CSL crowdsourcing algorithms aim to lift people from poverty

5/4/2016 August Schiess, CSL

Lav Varshney's team has built algorithms that more efficiently match employees to jobs that complement their talents.

Written by August Schiess, CSL

Crowdsourcing algorithms developed by a CSL research team are helping lift people out of poverty—one perfectly placed job at a time.

A team of CSL researchers, led by Assistant Professor Lav Varshney, has developed algorithms that help the nonprofit company Samasource efficiently connect individuals living below the poverty line with meaningful work.
A team of CSL researchers, led by Assistant Professor Lav Varshney, has developed algorithms that help the nonprofit company Samasource efficiently connect individuals living below the poverty line with meaningful work.
A team of CSL researchers, led by Assistant Professor Lav Varshney, has developed algorithms that help the nonprofit company Samasource efficiently connect individuals living below the poverty line with meaningful work.

In collaboration with Samasource, researchers led by CSL Assistant Professor Lav Varshney have built algorithms that more efficiently match employees to jobs that complement their talents. By more effectively allocating talent resources, the team stands to significantly increase Samasource’s ability to find meaningful work for those living in poverty.

Samasource is a nonprofit that connects people in countries like Kenya, Uganda, India, and Haiti to entry-level digital jobs that give them the experience to then pursue formal work, such as college and full-time positions. Since 2008, Samasource has employed 7,605 workers and increased their income by 3.7 times over four years.

Their customers, which include companies like Google, GettyImages, Microsoft, and TripAdvisor, have various digital needs, like tagging pictures or programming work. Samasource trains workers and then individually matches them to a job that fits their talents.

“We are trying to more efficiently match their talents and skillsets with job requirements,” said Avhishek Chatterjee, a postdoctoral scholar in Varshney’s group. “For example, a job might come along that requires knowledge of a certain programming language like C++. Should the job be given to anyone who knows C++? No—it should go to somebody who only knows C++, and then save another worker, with an additional specialized skill, for a job where they might be more useful later.”

This skillset sorting is difficult and time-consuming to do manually, but the team’s algorithms have automated the process and mathematically guaranteed that that workers’ skillsets are used in the optimal way. By perfectly utilizing worker capabilities, the team found that the work is done over 6 times faster.  

“Not only does the algorithm improve efficiency, but it also operates in a decentralized way so that people can choose what job they want to do rather than being controlled centrally. People like this freedom of choice,” said Varshney, an assistant professor of electrical and computer engineering.

Varshney and his team build algorithms that increase efficiency and reliability in many different types of systems, and they reached out to Samasource when it seemed like the team’s algorithms could be used effectively for Samasource’s platform.

“It feels good to know our algorithms are significantly minimizing delays and costs for Samasource and consequently finding many more people jobs than before,” said Chatterjee. “It’s a good cause and a great way to put our research to work.”

The team has also been exploring other applications, particularly in matching radiologists to patient cases. Each radiologist has specific specialties, and each patient has specific needs, so the algorithm could optimally match these two parties together as cases come in.

Chatterjee presented this work at the IEEE Infocom conference.  


Share this story

This story was published May 4, 2016.