Data Driven Investigation in Defense of Human Rights
summary
Field shows concrete results and methods not theory A project systematically finds and verifies detention infrastructure in Xinjiang and publishes the dataset It runs an automated browser scraper at scale built a scraper with Puppeteer running for 1.5 months on 25 machines producing 50 million tiles 5 million masked then filters by transport infrastructure down to ~800k candidates and culminates in verified 348 detention camp locations using a rapid annotation tool Tinder-like verification UI and satellite + eyewitness corroboration
AI-written summary · about
chapters (26)
- 0:00Welcome
- 1:13About the speaker and their work
- 2:39Data-driven investigation
- 3:22Key investigations
- 5:50Syrian Archive challenges
- 8:17Current state of investigations
- 9:19Why it matters
- 10:44Required fields
- 11:50Xinjiang camps background
- 13:36Built to Last methodology
- 15:41Censorship on Baidu Maps
- 18:28Automating detection
- 20:11Filtering camp data
- 21:54Verification process
- 23:43Satellite imagery analysis
- 26:35Ground-level confirmation
- 28:48Detention camp scale
- 31:29Methods over tools
- 33:34Tool development values
- 36:01Human rights tech values
- 37:26Tech and identity reflections
- 39:10Transitional justice
- 40:34Collaborative tools
- 41:58Scaling challenges
- 42:20Secure data collection
- 43:43Social media data reliability
related talks
01h 15mStrengthening trust in computing technologies for detecting illegal deception online
40mWebscraping Responsibly for Social Science Research
34mHow Big Data can foster development
40mProtests, Pandemics, and Privacy: The Rise of Surveillance States
15mDefending Open Data with Intelligence
26mOpen Data for All: Social Action, Peace Tech & Community Outreach
01h 02mSecurity for Humans: Privacy and Coercion Resistant Design
31mHelp! There Are Humans in My Data!