on how to improve
Let’s say we are interested in the most commonly used browsers. Wikipedia has a table with the “usage share of web browsers” or the proportion of visitors to a group of web sites that use a particular web browser from July 2017.
browsers <- c(
Chrome = 60, Safari = 14, UCBrowser = 7,
Firefox = 5, Opera = 3, IE = 3, Noinfo = 8
)
browsers.df <- gather(
data.frame(t(browsers)),
"browser", "proportion"
)Let’s say we want to report the results of the usage. The standard way of displaying these is with a pie chart:
pie(browsers, main = "Browser Usage (July 2022)")
If we look at the help file for pie():
?pieIt states:
“Pie charts are a very bad way of displaying information. The eye is good at judging linear measures and bad at judging relative areas. A bar chart or dot chart is a preferable way of displaying this type of data.”
To see this, look at the figure above and try to determine the percentages just from looking at the plot. Unless the percentages are close to 25%, 50% or 75%, this is not so easy. Simply showing the numbers is not only clear, but also saves on printing costs.
Having said that, see how we used a pie chart in a 2023 pre-print https://www.biorxiv.org/content/biorxiv/early/2023/02/15/2023.02.15.528722/F5.large.jpg. The published version is accessible from doi 10.1126/science.adh1938.
If you do want to plot them, then a barplot is appropriate. Here we use the geom_bar() function in ggplot2. Note, there are also horizontal lines at every multiple of 10, which helps the eye quickly make comparisons across:
p <- browsers.df %>%
ggplot(aes(
x = reorder(browser, -proportion),
y = proportion
)) +
geom_bar(stat = "identity")
p
Notice that we can now pretty easily determine the percentages by following a horizontal line to the x-axis.
While this figure is already a big improvement over a pie chart, we can do even better. When you create figures, you want your figures to be self-sufficient, meaning someone looking at the plot can understand everything about it.
Some possible critiques are:
Let’s explore how to do these things to make an even better figure.
To start, go to the help file for theme()
?ggplot2::themeWe see there are arguments with text that control all the text sizes in the plot. If you scroll down, you see the text argument in the theme command requires class element_text. Let’s try it out.
To change the x-axis and y-axis labels to be full names, use xlab() and ylab()
p <- p + xlab("Browser") +
ylab("Proportion of Users")
p
Maybe a title
p + ggtitle("Browser Usage (July 2022)")
Next, we can also use the theme() function in ggplot2 to control the justifications and sizes of the axes, labels and titles.
To center the title
p + ggtitle("Browser Usage (July 2022)") +
theme(plot.title = element_text(hjust = 0.5))
To create bigger text/labels/titles:
p <- p + ggtitle("Browser Usage (July 2022)") +
theme(
plot.title = element_text(hjust = 0.5),
text = element_text(size = 15)
)
p
p + theme_bw()
p + theme_dark()
p + theme_classic() # axis lines!
p + ggthemes::theme_base()
Please, avoid a 3D version because it obfuscates the plot, making it more difficult to find the percentages by eye.
Even worse than pie charts are donut plots.
The reason is that by removing the center, we remove one of the visual cues for determining the different areas: the angles. There is no reason to ever use a donut plot to display data.
Why are pie/donut charts so common?
https://blog.usejournal.com/why-humans-love-pie-charts-9cd346000bdc
While barplots are useful for showing percentages, they are incorrectly used to display data from two groups being compared. Specifically, barplots are created with height equal to the group means; an antenna is added at the top to represent standard errors. This plot is simply showing two numbers per group and the plot adds nothing:
If the number of points is small enough, we might as well add them to the plot. When the number of points is too large for us to see them, just showing a boxplot is preferable.
Let’s recreate these barplots as boxplots and overlay the points. We will simulate similar data to demonstrate one way to improve the graphic above.
set.seed(1000)
dat <- data.frame(
"Treatment" = rnorm(10, 30, sd = 4),
"Control" = rnorm(10, 36, sd = 4)
)
gather(dat, "type", "response") %>%
ggplot(aes(type, response)) +
geom_boxplot() +
geom_point(position = "jitter") +
ggtitle("Response to drug treatment")
Notice how much more we see here: the center, spread, range, and the points themselves. In the barplot, we only see the mean and the standard error (SE), and the SE has more to do with sample size than with the spread of the data.
This problem is magnified when our data has outliers or very large tails. For example, in the plot below, there appears to be very large and consistent differences between the two groups:
However, a quick look at the data demonstrates that this difference is mostly driven by just two points.
set.seed(1000)
dat <- data.frame(
"Treatment" = rgamma(10, 10, 1),
"Control" = rgamma(10, 1, .01)
)
gather(dat, "type", "response") %>%
ggplot(aes(type, response)) +
geom_boxplot() +
geom_point(position = "jitter")
A version showing the data in the log-scale is much more informative.
gather(dat, "type", "response") %>%
ggplot(aes(type, response)) +
geom_boxplot() +
geom_point(position = "jitter") +
scale_y_log10()
A common task in data analysis is the comparison of two groups. When the dataset is small and data are paired, such as the outcomes before and after a treatment, two-color barplots are unfortunately often used to display the results.
There are better ways of showing these data to illustrate that there is an increase after treatment. One is to simply make a scatter plot, which shows that most points are above the identity line. Another alternative is to plot the differences against the before values.
set.seed(1000)
before <- runif(6, 5, 8)
after <- rnorm(6, before * 1.15, 2)
li <- range(c(before, after))
ymx <- max(abs(after - before))
par(mfrow = c(1, 2))
plot(before, after,
xlab = "Before", ylab = "After",
ylim = li, xlim = li
)
abline(0, 1, lty = 2, col = 1)
plot(before, after - before,
xlab = "Before", ylim = c(-ymx, ymx),
ylab = "Change (After - Before)", lwd = 2
)
abline(h = 0, lty = 2, col = 1)
Line plots are not a bad choice, although they can be harder to follow than the previous two. Boxplots show you the increase, but lose the paired information.
z <- rep(c(0, 1), rep(6, 2))
par(mfrow = c(1, 2))
plot(z, c(before, after),
xaxt = "n", ylab = "Response",
xlab = "", xlim = c(-0.5, 1.5)
)
axis(side = 1, at = c(0, 1), c("Before", "After"))
segments(rep(0, 6), before, rep(1, 6), after, col = 1)
boxplot(before, after,
names = c("Before", "After"),
ylab = "Response"
)
The above plot was made using ggpubr::ggpaired(). Note that the title of the package is:
ggpubr: ‘ggplot2’ Based Publication Ready Plots
The figure below shows three curves. Pseudo 3D is used, but it is not clear why. Maybe to separate the three curves? Notice how difficult it is to determine the values of the curves at any given point:
This plot can be made better by simply using color to distinguish the three lines:
x <- read_csv("https://github.com/kbroman/Talk_Graphs/raw/master/R/fig8dat.csv") %>%
as_tibble(.name_repair = make.names)
p <- x %>%
gather("drug", "proportion", -log.dose) %>%
ggplot(aes(
x = log.dose, y = proportion,
color = drug
)) +
geom_line()
p
This plot demonstrates that using color is more than enough to distinguish the three lines.
We can make this plot better using the functions we learned above
p + ggtitle("Survival proportion") +
theme(
plot.title = element_text(hjust = 0.5),
text = element_text(size = 15)
)
We can also move the legend inside the plot
p + ggtitle("Survival proportion") +
theme(
plot.title = element_text(hjust = 0.5),
text = element_text(size = 15),
legend.position = c(0.2, 0.3)
)
We can also make the legend transparent
transparent_legend <- theme(
legend.background = element_rect(fill = "transparent"),
legend.key = element_rect(
fill = "transparent",
color = "transparent"
)
)
p + ggtitle("Survival proportion") +
theme(
plot.title = element_text(hjust = 0.5),
text = element_text(size = 15),
legend.position = c(0.2, 0.3)
) +
transparent_legend
By default, statistical software like R returns many significant digits. This does not mean we should report them. Cutting and pasting directly from R is a bad idea since you might end up showing a table, such as the one below, comparing the heights of basketball players:
heights <- cbind(
rnorm(8, 73, 3), rnorm(8, 73, 3), rnorm(8, 80, 3),
rnorm(8, 78, 3), rnorm(8, 78, 3)
)
colnames(heights) <- c("SG", "PG", "C", "PF", "SF")
rownames(heights) <- paste("team", 1:8)
heights SG PG C PF SF
team 1 68.88065 73.07480 81.80948 76.60455 82.23521
team 2 70.05272 66.86024 74.64847 72.70140 78.55640
team 3 71.33653 73.63946 81.00483 78.56787 77.86893
team 4 73.36414 81.01021 81.68293 76.90146 77.35226
team 5 72.63738 69.31895 83.66281 81.17280 82.39133
team 6 68.99188 75.50274 79.36564 75.77514 78.68900
team 7 73.51017 74.59772 82.09829 73.95492 78.32287
team 8 73.46524 71.05953 77.88069 76.44808 73.86569We are reporting precision up to 0.00001 inches. Do you know of a tape measure with that much precision? This can be easily remedied:
round(heights, 1) SG PG C PF SF
team 1 68.9 73.1 81.8 76.6 82.2
team 2 70.1 66.9 74.6 72.7 78.6
team 3 71.3 73.6 81.0 78.6 77.9
team 4 73.4 81.0 81.7 76.9 77.4
team 5 72.6 69.3 83.7 81.2 82.4
team 6 69.0 75.5 79.4 75.8 78.7
team 7 73.5 74.6 82.1 74.0 78.3
team 8 73.5 71.1 77.9 76.4 73.9Recall the plot we had before:
transparent_legend <- theme(
legend.background = element_rect(fill = "transparent"),
legend.key = element_rect(
fill = "transparent",
color = "transparent"
)
)
p + ggtitle("Survival proportion") +
theme(
plot.title = element_text(hjust = 0.5),
text = element_text(size = 15),
legend.position = c(0.2, 0.3)
) +
xlab("dose (mg)") +
transparent_legend
What type of caption would be good here?
When creating figure captions, think about the following:
A plot of the proportion of patients who survived after three drug treatments.
Figure 1. A plot of the proportion of patients who survived after three drug treatments.
Figure 1. Drug treatment survival. A plot of the proportion of patients who survived after three drug treatments.
Figure 1. Drug treatment survival. A plot of the proportion of patients who survived after three drug treatments (milligram).
Figure 1. Drug treatment survival. A plot of the proportion of patients who survived after three drug treatments (milligram). Three colors represent three drug treatments. Drug A results in largest survival proportion for the larger drug doses.
In general, you should follow these principles:
Create expository graphs to tell a story (figure and caption should be self-sufficient; it’s the first thing people look at)
Show as much information as possible, taking care not to obscure the message
Science not sales: avoid unnecessary frills (especially gratuitous 3D)
In tables, every digit should be meaningful
options(width = 120)
sessioninfo::session_info()─ Session info ───────────────────────────────────────────────────────────────────────────────────────────────────────
setting value
version R version 4.5.1 (2025-06-13)
os macOS Sequoia 15.6.1
system aarch64, darwin20
ui X11
language (EN)
collate en_US.UTF-8
ctype en_US.UTF-8
tz America/New_York
date 2025-10-24
pandoc 3.7.0.2 @ /opt/homebrew/bin/ (via rmarkdown)
quarto 1.4.550 @ /Applications/quarto/bin/quarto
─ Packages ───────────────────────────────────────────────────────────────────────────────────────────────────────────
package * version date (UTC) lib source
bit 4.6.0 2025-03-06 [1] CRAN (R 4.5.0)
bit64 4.6.0-1 2025-01-16 [1] CRAN (R 4.5.0)
cli 3.6.5 2025-04-23 [1] CRAN (R 4.5.0)
colorout * 1.3-2 2025-05-09 [1] Github (jalvesaq/colorout@572ab10)
crayon 1.5.3 2024-06-20 [1] CRAN (R 4.5.0)
curl 7.0.0 2025-08-19 [1] CRAN (R 4.5.0)
dichromat 2.0-0.1 2022-05-02 [1] CRAN (R 4.5.0)
digest 0.6.37 2024-08-19 [1] CRAN (R 4.5.0)
dplyr * 1.1.4 2023-11-17 [1] CRAN (R 4.5.0)
evaluate 1.0.5 2025-08-27 [1] CRAN (R 4.5.0)
farver 2.1.2 2024-05-13 [1] CRAN (R 4.5.0)
fastmap 1.2.0 2024-05-15 [1] CRAN (R 4.5.0)
forcats * 1.0.1 2025-09-25 [1] CRAN (R 4.5.0)
generics 0.1.4 2025-05-09 [1] CRAN (R 4.5.0)
ggplot2 * 4.0.0 2025-09-11 [1] CRAN (R 4.5.0)
ggthemes 5.1.0 2024-02-10 [1] CRAN (R 4.5.0)
glue 1.8.0 2024-09-30 [1] CRAN (R 4.5.0)
gtable 0.3.6 2024-10-25 [1] CRAN (R 4.5.0)
hms 1.1.3 2023-03-21 [1] CRAN (R 4.5.0)
htmltools 0.5.8.1 2024-04-04 [1] CRAN (R 4.5.0)
htmlwidgets 1.6.4 2023-12-06 [1] CRAN (R 4.5.0)
jsonlite 2.0.0 2025-03-27 [1] CRAN (R 4.5.0)
knitr 1.50 2025-03-16 [1] CRAN (R 4.5.0)
labeling 0.4.3 2023-08-29 [1] CRAN (R 4.5.0)
lifecycle 1.0.4 2023-11-07 [1] CRAN (R 4.5.0)
lubridate * 1.9.4 2024-12-08 [1] CRAN (R 4.5.0)
magrittr 2.0.4 2025-09-12 [1] CRAN (R 4.5.0)
pillar 1.11.1 2025-09-17 [1] CRAN (R 4.5.0)
pkgconfig 2.0.3 2019-09-22 [1] CRAN (R 4.5.0)
purrr * 1.1.0 2025-07-10 [1] CRAN (R 4.5.0)
R6 2.6.1 2025-02-15 [1] CRAN (R 4.5.0)
RColorBrewer 1.1-3 2022-04-03 [1] CRAN (R 4.5.0)
readr * 2.1.5 2024-01-10 [1] CRAN (R 4.5.0)
rlang 1.1.6 2025-04-11 [1] CRAN (R 4.5.0)
rmarkdown 2.29 2024-11-04 [1] CRAN (R 4.5.0)
S7 0.2.0 2024-11-07 [1] CRAN (R 4.5.0)
scales 1.4.0 2025-04-24 [1] CRAN (R 4.5.0)
sessioninfo 1.2.3 2025-02-05 [1] CRAN (R 4.5.0)
stringi 1.8.7 2025-03-27 [1] CRAN (R 4.5.0)
stringr * 1.5.2 2025-09-08 [1] CRAN (R 4.5.0)
tibble * 3.3.0 2025-06-08 [1] CRAN (R 4.5.0)
tidyr * 1.3.1 2024-01-24 [1] CRAN (R 4.5.0)
tidyselect 1.2.1 2024-03-11 [1] CRAN (R 4.5.0)
tidyverse * 2.0.0 2023-02-22 [1] CRAN (R 4.5.0)
timechange 0.3.0 2024-01-18 [1] CRAN (R 4.5.0)
tzdb 0.5.0 2025-03-15 [1] CRAN (R 4.5.0)
vctrs 0.6.5 2023-12-01 [1] CRAN (R 4.5.0)
vroom 1.6.6 2025-09-19 [1] CRAN (R 4.5.0)
withr 3.0.2 2024-10-28 [1] CRAN (R 4.5.0)
xfun 0.53 2025-08-19 [1] CRAN (R 4.5.0)
yaml 2.3.10 2024-07-26 [1] CRAN (R 4.5.0)
[1] /Library/Frameworks/R.framework/Versions/4.5-arm64/Resources/library
* ── Packages attached to the search path.
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────