Hurricane Ian Destroyed Their Homes. Algorithms Sent Them Money
When storm Ian churned Florida in late September, leaving a trail of devastation caused by high winds and flooding. But a week after the storm passed, some residents in three of the hardest-hit counties saw an unexpected glimmer of hope.
Nearly 3,500 residents of Collier, Charlotte and Lee Counties received a push notification on their smartphones offering $700 cash assistance, no questions asked. A Google algorithm implemented in partnership with the nonprofit GiveDirectly estimated from satellite images that those people lived in heavily damaged neighborhoods and needed some help.
GiveDirectly is testing this new way of targeting emergency aid in partnership with Google.org, the search and advertising company’s charitable arm. The individuals offered the funds were users of a welfare app called Vendor that manages food stamp payments. Targeting messages with help from Google’s AI software allows GiveDirectly to deliver aid only to people living in Ian-ravaged areas faster than manually sorting through the reels users of the application.
This is the first time GiveDirectly has used this technology in the US, but before it tested a similar idea in Togo in the months following the pandemic that crippled the world economy. There, households are given aid based on signs of poverty detected by imaging algorithms from researchers at UC Berkeley, and clues from cell phone bills.
The Florida project is supported by a mapping engine called Delphi, developed by four Google machine learning experts who have been working with GiveDirectly for over six months, starting in late 2019. The software highlights communities in need after disasters like hurricanes by overlaying live maps of storm damage. with poverty data from sources including the US Centers for Disease Control and Prevention. Hurricane damage data is provided by another Google tool, called Skaiuses machine learning to analyze satellite images from before and after the disaster and estimate the severity of damage to buildings.
“You now have a map that shows where socioeconomically vulnerable and where has been damaged,” says Alex Diaz, head of AI for Social Good at Google.org. aid distribution. “
Skai’s damage assessment algorithms are trained by manually labeling satellite images of several hundred buildings in the disaster area known to have been damaged. The software can then quickly detect damaged buildings across the entire affected area. A research paper on underlying technology presented at the 2020 academic conference on AI for Disaster Response stated that automatically generated damage assessments are in line with human experts’ assessments with a high degree of accuracy. accuracy from 85 to 98%.
In Florida this month, GiveDirectly sent out a push notification offering $700 to any user of the Vendor app with a registered address in the neighborhoods of Collier, Charlotte, and Lee County, where the system Google’s AI thinks more than 50% of buildings have been damaged. So far, 900 people have accepted offers and half of them have been paid. If every recipient accepts GiveDirectly’s offer, the organization will pay $2.4 million in direct financial assistance.