CAM: A Mobile Interaction Framework for Digitizing Paper
Processes in the Developing World
Tapan S. Parikh
Department of Computer Science
University of Washington
Seattle, WA 98195-2350 USA
tapan@cs.washington.edu
ABSTRACT
During our work with community microfinance groups in ru-
ral India, we found that paper plays a crucial role in many
local information practices. However, paper-based record-
keeping can be inefficient, so we need to link paper with the
flexibility of modern information tools. A mobile phone can
be the perfect bridging device. Here we present the CAM
mobile document processing system, in which a camera phone
is used as an image capture and data entry device. Our sys-
tem is able to process paper forms containing CamShell pro-
grams - embedded instructions that are decoded from an elec-
tronic image. By combining 1) paper, 2) audio, 3) numeric
data entry, 4) narrative scripted execution and 5) asynchronous
connectivity, we have synthesized our experience into a sys-
tem that we believe is well-suited for an important set of
users and applications in the developing world.
Keywords: rural development, paper user interface, docu-
ment processing, visual codes, literacy
INTRODUCTION
Recently we have seen a growing interest in the topic of the
digital divide - between people that take advantage of modern
digital tools and information services for personal or profes-
sional purposes, and those that do not. This has given rise to
the optimistic notion that if digitally disenfranchised people
can adopt information technologies in a sustainable way, they
can achieve many development objectives, particularly in ru-
ral areas of the developing world. Termed Information and
Communication Technologies (ICTs) for Development, this
vision carries a broad and pressing mandate - allowing bil-
lions of people access to services as important and varied as
health care, education, financial and governmental services.
During our work with community microfinance groups in ru-
ral India, we found that paper plays a crucial role in many
local information practices [15]. 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
Copyright is held by the author/owner(s).
UIST’05, October 23-27, 2005, Seattle, Washington, USA.
Copyright 2005 ACM 1-59593-023-X/05/0010 ...$5.00.
Figure 1: Paper-based information processes are
ubiquitous in the developing world.
people own and trust, and provides a greater sense of security
than rented or borrowed appliances. Based on these observa-
tions, we conclude that paper must be retained as an integral
part of this information ecology.
However, the introduction of some technology is desirable.
Not only are purely paper-based processes inefficient, but
they do not easily lend themselves to certain kinds of anal-
ysis. This can be a serious impediment to information man-
agement in developing countries, particularly in the govern-
ment, financial and health sectors.
The mobile phone has been shown to be the most likely mod-
ern digital tool to support economic development in develop-
ing nations [16]. The growth of mobile phone use in China,
India and Africa resembles the growth of the Internet in the
developed world in the 1990s. As shown in the example of
Grameen Phone [1], if a mobile phone is shared by a group of
people, it can be afforded by even the poorest communities.
So far mobile phones have primarily only been used for com-
munications in developing countries. Recent improvements
in mobile phone functionality and processing power make it
a compelling device for piggy-backing other kinds of infor-
mation services to areas thus far untouched by information
technology.

In this paper we introduce mobile document processing as a
way to provide information services to the developing world.
In our system a camera-equipped mobile phone is used for
image capture and data entry. Document interaction is spec-
ified using CamShell - a document-embedded programming
language that allows developers to add interactive audio, data
entry, validation, processing, and networking instructions to
otherwise inert paper documents.
THE CAM USER INTERFACE
In contrast to other tangible UI programming toolkits [2, 9],
CAM is not an API for software developers to build different
physical user interfaces and applications. CAM is more like
the World Wide Web, in that it unites a user interface, a pro-
gramming language and a delivery mechanism for accessing
networked information services. Like the web, our goal is to
allow many different services to be developed and deployed
using a common infrastructure.
Since CAM uses standard Internet transport protocols for de-
livery (HTTP and SMTP), we focus on the user interface and
the programming language that drives the user interaction.
CamBrowser
The CAM client application is called the CamBrowser, and
has currently been implemented for Nokia Series 60 camera
phones. CamBrowser is designed to process specially de-
signed CamForm documents (see Figure 2). CamForms con-
tain visual codes - two-dimensional data glyphs containing
up to 76-bits of data that can be decoded from a camera im-
age [18]. Visual codes serve as references to the interactive
content embedded within CamForms.
Other researchers have presented a rich set of interactions
made possible by tilting, rotating and translating the phone’s
camera relative to an individual visual code [19]. However it
also has been observed that most of these interactions are not
accessible to novice users [22]. In CAM we have simplified
the interaction to two primitives. The user can scan codes
from low-resolution images taken in real time, or click codes
by taking a high-resolution image using the joystick button.
Our focus was on designing a simple and intuitive interaction
model with well-defined affordances between actions.
CamShell
CamShell is the programming language that is used to define
interaction with visual codes embedded in the form. There
is one visual code in every CamForm that serves as a form
identifier. This code has a 0 as the lowest-order bit. The next
4 bits identify the form to the CamBrowser application. This
information is used to load the form description schema, an
XML file containing metadata about the form and the un-
derlying data elements (including any default or pre-existing
values). The remaining bits specify the protocol that should
be used to load the schema file (currently HTTP, bluetooth or
the local filesystem), and the required information to find the
schema (a network address or a file name).
Visual codes with a lowest-order bit of 1 are action codes.
Each action code invokes a separate callback function when
it is clicked or scanned. These callbacks are mapped to a
set of code entry points specified as XML elements in the
form description schema. The XML elements contain the
executable code that should be invoked when the callback is
activated. The executable code itself is written using a simple
scripted programming language including support for func-
tion calls, control flow, arithmetic and basic datatypes [21].
The mapping to the appropriate callback is provided by the
next 7 bits of the action code. Static data is contained in
the remaining bits, which is sent to the callback function as
a parameter. There is a default function that is called when
a form is first loaded. This is used to perform any required
initial processing on the form.
CamShell provides an API for accessing the mobile phone’s
functionality - including the phone’s user interface, network-
ing and telephony features. Listed below are the main func-
tions currently included in this API:
• User Dialogs - These instructions launch various user di-
alogs to collect data, get confirmation or convey a message
to the user. Each dialog can be associated with audio and
graphical prompts. Audio feedback has been found very
useful by potential users.
• HTTP Post, Get - These instructions are used to make
HTTP requests using the phone’s built-in connectivity.
• SMS, MMS, Email - These instructions transmit asyn-
chronous network messages. In the case of MMS and
Email the current form image can be attached.
• Phone Call - This instruction is used to make a phone call
to a particular number.
• Applications - These instructions are used to launch other
applications, such as the Web or WAP browser.
Each callback function can contain any number of sequential
actions, some of which may be executed conditionally. Con-
ditionals are useful when performing form validation. An
example callback function is given below.
<function name="u_click" params="param1">
seq = input_int("Please input Form ID");
if (!seq) return false;
uri = "http://abc.com/reload.php?"
."seq=".seq;
return http_get(uri);
</function>
CamForms
Although there is nothing in the specification of visual codes
or CamShell that dictates how CamForms should be orga-
nized, there is a convention that we have been developing to
build a consistent and understandable metaphor for users.
An example CamForm is shown in Figure 2. This CamForm
is organized into the following three sections:
• Header - The header contains the form identifier code and
other codes containing static data (for example identify-
ing the sequence number of a particular form instance).
Header codes need to be clicked only once at the begin-
ning of a form interaction. A default form action can also
be included in the header that triggers all of the required
data entry tasks. This is useful for forms that have a single
execution path.

Figure 2: An example CamForm, showing the three
sections the user can interact with.
• Body - The body contains input codes that are like HTML
input tags. These are co-located with form fields, and
are used to transcribe data from the document using the
phone’s keypad. To edit a form field the user must click on
the input code, which brings up an editable dialog window.
When a set of input codes is clicked together in one image,
their respective dialogs are displayed in sequence. When
the CamBrowser scans a visual code, the value returned
by the associated scan callback function is shown on the
screen. This is usually the value of the field.
• Buttons - Buttons are codes identified by text and/or icons
that allow the user to perform various actions. Some ex-
ample actions include submitting the form data to a web
server, reloading the data from an online source or voiding
the currently stored data for the form.
USABILITY EVALUATION
We conducted an initial evaluation of a prior version of this
system [14] with microfinance staff and group members in
rural Tamil Nadu. This helped us identify the basic usabil-
ity issues, many of which have been rectified in the current
design. The users’ response to the system was very positive,
particularly when compared to our earlier experience design-
ing a PC interface for a similar user population [15].
We are now conducting a more detailed evaluation for simple
data entry tasks using both CAM and a web-based interface
on a PC. We are making this comparison because the alter-
native to CAM is that data is collected on paper forms in
the field and entered using a PC at the branch office. The
field staff waste time in travelling to the branch office for
this task. This reduces the number of clients served by each
staff (a benchmark for efficiency in the microfinance indus-
try), and takes time away from other tasks such as recruiting
new clients and developing the capacities of existing ones.
There is also a delay between when the data is collected and
when it is available for reference.
A mobile interface is preferred because it would allow field
staff to enter data as it is generated in the context of their reg-
ular documentation activities. Clients would also gain con-
fidence by observing this process, particularly when hear-
ing the audio confirmation that their transaction has been
recorded. Earlier efforts to automate microfinance data col-
lection using handheld devices faced problems in efficiency
and cost. If we can demonstrate that CAM has comparable
efficiency, accuracy and learnability to a PC-based system,
without a significant increase in overall cost, it could have
important ramifications for the microfinance industry.
Current mobile data entry interfaces are known to be inef-
ficient and difficult to learn. Much of this difficulty is due
to the limited screen space of mobile devices, which makes
navigation between different tasks and data fields difficult.
Our system addresses this issue by expanding navigation to
the physical domain of paper documents.
APPLICATIONS
We hypothesize that these benefits can also carry over to
other similar paper-based application domains in the devel-
oping world. We have met with banks, private companies,
NGOs and governmental agencies that have expressed inter-
est in using the CAM framework. Listed below are some of
the other applications that we are exploring:
• Sales and Distribution Tracking - CAM can be used to
track sales and distribution of inventory over a wide area.
• Mobile Cash Register - CAM can be used as an impro-
vised cash register for rural provision shops.
• Health Data Collection - CAM can be used to collect
health information from underserved rural populations. This
data could be useful for primary health clinics, NGOs and
government agencies.
• Transaction Processing - CAM can be used to securely
process remote financial transactions. This would allow
for the establishment of low-cost human-mediated banking
proxies. Several Indian banks are exploring this method of
service delivery using alternate technologies, including PC
kiosks and smartcards [13].
• Rural GIS - CAM can be used in conjunction with a
location-sensing technology such as GPS or PlaceLab [10]
to populate Geographic Information Systems, useful for
the government and other civil service organizations.
RELATED WORK
Other researchers have commented on the importance of pa-
per in workplace settings [20, 12], and sought to improve
coordination between paper and digital media [6, 4, 8, 3].
Companies have extended these efforts to develop commer-
cial document processing systems that recognize structure
and text from scanned paper forms and integrate with back
end data sources [5]. None of these systems have used a
mobile phone as the primary image capture and data entry
device, and none provide a general purpose networked pro-
gramming environment for specifying user interaction and
client processing.
The Mobile Server Toolkit described by Toye et al. [22] is
a visual-tag based system that allows mobile devices to ac-
cess site-specific information services. The MST transmits
individual UI events directly to an application running on

a bluetooth-connected server. CAM handles all user inter-
action on the phone itself - messages to the server contain
data elements only. CAM can also work offline and send
messages asynchronously using standard Internet protocols
(HTTP and SMTP). These features make it more suitable
for accessing remote information services without access to
other nearby devices and only limited connectivity.
The Cooltown project [7] at HP Labs uses RFID, barcodes
and and other sensing technologies to retrieve specific HTML
documents. The visual code widget library developed by
Rohs [17] includes a declarative UI specification that links
visual codes with a rich set of data entry widgets. Neither
of these systems support narrative interfaces consisting of
sequences of actions and conditional execution. Mobile UI
designers have recognized that sequential execution is more
suitable for devices with limited screen space [11].
CONCLUSION
In this paper we have described a mobile interaction frame-
work for bringing information services to the developing
world. By combining 1) paper, 2) audio, 3) numeric data en-
try, 4) narrative scripted execution and 5) asynchronous con-
nectivity via the medium of an increasingly ubiquitous mo-
bile device (the mobile phone), we have synthesized our ob-
servations into a usable and locally harmonious system that
we believe is well-suited for an important set of users and
applications in the developing world.
REFERENCES
1. Grameen Phone home page.
http://www.grameenphone.com, March 2005.
2. S. Greenberg and C. Fitchett. Phidgets: easy devel-
opment of physical interfaces through physical wid-
gets. In UIST ’01: Proceedings of the 14th annual
ACM symposium on User interface software and tech-
nology, pages 209–218, New York, NY, USA, 2001.
ACM Press.
3. F. Guimbretie`re. Paper augmented digital documents.
In UIST ’03: Proceedings of the 16th annual ACM
symposium on User interface software and technology,
pages 51–60, New York, NY, USA, 2003. ACM Press.
4. J. M. Heiner, S. E. Hudson, and K. Tanaka. Linking and
messaging from real paper in the paper pda. In UIST
’99: Proceedings of the 12th annual ACM symposium
on User interface software and technology, pages 179–
186, New York, NY, USA, 1999. ACM Press.
5. IFP success stories, June 2003.
http://www2.clearlake.ibm.com/GOV/ifp/success stories.html.
6. W. Johnson, H. Jellinek, J. Leigh Klotz, R. Rao, and
S. K. Card. Bridging the paper and electronic worlds:
the paper user interface. In CHI ’93: Proceedings of
the SIGCHI conference on Human factors in computing
systems, pages 507–512, New York, NY, USA, 1993.
ACM Press.
7. T. Kindberg. Implementing physical hyperlinks using
ubiquitous identifier resolution. In WWW ’02: Proceed-
ings of the 11th international conference on World Wide
Web, pages 191–199, New York, NY, USA, 2002. ACM
Press.
8. S. R. Klemmer, J. Graham, G. J. Wolff, and J. A. Lan-
day. Books with voices: paper transcripts as a physical
interface to oral histories. In CHI ’03: Proceedings of
the SIGCHI conference on Human factors in comput-
ing systems, pages 89–96, New York, NY, USA, 2003.
ACM Press.
9. S. R. Klemmer, J. Li, J. Lin, and J. A. Landay. Papier-
mache: toolkit support for tangible input. In CHI ’04:
Proceedings of the SIGCHI conference on Human fac-
tors in computing systems, pages 399–406, New York,
NY, USA, 2004. ACM Press.
10. A. LaMarca, Y. Chawathe, S. Consolvo, J. Hightower,
I. Smith, J. Scott, T. Sohn, J. Howard, J. Hughes,
F. Potter, J. Tabert, P. Powledge, G. Borriello, and
B. Schilit. Place lab: Device positioning using ra-
dio beacons in the wild. In Proceedings of PERVA-
SIVE 2005, Third International Conference on Perva-
sive Computing, Munich, Germany, 2005.
11. C. Lindholm, T. Keinonen, and H. Kiljander. Mobile
Usability: How Nokia Changed the Face of the Mobile
Phone. McGraw-Hill, NY, NY, USA, 2003.
12. W. E. MacKay. Is paper safer? the role of paper flight
strips in air traffic control. ACM Trans. Comput.-Hum.
Interact., 6(4):311–340, 1999.
13. T. S. Parikh. Rural microfinance service deliv-
ery: Gaps, inefficiencies and emerging solutions.
http://www.cs.washington.edu/homes/tapan/papers/mf-
challenges.pdf, April 2004.
14. T. S. Parikh. Using mobile phones for secure, dis-
tributed document processing in the developing world.
IEEE Pervasive Computing, 4(2):74–81, April 2005.
15. T. S. Parikh, K. Ghosh, and A. Chavan. Design stud-
ies for a financial management system for micro-credit
groups in rural india. In CUU ’03: Proceedings of the
2003 conference on Universal usability, pages 15–22,
New York, NY, USA, 2003. ACM Press.
16. The real digital divide. The Economist, Mar. 2005.
17. M. Rohs. Visual code widgets for marker-based interac-
tion. In IWSAWC’05: Proceedings of the 25th IEEE In-
ternational Conference on Distributed Computing Sys-
tems – Workshops (ICDCS 2005 Workshops), Colum-
bus, Ohio, USA, June 2005.
18. M. Rohs and B. Gfeller. Using camera-equipped mo-
bile phones for interacting with real-world objects. In
A. Ferscha, H. Hoertner, and G. Kotsis, editors, Ad-
vances in Pervasive Computing, pages 265–271. Aus-
trian Computing Society, Vienna, Austria, 2004.
19. M. Rohs and P. Zweifel. A conceptual framework for
camera phone-based interaction techniques. In Per-
vasive 2005, Third International Conference, Lecture
Notes in Computer Science (LNCS) No. 3468, Munich,
Germany, May 2005.
20. A. J. Sellen and R. H. Harper. The Myth of the Paperless
Office. MIT Press, Cambridge, MA, USA, 2003.
21. Simkin language, May 2005.
http://www.simkin.co.uk/.
22. E. Toye, R. Sharp, A. Madhavapeddy, and D. Scott. Us-
ing smart phones to access site-specific services. IEEE
Pervasive Computing, 4(2):60–66, April 2005.