Mosoko
People in developing regions have limited knowledge of the marketplace of goods and services beyond their immediate friends and neighbors. This limited knowledge contributes to inefficiencies in local markets, which, in turn, affect their larger economies. The central goal of this project is to build software and services that help create localized marketplaces in developing regions.
Two recent examples show how mobile phones can improve peoples' access to information in developing economies. Robert Jensen studied the effect of the growth in mobile phone use in fishing villages in the Indian state of Kerala. As cellular coverage grew from nothing to 100% from 1997 to 2000, the fish market became more efficient: fishermen knew where that days' catch would fetch a good price, price fluctuations between villages diminished, and fewer fish were discarded at the end of each day.
A second example shows how markets for agricultural goods -- accessed via phone -- aided farmers in East Africa. The Kenyan Agricultural Commodity Exchange makes nationwide prices available through text messages. Prior to the Exchange, the main source of pricing information was the middleman to whom the farmers were selling -- people who were motivated to buy the commodity as cheaply as possible. Armed with better pricing information, farmers can now sell their goods for prices closer to market rates.
As both examples show, providing people with better information helps build the "long tail" of developing markets. Like the Kenyan Exchange, we aim to develop long-running services that extend the flow of information one or two degrees beyond peoples' current friends and neighbors, improving the efficiency of developing markets.
Craigslist for the Next Billion
Billions of people have mobile phones but only a small fraction of those people have access to the Web. We are focusing on three main types of services that are widespread on the Web, but are absent in developing regions: classified advertising, social networks, and information sharing through Wikis. In all three service aspects, we seek to provide an intuitive and dynamic infrastructure for user-generated content.
Craigslist, in the US, and similar free classified services in other developed countries have created marketplaces for exchanging goods and services where none existed before. Their web-based access cannot be brought to developing countries in the near future. Instead, what if we could assist people in forming similar types of connections using only their mobile phones? Because this classifieds functionality is our initial primary focus, we call our project Mosoko, for Mobile Soko. In Swahili, soko is a local market where people go to sell their goods.
A few example scenarios:
- A carpenter posts that he is available to work tomorrow within 10km of his home village. Later in the day, jobs that match this description are posted and he receives a message describing them.
- A farmer posts that he wants to sell his tractor (or farm animals). This ad is delivered to locals who signal their interest before or after his ad is placed.
Carpenters in an area join a distribution list that acts as a kind of Small and Medium Enterprise (SME).
Short-term goals:
- Focus on one domain: apartment listings.
- Build a prototype and deploy it for a month in Nairobi for free.
- Review how the prototype was used.
- Add other market "topics" such as jobs, goods for sale, and roommates.
- Expand to other cities in East Africa.
Long-term goals
- Provide more powerful interfaces to users who do have Internet access. These interfaces will be both through standard computer web browsers and through customized applications that run on higher-end phones. Currently, the vast majority of phones in developing regions are not capable of running customized applications (e.g. Java).
- Use speech recognition to improve matching process.
- Work with Nokia's Consumer Insight to document what the Mosoko project learns about developing markets.
- Work with NRC groups in Africa and India to expand the project there, if trials prove successful.
Even as the barrier to Web access through mobile phones gradually fades, it remains unclear if the best modes for interacting with it will be through the tiny third screen. We plan to compare our work on audio interfaces to the visual Web as accessed through the mobile screen.
We are currently seeking Computer Science and Anthropology students to work on the Mosoko project.
Audio Wiki
In addition to currently working on the Mosoko Craiglist-like functionality, we at NRCC are collaborating with the Audio Wiki project at MIT CSAIL, lead by Prof. Saman Amarasinghe.
This project seeks to develop an "audio wiki" -- an open platform (like Wikipedia) that people can freely access and contribute to, but rather than using a desktop computer, they use a cell phone. This means that all content will be spoken rather than written, and there are a host of interesting challenges in user interfaces, speech recognition, and audio processing that need to be tackled.
The Audio Wiki project has a working prototype. This prototype allows users to post and edit movie reviews through an audio interface using their phones. Jonathan Birnbaum will improve the prototype's functionality for his MIT UROP in January 2008.
Audio Anthro
Using similar technology developed to support the Mosoko Marketplace and the Audio Wiki, we plan to build an audio-driven NPR Storycore-like application that helps people in remote regions record stories from their lives for free.
People
Nokia Research Cambridge
- Primary contact
- Smart matching of posts using voice recognition
- Harri Hansen
- Functionality and business uses.
- NRCC Lab Head
MIT CSAIL and Media Lab
Nathan Eagle (MIT Media Lab)
- On-site person in Nairobi
Saman Amarasinghe (MIT CSAIL)
- Audio Wiki Project Head
Bill Thies (MIT CSAIL)
- Audio Wiki Graduate Student
- Jonathan Birnbaum (MIT CSAIL)
- Audio Wiki Undergraduate Student
Harvard
- Web and Audio interface developer
NYU
- Marc Covitz
- Asterisk expert (taught Asterisk course at UoN)
East Africa
- David Kiania
- Asterisk and DID
- Saidi Apale
- Local developer
- Billy Odero
- Data collection (recent UoN graduate)
- Titus Mburu
- Data collection (recent JKUAT graduate)
Related Work
Companies and Organizations
Tellme's user-generated audio content Currently reviews of restaurants in limited areas in US.
Kijiji Web-based Classifieds in India (run by Ebay)
GrandCentral Lots of nice features for handling mobile phone messages.
Burrp Web-based Reviews of nightlife in Indian cities
Babajob Indian job seeking web site, currently exclusively in Bangalore. Model is that others post job wanted ads on the site for a friend or relative and then receive a small amount of money when that person is hired. Web-based searching and interaction.
Zunguka East African web and phone portal.
Projects in Developing Regions
MobilEd Initiative includes phone access to Wikipedia
Classes, Reading Groups, and Research Groups
News Articles
Status
Dec 3 2007
Setting up DID in Nairobi with contact who set up Nokia CareLine there
- Two UoN students bootstrapping apartment database
- Initial DB schema is complete
- 617 callbacks work
- mosoko.nrcc.noklab.com rack server is up
- Setting up weekly phone meetings with implementation group
- Subversion access set up and initial code imported
- Nathan leaving Nairobi Dec 12 - May 2008
Dec 17 2007
- Nairobi developers held second meeting to discuss design and assign tasks
- Users can post real estate apartment ads
- Tierney will develop web site in January
- Working on tie-in with Lynn Stein's UI course at Harvard
- Audio Wiki/Mosoko meeting this week
Jan 14 2008
- David Kiania is setting up a test machine in Nairobi.
- Basic posting and searching for apartments works.
- Matt Tierney will be working on the demo throughout January.
An MIT D-Lab course, SP 716, may have students doing projects related to Mosoko.
Feb 7 2008
- Still working on getting test machine up in Nairobi (certainly slowed some by conflict there).
- Demo machine in Cambridge up (usually, unless we're doing work on it right then): +1 617 453 2262.
- Matt Tierney finished with initial set of voice prompts.
Ledlie meeting up with folks on site later in the month and talking about project at DIS workshop in Cape Town.
Jun 20 2008
- Worked with MIT D-Lab course during Spring '08 to allow voice searchable postings.
- Student from course, Brian Omwenga, is interning with us over the summer.
- Taking second pass at the bulk of the code in order to make it more usable.
- Working through relations with ISPs and carriers on-site.