Data for good: Lessons from the frontline
summary
DataKind UK describes concrete results: a WeFarm data dive processes 64,000 users / 240,000 messages, uses topic modeling and simple classifiers to detect question topics and route queries to likely answerers, producing a weekend prototype that reduces frustrated loops. A Centrepoint project gathers FOI returns, merges ONS covariates and runs multivariate regression to impute missing councils, moving an official 13,000 estimate to imputed 90,000 then revised 136,000, with open code, maps and documented assumptions. Lessons: scope problems tightly, keep stakeholders in the loop, and prioritize reproducible checks over cleverness.
AI-written summary · about
chapters (7)
- 0:00Data for Good intro
- 2:30Charity data science uniqueness
- 10:55WeFarm case study
- 21:14Centerpoint case study
- 34:30Key lessons from charity data
- 39:51Q&A on ethics
- 46:11Final thoughts
related talks
52mHow to Do Data for Good
24mData science, apps and civic responsibility
41mUsing Data Riches: A Tale of two projects
01h 26mSaturday Keynotes
16mThe importance of representative data
11mConnecting Researchers to Patient and Person Generated Data
23mBig data and big insights
46mData Discussion - AI for Good