WearITReadR
Software documentation for the WearITReadR Shiny app and R package.
Table of contents
- Table of contents
- What is WearITReadR
- Running WearITReadR in R
- The Shiny App
- Using WearITReadR with other data collection apps
- AI Study Generator
- Screenshot Scraper
- Examples
What is WearITReadR
The WearITReadR repo contains a Shiny app and open-source R package, currently in development. Functionality includes:
- A Wear-IT API pull (Wear-IT is a smartphone EMA data collection app)
- A customizable codebook, data dictionary, and study flowchart generator
- An AI study generator and modifier, which generates JSON following WearITReadR’s data description and storage spec
- A screenshot scraper, which walks a study block map to generate YAML that walks a study and scrapes screenshots (which can then be embedded inside a generated codebook)
- A Qualtrics translator, which translates from the Qualtrics API into our internal JSON storage format, then parses it — allowing all of the above to be generated from Qualtrics surveys
See the examples folder in the WearITReadR repo for more.
Accompanying the app is a JSON specification that acts as an initial draft of a data description standard for complex study designs — see inst/schema/0.1.0/study.json in the repo, and JSONSchema on this site for a walkthrough.
Running WearITReadR in R
devtools::install_github("ymirssbm/wearitreadr")
library(wearitreadr)
run_app()
See GettingStarted for full installation details.
The Shiny App
The Shiny app is the primary interface for WearITReadR. It connects to a study’s Wear-IT API and lets you generate:
- Codebooks, filtered and customized by topic or subset
- Data dictionaries, one per data file
- Study flowcharts describing the flow of participants and assessments through the study design
Using WearITReadR with other data collection apps
Although WearITReadR is designed primarily to interface with the Wear-IT API, it can be adapted to work with other data collection APIs as well.
As a proof of concept, we built a simple translator that converts studies run on Qualtrics into WearITReadR’s format via an API translator function. This lets any of WearITReadR’s functionality be used on data collected through Qualtrics. This translator was intentionally “vibe coded” — written quickly with the help of a generative AI assistant — to demonstrate that even researchers with limited technical background could build an API translator from their preferred data collection tool to the WearITReadR spec. The full process took only a couple of minutes.
Similar translators could, in principle, be built for any data collection API or software. Because WearITReadR is built around a JSON specification, it’s relatively easy to build and test new translators: JSON validators let both humans and generative AI formally check whether an ingested JSON follows the required specification. If it doesn’t, an informative error is thrown. This makes the process of building translators with AI assistance increasingly seamless, since the model can test whether, where, and how a translator is failing, and correct it.
Note: The study screenshot scraper and Qualtrics translator have not yet been implemented into the Shiny app. Both currently exist as standalone R functions. The screenshot scraper additionally requires Maestro Studio.
AI Study Generator
The AI study generator and modifier produces JSON following WearITReadR’s data description and storage specification. This allows a study design to be drafted, or an existing study JSON to be modified, using natural language instructions rather than hand-editing JSON directly. Generated JSON can be validated against the JSON schema to confirm it conforms to spec before use.
Screenshot Scraper
The screenshot scraper walks a study’s block map and generates YAML instructions to navigate the study, capturing a screenshot at each step. These screenshots can then be embedded directly inside a generated codebook, giving readers an exact visual record of the UI a participant saw — useful given how much small interface decisions (layout, slider behavior, labeling) can affect responses. See UIAndDelivery for why this matters.
The screenshot scraper requires Maestro Studio and is not yet integrated into the Shiny app.
Examples
- Codebook, generated from a real study — Recovery Community Center (RCC) Project
- Study flowchart for the RCC Project (see the Examples folder in the repo)
- Data dictionary, generated for an AI-generated example study — DataDictionary.csv
See the Examples folder in the repo for additional examples of AI study generation, the Qualtrics translator, and more.