Trustworthy Data Visualization
summary
Shows that trustworthy visualization is not a single image but a social pipeline: R and ggplot v4 map numbers to shapes reliably, gganimate produces an animation of daily sea surface temperatures computed from 17 million observations, and NOAA's bucket correction story explains a surprising midcentury plateau. Reproducibility tools like renv and targets and Quarto improve reliability, but fabricated data, LLMs that overcommit, and shifting data collection practices remain human problems software cannot fully solve.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:06Talk purpose
- 2:09Data visualization in data science
- 8:56Technical analysis in stock markets
- 12:26Storytelling dangers with graphs
- 15:17Human perception limits
- 18:55Reliability in visualization
- 23:10Visualization design commitments
- 27:07Social trust in data work
- 32:52Trust challenges in data practices
- 38:51Infrastructure for trustworthy data
- 43:32Visualization role in data science
- 45:43Question and answer session begins
- 46:05Custom ggplot themes
- 46:26Politicized data environments
- 47:32Visualization in hostile environments
- 48:39AI impact on reproducibility
- 50:28Accessibility vs reliability in AI
- 51:52Trusting AI for data problems
- 52:58Convincing leadership on axis scaling
- 53:45Closing remarks
related talks
31mFundamentals of Data Visualization
43mBehind the Scenes at Gapminder: A Story about Data and Visualizations
18mFrom idea to code to image: Creative data visualizations in R
15mThe Future of Data Visualization on the Web
40mData Visualization for Humans
10mThe Future of Data Visualization
05mThe Art of Data Visualization
01h 38mEffective Data Visualization