Using AI to fight corruption in the Brazilian government

talk I gave at Evidence2Action: Smart Government, on May 10th, 2017, San Francisco, CA

Transcript

1. Using AI to fight corruption in the Brazilian government Thiago Marzagão data scientist OPS’ deputy head
2. the last few years in Brazil...
3. the Fachin list (4/11) - 17 political parties - 24 out of 27 states - all former presidents - 3 governors - 8 ministers - 39 representatives - 24 senators - ...dozens of other bigshots
5. using AI to find corrupt employees
6. what the app looks at: - how did he/she got in? (political connections vs entrance exams) - owns company? - punished before? - position? - ... hundreds of other variables
7. using AI to predict procurement risk
8. what the app looks at: - campaign donations? - how many employees? - has been around for how long? - how many bids? - scope of activities?
9. cooperation w/ academia - how common are frauds in public procurement? - in 2014 alone: - 30,302 bidders w/ same phone numbers - 2,882 bidders w/ same addresses - 532,986 identical bids - 154,911 bids from same IP addresses - what’s the damage caused by frauds in public procurement? - (working on it) Claudio Ferraz, PUC-RJ Dimitri Szerman, PUC-RJ Frederico Finan, UC-Berkeley
10. @tmarzagao http://thiagomarzagao.com thiago.marzagao@cgu.gov.br