The global spread of mobile phones has brought new opportunities to many poor people around the world, but an estimated 800 million have trouble with text entry or automated voice systems because they are illiterate or only partly literate. And training programs are difficult to get going at sufficient scale.
In Pakistan, researchers are using a silly voice game to motivate hundreds of thousands of people to master an automated voice system—and then move on to scroll job listings this way, too.
The effort, led by researchers at Carnegie Mellon University and Pakistan’s Lahore University of Management Sciences, started last May, when just five people in service jobs at Lahore University were given a phone number to get started playing the game. As of this week, the game has more than 156,000 users involving nearly 600,000 calls, including 27,000 inquiries to a job service, all of it sustained by viral spreading alone.
Umar Saif, a computer scientist at Lahore University who currently heads the Punjab Information Technology Board, an agency of a Pakistani province, and who co-led the research (see “Improving Connectivity in Poor Nations”), says viral spread is important to any technology that hopes to move beyond isolated pilot tests in the poor world. “This is the first virally spreading speech-based service in the developing world,” he says.
In the game, called Polly, a person calls the game’s phone number and the system calls right back, gives commands to record a message to a friend, plays it back with funny effects added, and allows the person to forward the goofy version to a friend, all in the native Urdu language.
Doing this means following voice and key commands: you are instructed to say something, after which you must press #. After hearing the funny effect, you press 1 to hear it again, 2 to send it to a friend, 3 to try another effect; to send the funny message to your friend, you must enter the friend’s phone number and press #.
Thus, the user has just been trained in the basics of using a customer service line or other automated voice system. And after hooking users, the researchers scanned Pakistani newspapers for jobs appropriate to low-skilled, low-literate workers, recorded them in Urdu, and made them available for audio-browsing. In just four months, users forwarded listings to friends 22,000 times.
“This platform is the first one that I know of that demonstrates social media can spread virally amongst low-income populations,” says Bill Thies, a researcher at Microsoft Research in India. “It shows ‘viral media’ can reach beyond the Internet. Even a rural farmer who lacks a smart phone, an Internet connection, and is non-literate can record and spread voice messages through a network of almost 100,000 people.”
A paper describing the work is scheduled to be presented at a conference next month in France. Updates on the project’s progress can be found here. The work is part of a trend in research aimed at using mobile phones as a training tool to improve literacy. Such work also helps people learn how to use phones, regardless of their literacy. Other work at Microsoft Research in India, for example, built a user interface based mostly on graphical representations and giving voice feedback; no text was involved. The system was used to supply job listings to domestic workers.
Following its success in Pakistan, Polly is now being launched in the Kannada language, which is used in southwest India. There, Polly will connect users to Babajob, an agency in Bangalore that specializes in low-skill jobs. The researchers are also looking for commercial, governmental, and nongovernmental partners for other applications, such as spreading public health messages.
“My long-term goal is to provide service to low-literate and illiterate people in developing countries,” says Roni Rosenfeld, a computer scientist at CMU who co-led the research with Saif. “But how do I reach these people and make them comfortable interacting with an automated system? This spreads and, at the same time, trains people.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
How to fix the internet
If we want online discourse to improve, we need to move beyond the big platforms.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.