The importance of communication in data science is often explained with a diagram like the one below:
The standard explanation says that communication, as the final step in the data science process, is crucial because it determines whether the rest of the work is ultimately used or wasted. To quote Hadley Wickham, "it doesn’t matter how great your analysis is unless you can explain it to others: you need to communicate your results."
In fact, communication is even more important than that. When data science is done well, communication happens at every stage of analysis. Before beginning a project, data scientists must communicate with decision-makers and domain experts to ensure that their analysis pursues the appropriate goal. While programming and analysing, data scientists must communicate with each other, with their managers, and with their future selves to ensure that their work is efficient, effective, and reproducible. The final stage, in which data scientists present their results to others, is of course crucial too. But communication is not just the last bottleneck. It can make or break the project at every stage.
This course prepares students for the many communication-related challenges they will face as professional data scientists. Students will use industry-standard software for data science communication. They will practice presenting data science concepts and results in various modes to various audiences. And they will learn to apply the principles of argumentation and reasoning that every good data scientist knows.
By the end of the course, students are expected to be able to:
- Productively and skeptically consume information about data science concepts and analyses.
- Design clear explanations of technical data science concepts.
- Write effectively about data science topics for a variety of audiences.
- Apply a framework that relates data science work to decision-making.
- Make calibrated claims about data science results.
- Reason about and apply principles of narrative when reporting on data science results.
- Design and deliver effective oral presentations of data science concepts and analyses.
| Day | Topic |
|---|---|
| lecture 1 | Overview of the course, and guidance on how to be a self-aware and skeptical consumer of communication in data science. |
| lecture 2 | Technical explanations. Effective use of analogies, diagrams, examples, and other representations of abstract (especially statistical) concepts. |
| lecture 3 | Writing for data science: clarity and flow at the level of the word, the sentence, the paragraph, and longer passages. |
| lecture 4 | Communication challenges and strategies at the beginning of a data science project. Understanding the client's needs. Knowing what to ask. Making plans and managing expectations. |
| lecture 5 | Communication challenges and strategies in the middle of a data science project: documentation; getting help; coordination, delegation, and leadership within a data science team. |
| lecture 6 | Communication challenges and strategies at the end of a data science project. Advising decision-makers. What to put, and what NOT to put, in a final report. Argumentation, persuasion, narrative, confident uncertainty. |
| lecture 7 | Speaking in data science: designing a presentation, managing information-density constraints, mitigating risks of being misunderstood, managing nerves. |
| lecture 8 | (Optional) Come practice your final presentation in front of an audience. |
| Deliverable | Grade | Description |
|---|---|---|
| Lab 1 - Critical consumption of a data science article | 17% | Read a data science blog post and reflect on the author's communication and argumentation. |
| Lab 2 - Blog post #1 | 17% | Write a blog post to explain a technical data science concept. |
| Lab 3 – Communication at the beginning and middle of a data science project (+ blog setup) | 17% | Choose an empirical question that can be answered with data from accessible and reliable sources. Make a prediction, and take detailed notes on your findings. |
| Lab 4 – Communication at the end of a data science project (Blog post #2 & presentation slides) | 17% | Write a blog post to report on your findings from lab 3. |
| Final presentation | 25% | A 5-minute presentation on the same topic that you wrote your second blog post on. See here for groups, times, rooms and order. |
| Portfolio | 7% | A series of activities and reflections, intended to be completed throughout the course during class time. |
All supplementary – no need to buy anything, unless you're interested.
- Writing: The Sense of Style, Steven Pinker (If you buy anything, buy this!)
- Tools & Technology: Chapters 26-30 of R for Data Science, Garrett Grolemund & Hadley Wickham
- Writing & Speaking: Houston, We Have a Narrative: Why Science Needs Story, Randy Olson
- Heuristics & Biases: Thinking, Fast and Slow, Daniel Kahneman
- Persuasion: Influence: The Psychology of Persuasion, Robert Cialdini
- Technical explanations: Better Explained, Kalid Azad
- Technical explanations: My favourite pedagogical principle: examples first!, Gowers's Weblog
- Writing: Nonfiction Writing Advice, Scott Alexander
- Speaking: The Cognitive Style of PowerPoint, Edward Tufte
- Speaking: How to Speak: Lecture Tips from Patrick Winston, Patrick Winston
- Writing: Writing (aka rewriting), The Effort Report (Elizabeth Matsui & Roger Peng)
- Biases & Heuristics: Cognitive Biases in Data Science, Drew Conway on This Week in Machine Learning and AI
- Writing & Speaking: Houston, We Have a Narrative: Why Science Needs Story, Randy Olson on New Books Network
I hope it goes without saying that I don't personally agree with everything said in every one of these resources. But they all contain good perspectives to expose yourself to.
-
How to be a self-aware and critical consumer of communication in data science contexts.
By the end of the lecture, students are expected to be able to:
- Explain the role and importance of communication and argumentation in data science.
- Notice when they are surprised, confused, impressed, or unphased by claims made about or with data.
- Consider the values of, and potential relationships between, relevant variables when someone else is describing their analysis.
-
How to explain technical data science concepts.
By the end of the lecture, students are expected to be able to:
- Understand the role and importance of concrete examples in explanations.
- Design the overall structure of an explanation by anticipating the audience's state of mind at every turn.
- Describe the strengths and weaknesses of the main ways in which data, models, and concepts can be represented: text, speech, diagrams, graphs, tables, code, and mathematical notation.
- Explain which combinations of representations are most useful for communicating with various audiences: decision-makers, stakeholders, domain experts, engineers, members of the public, other data scientists, and even oneself.
-
How to write clearly and elegantly about data science topics.
By the end of the lecture, students are expected to be able to:
- Explain considerations for communication that are unique to writing (as opposed to speaking)
- Craft clear sentences through the effective choice and use of words: avoid garden path sentences, apply the context-comment principle, know when and why to use the passive voice.
- Avoid common pitfalls when trying to achieve balance between uncertainty and clarity, particularly when writing about statistics.
- Solve problems of structure and flow in paragraphs and longer passages.
-
Communication at the beginning of a data science project.
By the end of the lecture, students are expected to be able to:
- Apply strategies for communicating with clients and domain experts at the beginning of a data science project.
- Understand considerations for a) when you have a specific question and want to answer it with data, versus b) when you have a bunch of data, and you want to know what to do with it.
- Avoid common errors when setting timelines and managing expectations for users and decision-makers.
- Understand counterfactual reasoning, and apply it to the process of choosing the right project.
-
Communication in the middle of a data science project.
By the end of the lecture, students are expected to be able to:
- Produce effective documentation for analysis, research, and code.
- Describe and apply strategies for translating between modes of representation, e.g. when translating mathematical notation from a research paper into code.
- Describe and manage the most prevalent challenges that data science teams face when communicating with each other: coordination, delegation, evaluation, leadership, tact.
-
Communication at the end of a data science project.
By the end of the lecture, students are expected to be able to:
- Apply strategies for advising decision-makers at the end of a data science project.
- Understand tradeoffs between good narrative and good reasoning.
- Confidently express uncertainty.
-
How to give effective presentations about data science topics.
By the end of the lecture, students are expected to be able to:
- Explain considerations for communication that are unique to speaking (as opposed to writing)
- Design and deliver an effective oral presentation.
- Reason about the use of visuals in an oral presentation: using text as a supplement rather than as a crutch; describing graphs and drawing insight from them.
- Engage an audience and avoid creating an environment of boredom or irrelevance.
During each lecture period, students will be given time to complete short activities that will reinforce the main insights from the lesson. These activities will usually be followed by short periods of discussion, in which students can share what they have produced. Not all students will have a chance to speak during every lecture, but every student should aim to share their work at least once during the course. At the end of the course, students will submit their full portfolio, which will be graded for completion.
Note: the portfolio does not need to be polished. It just needs to show evidence of consistent engagement with course material, and a sincere effort to improve communication skills.
All presentations will happen on October 3rd, in parallel sessions in both the morning and afternoon. Presentations are 3.5-5 minutes long, and the topic can be the content from either lab 2 or lab 4.