EDK All work

Computational media research · UC Santa Cruz

StimScapes

How can computational media support stimming rather than treating it as behavior to suppress?

In brief

Under Professor Kate Ringland and PhD mentor Yihe Wang at UC Santa Cruz, I worked with two other high-school interns to study neurodivergent-led media practices. We asked what online stimming content could teach designers about building computational tools that support—rather than suppress—self-regulation and sensory expression.

A quieter view

Calm Mode softens the page with slow color and shape.

Method overview

Research questions
Data collection
Hashtag network
Content analysis
Design framework
Figma prototype

Hashtag network workflow

Scrape TikTok73 creators
20,523 videos
Extract hashtags505 parsed terms
Co-occurrence score
Semantic score
Similarity matrix
Spectral clustering
Analyze results
Dense network visualization connecting hashtags found in neurodivergent stimming videos
01The hashtag co-occurrence network: 505 parsed hashtags connected by how often they appeared in the same videos.
Figma prototype flow for the StimScapes educational mobile game
02The Figma prototype translated the research into lessons, progress paths, achievements, and an interactive stimming activity.

Prototype walkthrough

What I did

  1. 01Collected 20,523 TikTok videos from 73 neurodivergent creators, then extracted and parsed 505 hashtags into a network of recurring themes.
  2. 02Compared hashtag co-occurrence with BERT-based semantic similarity, built pairwise similarity matrices, and used spectral clustering to optimize and evaluate thematic groups.
  3. 03Coded videos in Dovetail by stimming type, content, and portrayal method, then translated the quantitative and qualitative findings into design principles and a Figma game prototype.

Key finding

Common content centered on vestibular, proprioceptive, and tactile stimming; information and practical tools; and multisensory presentation through demonstration and music. Across the analysis, three design priorities recurred: self-acceptance, support for varied preferences, and clear educational context.