Revision as of 16:14, 13 June 2023

Course Information

COM 682: Advanced Computational Communication Methods

Location: Discord

Class Hours: Tuesdays, 10–12 ET in the #General channel on Matrix

Instructor

Instructor: Jeremy Foote

Email: jdfoote@purdue.edu

Office Hours: By appointment

Course Overview and Learning Objectives

I teach an Intro to Programming and Data Science course that gives students an introduction to programming in Python, and some basic skills for gathering and analyzing data from the web.

There are many, many aspects of computational communication research that we don't cover in that class. This class is intended to take the next step in providing resources for students who want to do computational social science research.

That next step typically looks different for different students, depending on what they want to research. One goal of this class is to collect useful resources for learning many of the types of tools and methods that computational social scientists typically use. Some of these will be explicitly discussed in class, but others will not.

In particular, following conversation with group members, we will focus on a fairly deep dive into computational text analysis and reproducible workflows. Other topics will be more self-organized and self-directed.

I will consider this class a complete success if, at the end, every student can:

Understand in and engage in creating open, reproducible workflows for their academic research
Understand some of the key tools for doing computational text analysis, including topic models, word embeddings, machine learning classifiers, and LLMs
Learn how to identify the key texts, software libraries, and resources for learning a new computational method

Required resources and texts

Laptop

Most of this class will be asynchronous, so you will need access to your own laptop. It is assumed that you have experience running Python programs on your computer, and have the needed hardware and software to do so. We will mostly be using Jupyter Notebooks, but you are welcome to use other IDEs.

In order to participate in class, you will need to sign up for our Matrix "space", using this link. If you haven't used Matrix before (most people haven't), then you will need to create an account and choose software to use to connect to the room. The default software is Element, which I highly recommend using. There is a web version of Element, but I'd recommend downloading desktop and mobile apps at https://element.io/download.

Readings

We will be working together to identify readings, videos, and other resources

Course logistics

Note About This Syllabus

This is a very collaborative class, and one of the outcomes of the class will be to update this wiki with resources regarding topics of interest to class members. Therefore, the syllabus will be in flux both before the class and during the class.

Lectures

Our class time will follow a "flipped" classroom model. We will identify asynchronous materials (readings, recorded lectures, assignments, etc.) which you will work on before class and we will use our class time to review concepts, identify confusion, and synthesize.

Because we only have a few hours a week, some of this work will also happen outside of class, through conversations in the Matrix space.

For the first ~half of class we will all be working on the same topics. For the second half, we will use part of class time for students to present on either a topic or a project that they are working on.

Office hours and email

I will be traveling for much of June and part of July. I'm happy to make time to meet, but my schedule will be much less consistent than during a typical semester.
I am also available by email. You can reach me at jdfoote@purdue.edu.

Assignments

There will be two primary assignments for this class: one topic exploration, and a final project.

Topic exploration

One goal of this class is to learn to identify and assess tools, libraries, and learning resources. Each student will identify a topic that they would like to be in charge of. This will involve researching the conceptual background of the topic, identify and evaluating resources related to the topic, and sharing the best resources on this wiki page.

Ideally, this will include at least:

one or more academic papers about the topic
one or more Python libraries
one or more walkthroughs of using the method in Python (text and/or video)

Final Project

The final project can take one of two forms: either an explainer of a computational method or a research project.

Explainer

For the explainer, you will create a polished explainer of one of the topics covered in class. This should include an explanation of the topic/method, with a focus on how and why the topic is useful for doing social science research, as well as its limitations. You should also include an explanation of how to use the associated method, probably using public data.

Ideally, this explainer will fill a gap in the resources that already exist about a topic. One gap that often exists is resources specifically designed for communication scholars or social scientists more broadly. Or, resources to help more novice programmers, etc.

The format of this explainer is up to you, but it will likely at least include a Jupyter Notebook (or similar); I think that a video walkthrough of a notebook is also nice, but I'm open to other options.

Research Project

The other option for a final project is to push forward one of your quantitative research projects that would benefit from one or more computational methods. I strongly urge you to work on a project that will further your academic career outside of the class. There are many ways that this can happen. Some obvious options are to prepare a project that you can submit for publication, that you can use as pilot analysis that you can report in a grant or thesis proposal, and/or that fulfills a degree requirement. I prefer that you do projects on your own but it may be possible to work as a small team (maximum 3 people). Team projects are expected to be more ambitious than individual projects.

Planning Document

If you would like to take this option, you will submit a 2-4 page planning document of what you would like to do within the first few weeks of class, just to make sure we are on the same page about the scope of what you want to accomplish and so that I can give some initial feedback.

The project planning document is a basic shell/outline of an empirical quantitative research paper. The planning document should focus around three big questions:

Why are you planning to do this analysis? Make sure to introduce any background information about the topic, the community, your business, or anything else that will be required to properly contextualize your study.
How will you get the data to analyze? Describe the data sources will you collect and how they will be collected.
How will you analyze the data? Describe the visualizations, tables, or statistical tests that you will produce.

One approach that I have found helpful is outlined on this wiki page.

Project report

Final projects will likely look different, depending on the stage of the project when beginning the class. Typically, you will write a document or a Jupyter Notebook that will ideally provide the foundation for a high quality short research paper that you might revise and submit for publication. I do not expect the report to be ready for publication, but it should contain polished drafts of all the necessary components of a scholarly quantitative empirical research study. In terms of the structure, please see the page on the structure of a quantitative empirical research paper.

The great thing about a Jupyter Notebook is that it allows you to provide data, code, and any documentation sufficient to enable the replication of all analysis and visualizations. If you choose to write the report as a Word document, then you will need to include the code in a separate file.

Because the emphasis in this class is on methods and because I'm not an expert in each of your fields, I'm happy to assume that your paper, proposal, or thesis chapter has already established the relevance and significance of your study and has a comprehensive literature review, well-grounded conceptual approach, and compelling reason why this research is important. As a result, you do not need to focus on these elements of the work in your written submission. Instead, feel free to start with a brief summary of the purpose and importance of this research followed by an introduction of your research questions or hypotheses. If you provide more detail, that's fine, but I won't give you detailed feedback on these parts and they will not figure prominently in my assessment of the work.

Jupyter Notebooks do not have all of the tools for citations that Word or LaTeX or even Google Docs have, so while I expect you to cite related work your references section does not need to be as polished as citation management software would make it.

Grades

This course will follow a "self-assessment" philosophy. I am more interested in helping you to learn things that will be useful to you than in assigning grades. The university still requires grades, so you will be leading the evaluation of your work. At the beginning of the course, I will encourage you to think about and write down what you hope to get out of the course. Halfway throughthe course you will reflect on what you have accomplished thus far, how it has met, not met, or exceeded expectations, based both on rubrics and personal goals and objectives. At each of these stages you will receive feedback on your assessments. By the end of the semester, you should have a clear vision of your accomplishments and growth, which you will turn into a grade. As the instructor-of-record, I maintain the right to disagree with your assessment and alter grades as I see fit, but any time that I do this it will be accompanied by an explanation and discussion. These personal assessments, reflecting both honest and meaningful reflection of your work will be the most important factor in final grades.

I suggest that we use the following rubric in our assessment:

30%: class participation, including attendance, participation in discussions and group work
30%: topic exploration
40%: Final Project paper/Jupyter notebook.

My interpretation of grade levels (A, B, C, D/F) is the following:

A: Reflects work the exceeds expectations on multiple fronts and to a great degree. Students reaching this level of achievement will:

Do what it takes to learn the programming principles and techniques, including looking to outside sources if necessary.
Engage thoughtfully with an ambitious final project.
Take intellectual risks, offering interpretations based on synthesizing material and asking for feedback from peers.
Share work early allowing extra time for engagement with others.

B: Reflects strong work. Work at this level will be of consistently high quality. Students reaching this level of achievement will:

Be more safe or consistent than the work described above.
Ask meaningful questions of peers and engage them in fruitful discussion.
Exceed requirements, but in fairly straightforward ways
Compose complete and sufficiently detailed reflections.
Complete nearly all of the programming assignments at a high level

C: This reflects meeting the minimum expectations of the course. Students reaching this level of achievement will:

Turn in and complete required assignments on time.
Be collegial and continue discussion, through asking simple or limited questions.
Not complete assignments or turn some in in a hasty or incomplete manner.

D/F: These are reserved for cases in which students do not complete work or participate. Students may also be impeding the ability of others to learn.

Schedule

NOTE: This section will be modified throughout the course to meet the class's needs. Check back in often.

Each week will include the topic of the week. This is where we will gather and organize resources regarding each topic. Please be bold in editing this portion of the wiki to add or arrange resources.

Week 1: Welcome! (May 16)

Assignment Due:

None

Required Readings:

Agenda:

Class overview and expectations — We'll walk through this syllabus.
Make assignments for topic exploration

Slides:

Welcome slides

Week 2: Reproducible Research I (May 23)

Resources:

Paper: Wilson G, Bryan J, Cranston K, Kitzes J, Nederbragt L, Teal TK (2017). Good enough practices in scientific computing. PLoS Comput Biol 13(6): e1005510.
Paper: Gentzkow M, Shapiro JM. Code and Data for the Social Sciences: A Practitioner's Guide; 2014. https://web.stanford.edu/~gentzkow/research/CodeAndData.pdf.
Video: Reproducible Research: Concepts and Ideas. Roger Peng. YouTube

Slides:

Organization

Key ideas:

Folder structure
- Different options, but separate code from data
- Jeremy's approach:

my_cool_project
|
|-- README.md # Explanation of project and how to navigate it
|-- Snakefile # Or Makefile - workflow tool
|
|-- data/
|   |-- raw_data/
|   |-- processed_data/
|
|-- code/
|
|-- results/
|   |-- figures/
|
|-- papers/
|
|-- presentations/

Data Management

Key ideas:

Back up raw data
Keep raw data (and make it read-only)
Step one is to clean the data: create the data you wish you received
- Name variables well
- Use a tidy data structure
Share data (when possible)

Code management

Key ideas:

Version control
Don't repeat yourself (DRY)
Build at least a few high-level test cases

Week 3: Reproducible Research II (May 30)

Resources:

Blog post: Reproducible Workflows by Justin Kitzes
Introduction to Snakemake Tutorial (video)
An Introduction to Snakemake for social science
LaTeX introduction (video)
knitr introduction

Slides:

Reproducible analyses and papers

Key ideas:

Some big benefits (and some drawbacks) to using text-based tools (Markdown or LaTeX)
- Can be put in version control
- Tools like knitr can be used to put code directly into a document
Make figure creation part of your workflow, have documents point to your figures directory
Use citation management software that integrates with your document (use Zotero)

Sharing

Key ideas:

Share your code and data whenever possible!
Lots of options - OSF.io, Harvard Dataverse, etc.
Share preprints online

Advanced: Workflow Management

Key ideas:

Tools to reproduce as much of the workflow as possible
README file is much better than nothing
Even better is a "wrapper" script that runs everything
- Very clear exactly what is run and how
- Some fairly simple options:
  - Python file
  - GNU Make
  - Snakemake

Week 4: Computational text analysis: Introduction and Key Concepts (June 6)

Resources: Text As Data: A New Framework for Machine Learning and the Social Sciences (2022). Justin Grimmer, Margaret E. Roberts, and Brandon M. Stewart.

Read chapters 1-7

Week 5: Computational text analysis: Some "traditional" approaches (June 13)

Topic modeling

Embeddings

Resources:

Getting started with embeddings (Huggingface)

Classification

Semantic networks

NLP Tutorial Playlist Python (YouTube videos)

Week 6: Computational text analysis: using LLMs for research (June 20)

Due:

Final project proposal (details on Brightspace)

Resources:

[1] How ChatGPT Works Technically | ChatGPT Architecture

Agenda:

Discuss how things are working and a plan for the rest of the course.