Ready for R (2025-06-30): High Level Explanations
Learn about communicating through high level explanations.
Welcome to the Weekly Ready for R mailing list! If you need Ready for R course info, it's here. Past newsletters are available here.
Hi Everyone,
Hope everyone is having an ok summer. Just thought I'd write up a few things that have happened in my life that all have to do with offering high level explanations.
Explainable Machine Learning of Wine Reviews
I was in Salt Lake City last month for one of my favorite smaller data science conferences: ASA's Symposium on Data Science and Statistics (SDSS). One of my favorite machine learning talks at SDSS 2025 was called "Interpretable Sentiment Analysis using the Attention-Based Multiple Instance Classification Model: An Application to Wine Reviews". by Jing Cao's group.
What I really liked about this talk is that it took a pretty sophisticated approach to sentiment analysis. Classifying a wine review as positive or negative sentiment can be done with packages such as tidytext. But by using an attention-based mechanism in their approach, they were able to identify the words that contributed to positive and negative reviews across all reviews (see image above for words that are associated with positive reviews). For example:
Descriptive words: words that are used to describe the sensory experience of drinking wine, such as gorgeous, beautiful, ethereal, exquisite, sumptuous, luxuriant, and glistening.
Approaches that give insight into why machine learning algorithms classify as one or the other are incredibly important. And I can see the applications of this method beyond wine reviews.
Communicating: Practice High Level Explanations
This is part of a short series about effective communication for data scientists. Check out the previous newsletter for the first part.
For five years, we taught a Data Analytics course that was half organizational behavior, and half data analysis. One of the exercises we did as part of this course was that the students were responsible for a presentation with their healthcare executive sponsors/stakeholders.
They had built an analytical model predicting hospital readmissions and had to explain the impact of the model to the healthcare executives who hd sponsored the project.
The students found the exercise challenging - but many also found success with it. I think doing this roleplay was very helpful for our students, in that it showed them that the last mile of analysis is effective communication.
The exercise was effective because we emphasized the need for change management - students were implementing a metric to predict hospital readmissions, and they thought carefully how to present it and how to use it to healthcare executives. They identified a particular way to present the pilot study data.
Communicating: Practice Summarizing
How do we get better at presenting to our stakeholders? We need to work on our summaries of our projects and work. This is an iterative process, which mean we need to get into a feedback loop and roleplay with friends, especially those who aren't in our field.
Specialization often leads to the curse of knowledge - we don't know what others don't know. We need people who will be honest when they don't understand what you're communicating to them. Otherwise, we won't get better.
Using short template formats for summarizing your work can be helpful, such as the Problem/solution/Impact framework:
- Problem: What was the problem you were trying to solve?
- Solution: How did you solve the problem?
- Impact: What was the impact of your solution? Did it help your stakeholder?
Practicing this with others will help you hone your approach.
Cascadia R Conference 2025
Just got back from Cascadia R Conference 2025, which was back in Portland this year at both Portland State University and Oregon Health & Science University. Thanks to all of the conference organizers - I had a great time teaching my Intermediate Shiny workshop and tasting donuts and rating them live with the Shiny app!
I taught the Intermediate Shiny workshop on Friday, and it was great to meet everyone and introduce them to run_with_themer() - which is the built in Shiny interface to modfiying themes in the bslib library.
Thanks for Reading!
Hope everyone's summer is going ok. If it isn't, please take care of yourself.
Best, Ted