Semester Long Project 50%

The remainder of your grade will be determined through a semester-long project that has both primarily individual, but also some group components. The goal of this project is to expose you to how we create, uncover, document, and share culture as data. The final project is modeled on the Responsible Datasets in Context Project (RDC) https://www.responsible-datasets-in-context.com/, which was created to help students “work with data responsibly.” While we will be using these datasets to practice and learn how to programmatically work with data, they also provide an example of how best to curate and share data about complex cultural phenomena and objects.

As the authors of the project write in the mission statement:

“Data cannot be analyzed responsibly without deep knowledge of its social and historical context, provenance, and limitations. Anyone who works with data—from academic researchers to industry professionals—will know this claim to be true.

But despite its significance, social and historical knowledge and methodologies are one of the most neglected areas in undergraduate computing education. In classes, it is very common for students to use datasets that they find on websites like Kaggle, datasets that are poorly documented and that students thus don’t fully understand. This is a recipe for irresponsible data work and a bad habit that can become a dangerous habit as the stakes get higher.” 1

While you are not expected to create as polished or extensive of an output as the datasets available on the RDC Project, you will be working to create a first draft of what could eventually be part of this project.

Over the course of this semester, you will work with culture as data through two connected approaches:

  1. Creating culture as data: You will transform cultural materials into a custom dataset and document the interpretive decisions that shaped that process.
  2. Auditing existing culture as data: You will examine an existing dataset, archive, platform, API, collection, or source to understand how it already represents culture, what it makes visible, and what it leaves out.

Both approaches must be grounded in evidence. You will need to show the dataset itself, your computational process, your documentation, your GitHub history, and scholarly citations that help explain why your choices matter. AI tools can help with parts of this work, but they cannot replace the friction of making decisions, checking sources, testing code, documenting uncertainty, and explaining how your dataset came to be.

For your custom dataset, you must make visible at least two transformation steps: how cultural materials became a working representation and how that representation became structured data. For born-digital projects, the first step might involve selecting, saving, screenshotting, clipping, transcribing, or exporting materials. For analog projects, it might involve scanning, photographing, transcribing, or otherwise digitizing. The second step asks you to explain how those representations became fields, categories, values, labels, or records.

Click each section below for more detail about the two approaches your project will bring together.

Today, we are often handed datasets without ever experiencing the messy, interpretive process of creating one ourselves. This part of the assignment asks you to create culture as data. You will explore how complex cultural materials—say, a sculpture or a medieval manuscript—get transformed into structured data using computation. While this may sound straightforward, every step involves interpretive decisions that shape how others can engage with the data downstream.

For example, if you are interested in the history of children’s literature and working with a lesser-known author whose work has yet to be digitized, you could scan those materials and decide what to capture: Is it just the text? The illustrations? The paratextual material? You might use image models to extract drawings and text, enabling analysis of the author’s evolving style.

Alternatively, if you might be curious about how students are using AI in their daily lives and could turn to social media platforms like TikTok. While we cannot conduct human-subject research in this course (e.g., no surveys or interviews requiring IRB approval), you could analyze publicly posted videos. For instance, you might annotate TikToks that mention AI to track recurring themes or rhetorical patterns.

These are just a few possibilities, but whatever your focus, this approach requires you to define a clear methodology, justify your use of computational methods, and carefully document your decisions. You are not just collecting data. You are making data, and that comes with responsibility.

Not all responsible data work begins from scratch though. In fact, some of the most important contributions come from uncovering how existing datasets were made, what they omit, and how they might be improved. This part of the assignment asks you to audit existing culture as data. You will critically engage with an existing dataset—especially one that lacks documentation or transparency—and make its origins, structure, and limitations legible to others.

For example, you might find a movie dataset on Kaggle that includes scraped reviews and ratings, but provides little information about where the data came from or how it was cleaned. You could trace the dataset back to its source—say, IMDb or Rotten Tomatoes or even historic newspapers—and compare a sample of the original reviews to what appears in the dataset. Were the reviews shortened or misattributed? What patterns of exclusion or distortion emerge? How much has been lost in this process?

Or you might find a dataset that is widely reused but lacks key metadata or flattens the complexity of the phenomenon it represents, like popular economic benchmark datasets like GDP. Your goal then would be to try and reintroduce that complexity either through supplementing it with missing information, reconciling conflicting sources, or annotating entries to flag inconsistencies or bias. These contributions not only enhance the dataset, but also model what responsible reuse looks like.

Whatever dataset you choose, this approach requires you to investigate its provenance, reflect on its transformations, and document your findings. Just like with data creation, you’ll need a clear methodology, a justification for your interventions, and a well-documented data biography that makes your process transparent.

Your custom dataset and audited dataset should be connected by a shared cultural question, topic, source base, category, or comparison. You do not need to fully merge the two datasets, but you do need to explain how they speak to each other.

In some projects, that connection may involve joining datasets on shared fields like title, date, creator, place, genre, platform, or institution. In others, the most important finding may be that the datasets cannot be cleanly joined because they organize culture in incompatible ways.

Augmenting an existing dataset, such as adding missing metadata, annotations, corrected fields, or computationally derived features, is encouraged but not required.

At a Glance

Semester project milestones
Deliverable Moment Due Date Optional Extension Weight Main Question
Proposal: Collective Focus and Individual Plan September 15 September 22 Pass/Fail What are you making, why, and with whom?
Initial Custom Dataset and Audit Plan October 20 October 27 15% Have you created an initial custom dataset and a concrete plan for auditing existing culture as data?
Final Group Presentation December 8 None 5% What did your group learn, and how does the website communicate it?
Final Project Submission December 11 December 18 30% How strong, useful, documented, and reflective is the finished project?

A strong project does more than collect information. It shows how cultural materials become data and what is gained or lost in that process.

Your project should demonstrate:

  • Cultural complexity: Capture meaningful features of the cultural materials, not only the easiest categories.
  • Interpretive choices: Explain what you included, excluded, categorized, simplified, or left uncertain.
  • Visible transformations: Show how cultural materials became a working representation and how that representation became structured data.
  • Creative data design: Think carefully about what data you wish existed, not only what data is easiest to collect.
  • Meaningful computation: Use computation to inspect, audit, compare, clean, publish, or question your data.
  • Validation and limits: Check computational outputs and explain what your dataset cannot fully represent.
  • Scholarly relevance: Use published scholarship to justify your choices and show why your dataset matters.

There is no single correct kind of dataset for this project. The key is that your project should include both a custom-created dataset and an audit or comparison of existing culture as data. These two parts should speak to each other.

Examples of paired custom and audited datasets
Cultural Topic Creating Culture as Data Auditing Existing Culture as Data
Lesser-known children’s literature Scan or photograph selected pages, save them as working images or text, then create records for books, pages, illustrations, characters, themes, paratextual features, or visual styles. Decide what cultural details matter beyond title/author/date. Use OCR, image extraction, catalog metadata, or another computational method to compare your records with existing library metadata or to test what becomes visible at larger scale.
AI discourse on TikTok or other platforms Select and save public posts as screenshots, links, transcripts, or notes, then code themes, genres, rhetorical patterns, audio choices, visual cues, or claims about AI. Include uncertainty and context where needed. Scrape or collect platform metadata when appropriate, use text/image/audio tools to add checked metadata, or audit what platform metrics reveal and obscure about cultural meaning.
Movie reviews and ratings Select a close-reading sample of reviews, preserve source text or screenshots, then create records for tone, genre expectations, references, identity markers, or claims that standard ratings do not capture. Audit an existing movie dataset from Kaggle, IMDb, Rotten Tomatoes, Letterboxd, or another source. Trace provenance, compare records, and identify missing context, transformations, or distortions.
Sports analytics and media coverage Select and save social media posts about a team, player, league, rivalry, or event, then code themes, affect, memes, identity, controversy, or forms of fan expertise that box scores do not capture. Compare your custom dataset to historical newspaper coverage, sports-reference data, broadcast transcripts, or existing analytics datasets to ask what changes across media forms, time periods, and data sources.
Museum, archive, or digital collection Select objects, save images or catalog traces, then create custom records that capture interpretive details, uncertain descriptions, relationships, absences, or culturally specific context. Use API or catalog data to compare your records to institutional metadata. Audit classification systems, subject headings, dates, creators, descriptions, or gaps in the collection.

Milestone 1: Collective & Individual Topic Selection

Due Tuesday, September 15, 2026. Automatic extension available until Tuesday, September 22, 2026. Pass/Fail.

In the first two weeks of the course, you will be assigned to a group based on shared interests and complementary skill sets. Your first task is to collaboratively determine the overall thematic focus of your group, as well as brainstorm some potential topics for your individual dataset. You will have time to work on this planning document in class and will submit it as a Markdown file in your group’s GitHub repository.

This planning document should include:

  • Group Theme: What is your shared area of interest? What kinds of cultural objects, practices, or phenomena fall within this theme? What’s included and what’s beyond scope?
  • Individual Ideas: What dataset is each group member considering creating vs. auditing? You should detail both as much as you can here. You can post links or screenshots. You should also detail how your idea connects to the group theme. The core question for your dataset creation is: what cultural complexity are you trying to preserve or make visible? The core question for your dataset auditing is: how has cultural complexity been captured or erased in the dataset?
  • Scholarly Context: What published scholarship helps explain your topic, sources, categories, methods, or ethical concerns?
  • Collaboration Plan: How will your group communicate and share progress? What’s your GitHub organization strategy? How will you build the final group website and organize content?

Format & Submission

  • Markdown file (planning.md or similar) in the root of your GitHub repo or a planning/ subfolder
  • 500–750 words is sufficient, though you are welcome to be as creative and verbose as you like.
  • Use headings, bullet points, links, images, or tables if helpful
  • Be sure to include scholarly citations and explain how it relates to your project to receive full marks.

Milestone 2: Initial Custom Dataset and Audit Plan

Due Tuesday, October 20, 2026. Automatic extension available until Tuesday, October 27, 2026. 15%.

This is your first major deliverable for the semester-long project. For this check-in, you will submit your initial custom dataset and a structured plan for the audit/comparison work you will complete in the second half of the semester. In other words, you should have a substantial start on creating culture as data and a concrete plan for auditing existing culture as data.

The focus for your initial custom dataset is to understand what it means to make data carefully—item by item, decision by decision—while considering how computational methods can shape this work. You will create approximately 50-100 data items that reflect deep engagement with a particular cultural object, practice, or phenomenon.

Why start small? We start with small-scale, custom data creation for a crucial reason: this is where you learn that every dataset embeds interpretive choices. When you work carefully through 75 items, deciding what counts, what to capture, how to categorize, you experience the intellectual and ethical labor that gets hidden when you download a dataset or automate collection at scale. It is also an opportunity to get creative and think about what data you wish would exist for your cultural topic of interest.

What about computation? While your dataset is small and reflects close interpretive work, you are required to use computational tools to assist your process. This is not about automation—it is about understanding how computation can support even custom data work. You might use computation to structure records, extract text, create an HTML curation page, check consistency, compare categories, identify missing values, visualize early patterns, or use LLMs to suggest metadata that you then check manually.

You do not need to submit a completed audit at mid-semester. Instead, you need to identify the existing dataset, archive, platform, API, collection, or source you plan to audit or compare against your custom dataset. Your plan should explain why this existing data matters, how you will examine it, and how you think it might connect to the custom dataset you have created.

Format & Submission

Your submission includes six components:

Initial custom dataset and audit plan requirements
Component Requirement
Initial Custom Dataset Your custom dataset in a structured format of your choice, usually approximately 50-100 items. This is the first version of your work creating culture as data. You should have a rationale for how you organized your data.
Initial Documentation Your first attempt at documentation that explains the dataset, your process, your interpretive choices, relevant scholarly context, and how someone else should understand or reuse the data. This should document how your cultural materials became a working representation and how that representation became structured data.
Computational Process Evidence that you used computation to assist the custom dataset, such as a Python script, Jupyter notebook, HTML curation page, validation check, visualization, extraction workflow, or other relevant artifact.
Audit and Connection Plan A roadmap for the second half of the project: what existing dataset, collection, platform, API, archive, or source will you audit or compare against your custom dataset, and how will this help you understand existing culture as data?
Reflection Explain what is working, what is not working, what feedback you need, and whether the project needs to change direction.
Scholarly Citations Include at least 2-3 published scholarly sources that help explain your topic, cultural materials, categories, methods, or ethical concerns.

Your initial documentation should address questions such as:

  • What cultural materials are you working with and why?
  • How did those materials become a working representation, and how did that representation become structured data?
  • What published scholarship helps explain your topic, sources, categories, methods, or ethical concerns?
  • What computational tools did you use to assist your work? How did they help? What were their limitations?
  • What decisions did you make about what to include, exclude, or categorize? Why?
  • What challenges did you encounter, and how did you address them?
  • What patterns, questions, or tensions emerged from working closely with this data?

Your audit and connection plan should address questions such as:

  • What existing dataset, archive, platform, API, collection, or source will you audit or compare with your custom dataset?
  • What do you need to investigate about its provenance, structure, categories, omissions, or limits?
  • Can your custom dataset connect to the existing data through shared fields such as title, date, creator, place, genre, team, platform, or institution?
  • If the datasets cannot be cleanly joined, what does that failure reveal about how culture gets represented?
  • What computational methods might help you compare, audit, clean, visualize, or question the data?

Your initial dataset should be submitted in your group’s GitHub repository, and you should update any collective documentation needed to help users navigate the files and folders.

Milestone 3: Final Group Presentation

Tuesday, December 8, 2026. 5%.

During the final class meeting on December 8, each group will briefly present its website and collective principles. You do not need to present every individual dataset in detail. You should plan for a 10-15 minute presentation with 5 minutes for questions and feedback.

Focus on telling the story of what your group learned about your shared cultural data type, how you worked together, and how you made your labor visible. You should also explain how your website and GitHub repository are organized to help future researchers understand and reuse your work.

We will go over logistics closer to the date, but the focus here is on communicating what your group learned from building the website, developing collective principles, and making your labor visible.

Milestone 4: Final Project Submission

Due Friday, December 11, 2026. Optional extension available until Friday, December 18, 2026. 30%.

This is the culmination of your semester-long engagement with culture as data. By this point, you will have created a custom dataset with computational assistance, audited or compared it with existing culture as data, experimented with methods, and presented your work with your group. Now you will submit your complete dataset and documentation, along with writing that demonstrates what you have learned about the interpretive, technical, and ethical dimensions of representing culture as data.

The final submission should make both approaches legible: how you created culture as data and how you audited existing culture as data. Your documentation and reflection should tell the story of both approaches: how your custom dataset came to be, how you uncovered the history or limits of existing data, and how the two parts speak to each other. This means showing not only the final files, but also how your sources, categories, code, methods, revisions, citations, and uncertainties shaped what your data can and cannot say.

Your final submission includes both individual and collaborative components, all submitted through your group’s GitHub repository. The final group presentation is graded separately. The following criteria apply to the final project submission due on December 11.

Final Submission Priorities

Final project submission criteria
Component Individual or Collective Submission? Weight What You Submit / What It Values
Culture As Documentation & Reflection Individual 12% Documentation for future users and a process reflection that explains both approaches: how your custom dataset came to be, how you uncovered an existing dataset’s history or limits, what changed, what failed, what remains uncertain, and how others should understand or reuse the work.
Culture As Custom & Audited Data Individual 12% Final data files that include both your custom data creation and your audit, comparison, cleaning, or computational investigation of existing culture as data. These two parts should speak to each other through a shared question, source base, category, field, cultural object, community, or tension.
Collective Website and GitHub Repository Collective 6% A group website that presents your process and principles, plus a shared GitHub repository that is organized, documented, and reusable.

Each student should submit:

  • A final structured dataset.
  • A computational audit, comparison, cleaning, or transformation that connects to the custom dataset.
  • GitHub documentation that explains provenance, structure, categories, limitations, citation, and reuse conditions.
  • A process reflection that explains your choices, labor, revisions, failures, ethics, uncertainty, and use of computation.
  • A process page for the group website that explains how the dataset came to be.
  • Citations to published scholarship that help explain your topic, methods, categories, or interpretive choices.

Your final individual work should explain:

  • What cultural materials your dataset represents.
  • How you created, collected, audited, cleaned, or transformed the data.
  • How your custom-created dataset and audit/comparison shaped each other.
  • What evidence shows your process, such as source notes, commits, code, screenshots, annotations, data dictionaries, or revision history.
  • What the dataset helps us see, and what it hides, distorts, excludes, or leaves uncertain.
  • How AI or other computational tools were used and checked.
  • How your work connects to published scholarship and how that scholarship informed your choices.

Your group will collaboratively build a minimal-computing website, maintain a shared GitHub repository, and write collective principles that synthesize what you learned about working with your particular type of cultural data. This is not a repeat of the individual process reflections. It is methodological guidance and collective wisdom for future researchers.

Think of this as writing the documentation you wish had existed when you started this project. What should someone know before they attempt to represent music as data? Or social media? Or gaming culture? Or sports? What principles emerged from your group’s diverse approaches to similar materials?

Your GitHub repository and website should do different kinds of work.

GitHub repository and website roles
GitHub Repository Minimal-Computing Website
Documentation for future users and preservation. A story of how the datasets came to be.
Final data files, code, notebooks, schemas, licenses, and reuse guidance. Process narrative, choices, revisions, uncertainty, and labor.
Stable reference point for downloading, citing, and reusing the data. Public explanation of what your group learned by making the data.
Evidence needed to understand the dataset’s structure and provenance. Reflection on how cultural complexity was translated into data.

Each group should include:

  • A minimal-computing website that gathers the individual process pages.
  • A group overview explaining the shared theme.
  • A collective principles document about working with your cultural data type.
  • Links to the group GitHub repository and individual datasets.
  • Individual dataset folders for each group member.
  • Each student’s dataset, code, documentation, and reuse guidance.
  • A CONTRIBUTIONS.md file or clearly labeled section in the main README.md.

The collective website should help a reader understand why the individual datasets belong together and what your group learned by making them. The GitHub repository should help future users find, cite, preserve, and responsibly reuse the data.

Your group must make individual contributions visible. This is part of grading, but it is also part of the course’s larger question: how can we datafy a process without pretending that the data captures everything?

Your contribution log should include:

  • Who was responsible for each major website page, section, or technical task.
  • Who contributed to the collective principles/documentation.
  • Who reviewed, edited, or tested another group member’s dataset page.
  • Which GitHub commits, issues, pull requests, or file histories show those contributions.

Each student will also submit a brief individual contribution reflection that answers:

  • What did you contribute to the collective website or documentation?
  • What peer review, troubleshooting, coordination, or editing did you do?
  • Where is your work visible in GitHub history?
  • What labor became visible through commits, issues, reviews, or contribution logs?
  • What labor was harder to capture?

The collective website and documentation receive a group grade, but individual grades for the collective component may be adjusted if the contribution log, GitHub history, peer review, or reflections show substantially uneven participation.

Grade Flexibility

This project rewards thoughtful experimentation, not pretending that your first idea was perfect. If your project changes direction, document the pivot clearly. Strong work can grow from failed sources, broken scripts, messy data, difficult categories, and revised research questions.

If you receive a low mark on the Initial Custom Dataset and Audit Plan and your final work substantially demonstrates growth, you may request that the Milestone 2 grade be reconsidered. This is not automatic. To request this, contact the instructors before the final deadline with a brief explanation of how your final work addresses the earlier problems.

Footnotes

  1. Walsh, Melanie, Sylvia Fernandez, Miriam Posner, Anna Preus, and Amardeep Singh. “Mission.” Responsible Datasets in Context, n.d. https://www.responsible-datasets-in-context.com/mission.html.↩︎