Second Edition of Data Visualization Book Addresses LLMs and Automation

Kieran Healy has announced a forthcoming second edition of his book, Data Visualization: A Practical Introduction, expected to be published by Princeton University Press later this year. The revised edition builds on its predecessor with substantial updates to both content and code, reflecting major changes in the R programming ecosystem as well as the rapidly evolving landscape of data analysis and automation.

The book's draft is publicly accessible at https://socviz.co, with a sign-up form available for those interested in purchase notifications. The new edition brings all example code up to date with ggplot version 4 and R version 4.5 and newer, addressing warnings and compatibility issues that users of the first edition may encounter. Besides technical updates, the second half of the book has been extensively rewritten to incorporate significant advances in mapping with the sf package and extracting model results with the marginaleffects package.

Healy’s experience teaching the subject and feedback from users have influenced modest changes throughout, including a greater emphasis on data wrangling. Methodologically, the teaching approach has shifted toward a more pipeline-based workflow, encouraging best practices for reproducible and robust graphic design in R.

Crucially, the second edition also directly addresses the influence of large language models (LLMs) and code-generation agents in the analytic workflow. LLMs, such as OpenAI’s GPT models, are artificial intelligence systems capable of generating human-like code and text. Healy explores the growing question of whether these tools can or should replace manual, iterative coding in the field of data visualisation. While acknowledging the power and utility of code automation, he maintains the importance of building fundamental skills for identifying quality and errors—critical capabilities that pure automation cannot replace.

Quoting from the preface, Healy writes, "Large Language Models (LLMs) and coding agents are now part of the workflow of code generation and evaluation... But I also want you, the reader, to learn how to do good graphical work in a reproducible way. That means having a keen eye for quality and a good nose for error."

Healy emphasises that even with powerful new AI tools, foundational knowledge remains essential to discerning correct, effective, and ethical data analysis—much as practical know-how is needed to safely and correctly use any advanced tool.

Data Visualization: A Practical Introduction continues its mission as a hands-on entry point into the field, targeting both beginners and experienced users intent on refining their practice amid a changing technological environment.

r-bloggers.com{:target="_blank"}