Designing an Architecture for Delivering Mobile Information Services to the
Rural Developing World

Tapan S. Parikh
Department of Computer Science and Engineering
University of Washington, Box 352350, Seattle, WA 98195-2350
tapan@cs.washington.edu

Abstract

Paper plays a crucial role in many developing
world information practices. However, paper-based
records are inefficient, error-prone and difficult to
aggregate. Therefore we need to link paper with the
flexibility of online information systems. A mobile
phone is the perfect bridging device. Long battery life,
connectivity, solid-state memory, low price and
immediate utility make it better suited to developing
world conditions than a PC. However, mobile
software platforms are difficult to use, difficult to
develop for, and make the assumption of ubiquitous
connectivity. To address these limitations we present
CAM – a framework for developing mobile
applications for the rural developing world. CAM
applications are accessed by capturing barcodes using
the phone camera, or by entering numeric strings with
the keypad. Supporting minimal navigation, direct
linkage to paper practices and offline multimedia
interaction, CAM is uniquely adapted to rural user,
application and infrastructure constraints.

1. Introduction

Recently we have seen a growing interest in the
digital divide - the division between those people that
take advantage of digital tools and information services
for personal or professional purposes, and those that do
not. This has given rise to the optimistic notion that if
digitally disenfranchised people adopt information
technologies in a sustainable way, they can achieve
many development objectives, particularly in rural
areas of the developing world. Termed ICT for
Development, this vision carries a broad and pressing
mandate - allowing billions of people access to
services as important and varied as health care,
education, and financial and governmental services.
During our work with community microfinance
groups in rural India, we found that paper plays a
crucial role in many local information practices [2]. It
is used ubiquitously as a method of data storage,
exchange and establishment of trust between two
transacting parties. It is a medium that the local people
own and trust, and provides a greater sense of security
than rented or borrowed appliances. However, the
introduction of some technology is desirable. Not only
are paper-based processes inefficient, they are difficult
to aggregate and do not lend themselves to certain
kinds of analyses. This is a serious impediment to
information management in developing countries.
The mobile phone has been described as the most
likely modern digital device to support economic
development in developing nations [4]. As seen in the
example of Grameen Phone [1], if a mobile phone is
shared by a group of people, it can be affordable for
even the poorest communities. New mobile handset
models provide an open development platform and
significant computing capabilities. Given their
popularity with developing world populations, and the
immediate utility of voice communications,
smartphones represent an opportunity to bootstrap
computing in the developing world. Several features
(battery operation, solid-state memory, wireless
connectivity, affordable price) make it a better-suited
device for rural developing world conditions than a
conventional PC.
However, current mobile software applications are
difficult to use, difficult to develop, and often make the
assumption of ubiquitous connectivity. To foster an
information revolution in the rural developing world,
applications must be used by minimally educated
users, developed by minimally trained developers and
work in a variety of connectivity and power
environments. To address these requirements, we
present CAM - a framework for developing and
deploying mobile applications in the rural developing
world. Supporting minimal, paper-based navigation, a
simple scripted programming model and offline multi-
media interaction, CAM is uniquely adapted to rural
user, application and infrastructure constraints.

2. The CAM Framework

The driving element of the CAM architecture is a
mobile phone application called the CAMBrowser.
Users navigate within and between CAM applications
by capturing barcodes using the mobile phone camera,

or by entering numeric strings. Both can be printed
directly on paper forms for ready access. CAMForm
analogs of existing paper forms serve as offline clients
for CAM applications. Data is first entered on paper,
from where it is transcribed, validated and uploaded
using the CAMBrowser.



Figure 1. Part of a CAM-enabled loan application

3.1. Navigation

In an earlier experiment, we observed the
navigation difficulties encountered by semi-literate
rural users with a traditional WIMP interface [2].
These problems are exacerbated by the limited display
and input capabilities of a mobile phone. Even for
literate users, menu-based navigation is complicated,
requiring significant time to understand and convey.
Moreover, only a limited number of options can be
accessed.
To address these limitations, CAM applications and
application functions are indexed using numeric
strings, encoded either as barcodes to be captured via
the camera, or as numbers to be entered via the keypad.
Both are printed on paper forms. In this way navigation
is directly tied to a paper representation of the task.

3.2. Content

The CAMBrowser downloads and executes
applications written in an XML-based scripting
language. CAM provides an API for accessing the
mobile phone's user interface, networking and
multimedia capabilities. Figure 1 shows a loan
application designed for a microfinance institution. The
barcode in the top right, when clicked, activates a
function displaying a sequence of prompts for the user
to transcribe the values from the form. The prompts are
displayed in sequence rather then arranged spatially,
like in a web-based form. This is better suited to the
small screen of the mobile device. We call this wizard
interaction, as it resembles the task wizards used for
installing and configuring software applications. Users
just choose a high-level task and then follow the
prompts.
Each CAM prompt can be associated with arbitrary
audio and graphics. This increases the flexibility of the
system for dealing with unsupported languages or
semi-literate users. The audio contributes to the wizard
metaphor, making the interaction proceed like a
conversation between user and device. The device asks
a question, and the user answers.
After data has been entered, the user can review the
values before submitting them to the server. By
focusing the camera on a barcode associated with a
form field, the current value is displayed on the screen.
This reinforces the binding between the written values
on the form, and those entered on the phone. If a value
is not correct, the user can click on the barcode,
displaying a prompt to edit the value.

3.3. Networking

In many rural places, a wireless connection is not
available. The first time CAMBrowser encounters an
application, it attempts to download the code and data
from the server specified in the barcode or numeric
string. Using SMS, this request can be issued offline.
The SMS is cached in the phone's outgoing message
queue. The message will automatically be sent when
the phone is connected. When the server receives the
SMS, it sends back the appropriate XML code as a
MMS message attachment. The phone will
automatically download the message when it is
connected. The code for the application is cached on
the phone for offline use. The next time the phone
returns to the village, the application can be used.
Application data is stored in the phone’s local memory,
which serves as a cache of the server database. Data is
synchronized using get and put functions. This is also
done asynchronously with messages.

4. Evaluation and Current Work

The true test of CAM is the breadth and relevance of
the applications that can be developed. The first CAM
application is for data collection from microfinance
groups. This application’s usability was tested with
local field staff in Tamil Nadu, India [3]. Within three
days, users could process transaction receipts in an
average time of thirty seconds, with an error rate less
than 1%. We are starting a pilot deployment covering
400 groups and 24 staff.
We are designing and implementing several other
CAM applications. This includes remotely ordering
groceries from villages, accessing a patient’s electronic

medical history, and tracking the movement of goods
and people through a rural supply chain.

5. References

[1] Grameen Phone home page.
http://www.grameenphone.com, Mar. 2005.

[2] T. S. Parikh, K. Ghosh, and A. Chavan. Design
studies for a financial management system for micro-
credit groups in rural India. In Proc. CUU 2003, ACM
Press (2003), 15–22.

[3] T. S. Parikh, P. Javid, S. Kumar, K. Ghosh, and K.
Toyama. Mobile phones and paper documents:
Evaluating a new approach for capturing microfinance
data in rural India. In Proc. CHI 2006, ACM Press
(2006).

[4] The real digital divide. The Economist, Mar. 2005.