Credit-Giving Norms in Online Subcultures

ISEF Category: Behavioral and Social Sciences

Subcategory: Sociology and Anthropology  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: Full Year

The Hook

A small shift in how people give credit can change who gets seen as the expert in an online group. You can measure that shift with public posts, manual coding, and text embeddings that map language patterns. This topic lets you study status, norms, and belonging in the same dataset.

What Is It?

Computational ethnography means you study an online community the way an ethnographer studies a neighborhood, but you also use code to spot patterns at scale. You read posts by hand, tag them for things like praise, attribution, self-promotion, correction, and mentorship, then compare those tags across time or across communities. That mix helps you see both the meaning of a message and the bigger pattern behind it.

Embeddings are a way to turn words into numbers so similar posts sit near each other in a map. If two messages both thank a creator, mention a source, or signal respect, they may cluster together even if the exact wording changes. Credit-giving norms are the unwritten rules about when people name a source, when they do not, and how the group rewards or ignores that choice.

Why This Is a Good Topic

This is a strong science fair topic because you can turn a social rule into data you can count. You can connect it to real issues like recognition, community trust, and who gets credit for ideas online. A student can learn coding, text analysis, and basic statistics without needing a wet lab.

Research Questions

• How does the rate of explicit credit-giving change across different kinds of posts?
• What is the effect of post format on whether users name a source or collaborator?
• Does moderator activity change how often members give credit in replies?
• To what extent do newcomers and long-time members differ in credit-giving behavior?
• Which language patterns predict whether a post will include attribution?
• How does community size relate to the density of credit-giving links between members?

Basic Materials

• Laptop or desktop computer with internet access.
• Spreadsheet software such as Google Sheets or Excel.
• Python installed locally or in Google Colab.
• A public post archive or manually collected dataset from one online community.
• A codebook template for manual tagging.

Advanced Materials

• University computing access for running larger embedding models.
• Qualitative coding software such as NVivo or MAXQDA.
• Secure data storage approved for human-subjects work.
• A network analysis package or server access for large graphs.
• Institutional review board guidance for public-data research.

Software & Tools

• Python: Cleans text, runs statistics, and builds embedding features.
• Jupyter Notebook: Keeps code, notes, and figures in one place.
• pandas: Organizes post-level coding and time-based summaries.
• scikit-learn: Supports clustering, classification, and basic model checks.
• Gephi: Visualizes community networks and credit-giving ties.

Experiment Steps

1. Define one community, one time window, and one clear claim about credit-giving.
2. Build a codebook that separates direct attribution, indirect praise, self-credit, and no credit.
3. Sample posts and replies so your dataset reflects the whole community, not just viral threads.
4. Turn the text into numbers with embeddings, then compare those clusters with your hand-coded labels.
5. Choose network measures and statistics that test whether credit flows through a few members or spreads across the group.
6. Decide how you will separate a real norm shift from platform noise or a one-time event.

Common Pitfalls

• Mixing different time periods, which makes platform drift look like a change in community norms.
• Treating praise as credit, which inflates attribution rates and blurs the coding scheme.
• Using only the most active users, which turns a few loud voices into the whole community.
• Running embeddings on too few posts, which creates clusters that look real but do not hold up.
• Ignoring deleted, edited, or quoted messages, which breaks the credit trail and weakens your analysis.

What Makes This Competitive

A stronger version of this project does more than count mentions. You compare at least two communities or two time windows, prove your coding is reliable, and test whether network structure matches the language patterns you find. If you add a careful control group and a simple model that predicts when credit appears, your project starts to look like real social science, not just a summary of posts.

Project Variations

• Compare a hobby subreddit with a language-learning Discord to see whether credit-giving works differently across platforms.
• Track how credit changes before and after a moderation rule shift or community guideline update.
• Focus on newcomer replies and test whether people give less, more, or different kinds of credit when they first join.

Learn More

• Pew Research Center: Search for reports on teens, social media, and online communities to ground your research question.
• MIT OpenCourseWare: Search for social network analysis and computational social science lecture notes.
• Data & Society Research Institute: Find reports on platform governance, online norms, and community moderation.
• Journal of Computer-Mediated Communication: Search for peer-reviewed articles on online identity, norms, and group behavior.
• Proceedings of the International AAAI Conference on Web and Social Media: Look for open research on online communities, language, and network structure.