Social Computing

Student Name: Siddharth Singh

Roll Number: 2014105

BTP report submitted in partial fulfillment of the requirements

for the Degree of B.Tech. in Computer Science & Engineering
on 16/11/2017

BTP Track: Research BTP Track

BTP Advisor
Dr. Ponnurangam Kumaraguru

Indraprastha Institute of Information Technology

New Delhi
 Students Declaration

I hereby declare that the work presented in the report entitled Social Computing submitted
by me for the partial fulfillment of the requirements for the degree of Bachelor of Technology in
Computer Science & Engineering at Indraprastha Institute of Information Technology, Delhi,
is an authentic record of my work carried out under guidance of Dr. Ponnurangam Ku-
maraguru. Due acknowledgments have been given in the report to all material used. This
work has not been submitted anywhere else for the reward of any other degree.

Place & Date: Indraprastha Institute of Information Technology, Delhi, 16/11/2017

Siddharth Singh

Certificate

This is to certify that the above statement made by the candidate is correct to the best of my
knowledge.

Place & Date: Indraprastha Institute of Information Technology, Delhi, 16/11/2017

Dr. Ponnurangam Kumaraguru

2
 Abstract

An emerging trend in developing regions is to give users access to certain Internet services for
free (or zero-rated) over cellular connections from phones. This can help bring the Internet
to users who otherwise could not afford to access it, and potentially attract future paying sub-
scribers. Free Basics is an initiative backed by Facebook to provide users in developing countries
free mobile Internet access to selected services. By partnering with telecom providers in some of
the developing countries, the Facebook FreeBasics program offers a set of free Internet services
in these regions, with the cost borne by telecom providers. The approach of providing a subset
of Internet services to users for free has generated tremendous debate in developing countries,
as there are questions about what is made free and at what (non-monetary) cost to users. That
said, most of the debates are qualitative or theoretical in nature. In this project, our goal is
to provide quantitative data, through careful experimentation, measurement and analysis, to
inform these debates.

Keywords: Freebasics, facebook, social computing

Acknowledgments

I would like to express my gratitude to my supervisor Dr. Ponnurangam Kumaraguru and

Dr. Rijurekha Sen, Post-doctoral researcher, MPI-SWS and Dr. Krishna Gummadi, Head, Net-
worked Systems Research Group, MPI-SWS for the useful comments, remarks and engagement
through the learning process of this B-Tech thesis. Furthermore I would like to thank Vedant
Nanda fellow researcher under Dr. Ponnurangam Kumaraguru for his the support along the way
in developing the service and doing the analysis. I would like to thank our loved ones, who
have supported me throughout entire process, both by keeping me harmonious and helping me
putting pieces together. I will be grateful forever to everyone who helped with this project.

i
Contents

0.1 Internet.org . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
0.2 Technical Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
0.3 Free Basics Proxy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
0.4 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1 Research Methodology 4
1.1 NewsBugle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 MyVoice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Rss Feed Entitiy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Data Collection Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Results 10
2.1 Medium of Access to NewsBugle . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Kind of data accessed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Web Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Number of services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.5 Topic of Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6 Temporal evolution of services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Characterizing the visitors 16

3.1 Visitor dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Visitor countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 Visitor interests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

4 Implications 19
4.1 Gatekeeping by Facebook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Data privacy from Facebook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Leveraging positives, if any . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

5 Awaaz:An ICTD Service 21

ii
 5.1 Service Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.2 Technical Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.3 Future Enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.4 User interactions so far . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

6 Conclusion 26

iii
Introduction

0.1 Internet.org

Internet.org [1] is a consortium founded and led by Facebook since 2013 with the goal of
bringing affordable Internet access to everyone in the world. Free Basics is the flagship initiative
by Internet.org, offering free access to select Web services in partnership with mobile (cellular)
service providers in developing and under-developed countries around the world. As of November
1st 2016, Free Basics has been deployed in 53 countries in Africa, Asia, and Central and South
America, with a total population of over 1.26 billion (22% of worlds population) [2] Compared
to developed countries, these countries have low Internet penetration rate (on average less than
20.4% of their population access the Internet) [3]. However, mobile phone usage rates in these
countries are very high (on average about 101.7%2) [4], making mobile phone users a ripe
target for expanding Internet usage. Free Basics, if successfully deployed worldwide, has the
potential to bridge the digital divide.However, data-driven studies of Free Basics are necessary
to help inform the public debates about Free Basics, which in early 2016, have led to Indian
telecommunication regulators blocking Free Basics to over a billion potential users [5] [6] .
Our study of Free Basics is motivated by the following high-level question : What is the scope
of Facebook FreeBasics platform in increasing web outreach in developing countries?
How does Free Basics work? Before we outline the research questions motivating our study,
it is important to understand the Free Basics architecture. As shown in figure 1, the Free Basics(
freebasics.com) service comprises three independent service providers:
(i) network service provider:The cellular carriers that agree to carry data for any Free Basics
service at no cost to the end user
(ii) Free Basics proxy service provider:All Free Basics traffic is routed via proxies that are
currently run by Facebook
(iii) web service providers:To have their services accessed by Free Basics users, web site operators
are required to first re-design their services following a set of technical requirements [7] and next
apply to have their service approved by the proxy service provider [8].
Any mobile subscriber of the participating network service providers can access (free of charge)
the list of approved web services by going to freebasics.com using their mobile browser or by
installing the Free Basics mobile application(while connected to their cellular providers net-
work).

0.2 Technical Guidelines

To analyze Free Basics ecosystem, we need to gather information (a) on how to deploy a service
on Free Basics, (b) on how does Facebook proxy works. Freebasics websites can only be accessed
through mobile devices that have a SIM registered to a service provider which is in partnership
with Free Basics. These restrictions make the access of freebasics content difficult to access
outside the walled garden. Facebook provides a free of cost developer platform to develop
services for it. Anyone can add their services on the platform as long as they by the participation

1
guidelines which exist to deliver performance on older phones and slower networks. Some of the
participation guidelines are mentioned below:
1. Efficiency : To efficiently deliver content to the users, the apps must not be data hungry
as it will be slow to load and might miss out on certain features of the app. Telecom providers
have made a huge investment to maintain the infrastructure to provide these set of services for
free. Therefore, the services should not use VoIP, video, file transfer, or photos larger that 200
KB.
2. Technical Specifications : Websites must be built to be optimized for browsing on both
feature and smartphones and in limited bandwidth scenarios. Here are some restrictions:
a. JavaScript:In order to offer an experience that can work across devices, the service must
work in the absence of JavaScript.
b. WOFF font types:If your service makes use of WOFF webfonts, ensure that the appearance
of your site is acceptable if these files are suppressed.
c. Iframes:Some suppression may occur for iframes so make sure that your content doesnt rely
on them. It is recommended to not to use iframes.
d. Video and large images: To comply with slower and weak networks, images and videos
must be truncated to 200 KB.
e. Flash and Java applets:Free Basics does not support Flash resources or Java applets.
f. HTTPS support: Freebasics supports end to end encryptions i.e user to proxy server and
proxy server to your server. Even if your service doesnt supports HTTPS, it still encrypts
user-proxy server connection.
Such restrictions are put in place to support the target population of mobile users in developing
regions, where mobile devices may not have full Web browsers and must access the Internet via
WAP or similar technologies [9] .
Any mobile subscriber of the participating network service providers can access (free of charge)
the list of approved web services by going to freebasics.com using their mobile browser or by
installing the Free Basics mobile application [10] [ (while connected to their cellular providers
network).

0.3 Free Basics Proxy

Since the requests passes through a proxy, the source IP always be a Facebook proxy server, the
actual source IP i.e. a freebasics user IP can be found in the HTTP header i.e. X-IORG-FBS-
UIP, the same IP can also be found in X-FORWARDED-FOR field to trace the actual user.
The platform also supports cookies which promotes user engagement. According to Facebook,
the User-Agent string is also not altered by the proxy. Below is the architecture diagram of
freebasics.

2
 Figure 1: Freebasics Architecture

0.4 Testing

Facebook also provides developers with Testing emulators to test their websites on wide variety
of devices to see if the site loads properly on all devices. It provides a Chrome Device Emulator
and Opera Emulator to test the mobile experience.

3
Chapter 1

Research Methodology

To get an insiders view of the application process and do fine-grained platform research and
user behavior analysis at the web server, we hosted our own web service named NewsBugle
in the Freebasics Ecosystem. We develop a suite of measurement techniques to improve the
transparency of Free Basics, to understand its scope and outreach in the context of developing
countries. Independent of the publicity of the program by Facebook and concerns raised by op-
ponents of the program, we aim to do an empirical audit as third party independent researchers.
We seek answers to the following key questions covering different aspects of the program
Free Basics services: What services constitute the current walled garden of Free Basics?
Are these services same across countries? Are these services growing over time?
Free Basics users: How many visitors does a typical service get, and from which countries
and demographic backgrounds?
We answer these questions using the following measurement methodology. First, we implement
an Android app to scalably collect Free Basics service lists across countries and over time. We
recruit participants in 15 countries across Asia, Africa, South and Central America, who collect
these service lists using our Android app. Second, we deploy our own Free Basics services
and collect empirical data to understand Facebooks approval and deployment process, and to
characterize users who visit our services.
Our measurements provide insight into Free Basics developer portal, users and content providers.
Specifically, we highlight the following key findings:
There are currently 200-450 Free Basics services across countries. Most of them are country-
specific, and they have grown by 100-150 new services over the last nine months.
Services need to be submitted to Facebook to be included in Free Basics. But this gate-
keeping process by Facebook is fast and mostly checks for technical conformity of the content.
Our experience while deploying our own Free Basics services has been highly positive.
One of our Free Basics services Bugle News, an RSS news feed aggregator offered in English,
Spanish and French, attracted 49.2K unique visitors from 60+ countries between September
2016 and October 2017. We characterize the nationality, demographics and interests of these

4
visitors.

5
1.1 NewsBugle

We built an RSS aggregator service called Bugle News that fetches RSS feeds from news orga-
nizations including BBC, CNN, and Reuters, and provides users with corresponding headlines.
The news stories are organized by topic and country. Users using Bugle News would see these
news snippets aggregated from different sources, categorized into eight high level topics (World
News, Africa, Asia, America, Science, Sports, Movies&Entertainment and Jobs&Career). There
were additional lower level topics for nine countries: South Africa, Zambia and Nigeria within
Africa; Philippines, Bangladesh and Pakistan within Asia; and Colombia, Peru and Mexico
within America; and two specific sports: Cricket and Football within Sports. The service was
offered in English between September 17th and December 15, 2016 and has been available in
English, French and Spanish since December 16, 2016.

Figure 1.1: newsbugle.mpi-sws.org

6
1.2 MyVoice

We Designed and developed another web service (MyVoice). We aim at providing users of this
application a platform to report daily problems in their life and surroundings like lack of hygiene,
no water supply, bad transportation, no electricity etc. This app lets a user post the complete
detail of the problem along with a picture and the address of the problem location just by the
click of a button. The user can also see the problems posted by other users using the same
platform. The goal of this application is to get all the issues in one place so that they can be
sorted out by the concerned agencies of the region.

Figure 1.2: myvoice.mpi-sws.org

7
1.3 Deployment

This Web application has been deployed in MPI-SWS(Max Plank Institute for Software Systems,
GermanyMPI-SWS(Max Plank Institute for Software Systems, Germany) and built on Django
Framework running on apache. We are also using redis-server to automate the addition of new
feeds into the system and authenticates the suggestions given by the user.
Figure 1.3: Server

1.4 Rss Feed Entitiy

We are using RSS(Really Simple Syndication) feeds to populate our database of news feed. RSS
feeds are found in XML format. We made a XML parser to parse the data and add it to the
database. Each news entity contains several fields but the ones that we use use are: title, source,
time, image and a summary of the news content. Below is an example entity of a RSS data
Figure 1.4: Rss Feed Entity

8
1.5 Data Collection Methodology

We built an Android app, that opens the Free Basics web- site and crawls through the service lists
in available lan- guages. We recruited participants in fifteen countries - five in Asia (Bangladesh,
Indonesia, Pakistan, Philippines and Thailand), three in South and Central America (Colombia,
Mexico and Panama) and seven in Africa (Benin, Kenya, Niger, Nigeria, South Africa, Tanzania
and Zambia). Each participant got some remuneration to cover the expenses of the Free Basics
SIM card in their respective country and as a service charge to run our Android app and send
us the col- lected service lists over their normal Internet connection.

9
Chapter 2

Results

2.1 Medium of Access to NewsBugle

Figure 2.1: Active User Data

The above data shows that majority of users that came to the website were accessing free basics
from a browser on the feature phone, the percentage of android users is very low. Hence, the
claim that facebook makes that it is empowering the parts of society which never had an access
to Internet can be somewhat seen from this result because a person using an android smart
phone is highly expected to be buying data packs for Internet usage to get the full usage of
a smart phone. Hence there are a lot of first time users of Internet who are using our service
freely using free basics. There are two more important things that I would like to point out by
referring the above graph.
1. As you can see in the graph, on 23rd September there has been an exponential dip in the
number of active user which is pretty unusual, on counter checking with the free basics emulator,

10
we saw that the whole platform i.e. Free Basics, experienced a huge downtime on that date.
We counter checked with the graph of another service i.e. learnbasics and it also reported a
sudden decline in the number of users. A question here arises on how reliable is the facebooks
infrastructure because such kind of downtime has never been seen on facebook.com and are they
using the right amount of resources to make this platform available to the people.
2.On 9th November, there was a exponential rise in the active user data most likely due to the
US Elections. This was later confirmed by our analysis on the traffic of that day which conveyed
that America was the most popular link clicked from each continent on our service when usually
the source continent feeds used to top the charts.

2.2 Kind of data accessed

(a) Links clciked in Africa (b) Links clicked in Asia

(c) Links clicked in South America

Figure 2.2: Continent-wise links clicked data

Inferring Figure 3.2 gives us the insight that users are highly interested in localized news content
which is limited to their continent, country and the popular local sport that is played in their

11
country. Another analogy that is a point of interest is their knowledge about whats happening
on the global level, though they may be interested in localized content but from the previous
section we saw that they knew about the US Elections which shows us their knowledge on
geographical regions, continents and global political events.

2.3 Web Interactions

Figure 2.3: User Responses

Looking at the background of the type of people who are actually using our service, we get an in-
triguing question, whether these people are well conversant in English and with web interactions
like browsing and clicking?
We divided all the user suggestions into 4 categories: 1. Spam : People trying/experimenting
with forms for the first time.
2. Curious about Internet : People who know something about Internet but are incorrectly
relating to the service that we are offering. For example, suggestions like YouTube, WhatsApp,
Google etc were received in the user suggestions. We asked them to enter RSS feed links that
they would like to see but they entered name of services which they might have heard of and
wanted us to host it for them on freebasics.
3. Legit suggestions: proper user suggestions that helped us add new sources for the feeds.
4.Security Leak : People entering their phone numbers in the text boxes unknowingly. Since
these people are new to elements of Internet, we dont expect them to fill HTML forms correctly
too. This can be inferred also from the fact that 55% of user suggestions were actually spam.
The fact that people entered their name and email address too even when it was not mandatory
and entering web-address instead of email-address in the email text box makes our point stronger
that these set of people are particularly naive when it comes to differentiating between UI
elements.

12
2.4 Number of services

Figure 2.4: Number of Free Basics services in different countries offered in English, French and local
languages

400
English/French
350 Local language
Number of services
300
250
200
150
100
50
0
Ba Be Co Ind Ke M N N P P P S T T Z
ng n lom on ny exic iger iger akis ana hilip outh anza haila amb
lad in bia esia a o ia tan ma pin A n nd ia
es
h es frica ia

Country

Figure 3.4 shows the number of services offered in different countries. Benin and Niger have
services offered only in French. Kenya, South Africa, Tanzania and Zambia have services offered
only in English. The rest of the countries have services offered in two languages: English
and a local language like Spanish in Mexico, Bahasa in Indonesia and Kiswahili in Tanzania.
Bangladesh and Nigeria have the highest number of services, exceeding 200, while the median
is around 100 services across the countries. Local lanaguage offerings are as many or more than
English offerings, except for Tanzania.
Figure 2 shows number of countries where a particular service is offered. A small percentage

Figure 2.5: Overlap of Free Basics services across countries

100
Free Basics services

10
Percentage of

0.1
1 2 3 4 5 6 7 8 9 101112131415
Number of countries where offered

of services are offered in more than one country. AccuWeather, BBC News, Bing, Baby Center,
Girl Effect, Thesaurus,Wattpad and Wikipedia are some services offered in all fifteen countries,
so this is probably a common list of services that the Free Basics pro- gram in any country starts
with. But 80% of the services are offered only in one or two countries. This observation and
the high number of offerings in local languages, indicate that local web service developers in a
country are using the Free Basics platform to offer services customized for their country in their
local language.

13
 Figure 2.6: Topic categorization of Free Basics services in different countries.

100 Communication
Learning

Free Basics services

80 Sports&Entertainment
News&Weather
Percentage of
Women&Girl
60 Buy&Sell
Job
40 Finance
Health&Safety
Transportation
20 Others
0
Ba

Be

In

Ke sia

Pa

Pa an

Ph a

So ine

Ta fric

Th nia

Za nd
ol

ig

ig
do a

ex

nz

ai

m
ng

ni

ny

ki

na

ilip

ut
om

er

er
ic
ne

st

la
n

bi
ha

a
a
la

m
ia

p
o

a
bi
d
es

a
h

Country

2.5 Topic of Services

Figure 3.7 show the proportion of Free Basics services in different countries categorized into
ten topics. The Free Ba- sics landing page lists these ten topics and lists 2-3 services under
each topic heading. The rest of the services are shown as a long list, without any topic catego-
rization. We manually categorized all the services offered in English (and in French for Benin
and Niger), using the service name and the one line service description. Some services did not
match any of the ten topics and have been listed as Others. Learning, News&Weather and
Health&Safety are the dominant topics in most countries. An interesting observation is, other
than in categories Communication (which has some instant messaging services and social net-
working service like Facebook) and Buy&Sell, almost all services are textual web- sites giving
information on the concerned topic. This is true even for topics like Sports&Entertainment,
where more than interactive games and live audio or video streaming, there are news articles in
sports and entertainment categories. Thus the web services are dominantly textual, providing
news and information on a variety of topics.

2.6 Temporal evolution of services

Figure 2.7: Histogram of Free Basics services and topics over time in Bangladesh (BD), Pakistan (PK)
and South Africa (SA).

100 Communication
Learning
Free Basics services

80 Sports&Entertainment
News&Weather
Percentage of

Women&Girl
60 Buy&Sell
Job
40 Finance
Health&Safety
Transportation
20 Others

0
Ba

Be

In

Ke

Pa

Pa

Ph

So

Ta

Th

Za
ol

ig

ig
do

ex

nz

ai

m
ng

ni

ny

ki

na

ilip

ut
om

er

er
ic
ne

st

la
n

bi
ha

an
a
la

m
ia

pi
o

an

nd

a
bi

si
de

fri
ne

ia
a

ca
sh

Country

Figure 2.6 shows the number and type of English-language services offered in Bangladesh (BD),
Pakistan (PK) and South Africa (SA) over time. The x-axis shows the country and time, while
the y-axis shows the number of listings. Between the nine months of May 2016 to January

14
2017, Bangladesh has added 179 new services, while South Africa has added 113 and Pakistan
95 during the same period. In summary, the contents of the Free Basics walled gar- den vary
depending on the country (but not on the ISP in a country), are growing rapidly, and are
dominated by in- formation services such as education, news, and health. We further find that
only a small number of services are offered in more than one country. The deployed services are
highly customized in each country, with 80% of the observed services offered only in one country
(and 20-30% offered only in local languages).

15
Chapter 3

Characterizing the visitors

In this section, we use our deployed services to characterize the Free Basics users who visit our
sites in terms of where the users are located, how they interact with our sites, and their demo-
graphic information. We leverage the fact that Free Basics users visit our site via a Facebook
proxy server that sets the x forwarded for HTTP header that indicates the mobile IP address
of the Free Basics user. This allows us to geolocate the users to a country based on the cellular
providers registered address. We associate visited content with unique users by setting browser
cookies that are unique to each user. Last, we use the Facebook developer portal analytics
demographic information that anonymously reports statistics about visitors age and gender.

3.1 Visitor dynamics

Figure 3.1: Daily Bugle News visitors, showing the impact of a Free Basics outage (left box), the US
election (middle box), and deployments in multiple languages (right box).

1600
Number of Bugle News

1400
visitors per day

1200
1000
800
600
400
200
0
0 15 30 45 60 75 90 105 120
Days since Bugle News deployment
on Sep 17, 2016
visitors overall
visitors using browser
visitors using android app

The number of daily Learn Basics visitors range from 300-500, while those for Bugle News
range between 800- 1,200. Although we do not have evidence to identify why Bugle News is
more popular, we posit that rank on the service listing (Learn Basics appears lower in the
listing alphabetically) and content relevance (news is more popular than educational reference
material) are contributing factors. Given the relative popularity of Bugle News, we focus only
on this service in the subsequent analysis. Figure 3.1 shows the daily Bugle News visitors since
deployment along y-axis with time along the x-axis. We con- firmed that the number of visitors

16
reported from Facebooks developer portal match the same recorded at our server. We see the
number of requests from the Free Basics app is small and relatively constant over time, and
the vast majority of visitors use a mobile web browser to access our service. We use rectangles
to highlight interesting events in Figure 3.1. The left box shows a Free Basics service outage
resulting in a sharp drop in traffic in September3. There is a sharp spike on the day after the
result of the US presidential election in early November (the middle box). In the right box,
we highlight sustained increases in number of visitors after the launch of French and Spanish
versions of Bugle News in December. The drops in visitors correspond to widely celebrated
holidays on Dec 24th, 25th and 31st. Figure 3.2 shows an empirical cumulative distribution
function (CDF) of the number of days each user visits the site. The x-axis is the number of
days, x, a user visits, with the y-axis representing the fraction of users who have visited a site at
most x days. We see that a majority of users are transient, with 78% of the users visiting Bugle
News only once and 14% visit over two days. Thus, user retention of Bugle News has been low.
Fairly high number of daily visitors (Figure 3.1) and low user retention (Figure 3.2) means a
large fraction of hits are from unique visitors. We counted 49.2K unique visitors to Bugle News
overall. Interestingly, even with zero promo- tional efforts, a reasonably practical Free Basics
service can see tens of thousands of visitors in a short period. This gives a ballpark lower bound
on the Free Basics population size, as only a subset of them would visit Bugle News (given many
competing news services including BBC itself).

Figure 3.2: Number of days each visitor uses Bugle news

100
Cumulative %
of visitors

95
90
85
80
75
0 5 10 15 20 25 30 35
Number of days using Bugle News

3.2 Visitor countries

Figure 3.3: Proportion of countries among Bugle News visitors, before and after adding French and
Spanish languages.

1000
Bugle News visitors per day

Sep 17 - Dec 15, Bugle News only in English

Dec 16 - Jan 5, Bugle News in English, Spanish and French
100
Average number of

10

0.1

0.01
MX
NG
PH
PK
IQ
ZA
TH
BD
ID
LR
PE
MW
CO
ZM
CD
RW
KE
TZ
GH
SN
AO
GT
JM
GN
SV
PA
MM
KH
TT
CV
BJ
MN
KW
MV
SR
MZ
SC
GW
CW
BB
NE
MR
MG
GA
CG
unknown

Country code

Figure 3.3 shows the visitor countries along x-axis and the percentage of total visitors from the
country along y-axis (log scale). We divide the time since Bugle News deployment into two

17
periods: the service initially offered only in English (3 months) and later in English, Spanish
and French (3 weeks). Bugle News has received visitors from 46 countries so far. Mexico,
Nigeria, Philippines, Pakistan, Iraq, Zambia, Thailand and Bangladesh are some of the countries
with highest visitor proportions. After introducing French and Spanish support, countries like
Benin, Niger, Mauritania, Madagascar, Gabon and Re- public of Congo have increased visitor
proportions. We can attribute this to the colonial French past of these African countries, making
French one of the official languages there. Some countries show a decrease in visitors after
introducing the new languages, compared to only English. This is possibly due to the smaller
measurement window of three weeks with the new languages, so that no visitors from those
coun- tries came within that window.

3.3 Visitor interests

Figure 3.4: Proportion of 18 news topics accessed from 8 countries. Bugle News has specialized news
feeds for these countries.

100 Entertainment SouthAfrica

Football Zambia
Percentage of requests

90
Cricket Nigeria
from the country

80
Sport Philippines
70 Science Bangladesh
60 America Pakistan
50 Asia Peru
40 Africa Mexico
30 World Career
20
10
0
M

Pe

Pa

Ba

Ph

Za

So
ig
ex

m
ru

ki

ng

ilip

ut
er
ic

st

bi

hA
la

ia
pe
o

an

a
de

fri
ne

ca
sh

Country

Figure 3.4 shows the proportion of users (along y-axis) in eight countries (along x-axis) requesting
specific news feeds on Bugle News. The patterns in the bars denote the 18 different news feeds
that Bugle News currently offers. World, Entertainment and Career are some of the popular
categories across all countries. Cricket is popular in Pakistan and Bangladesh while Football
(i.e., soccer) is popular in the American and African countries. Finally, all countries show.

18
Chapter 4

Implications

The Free Basics program has grown substantially, but has also created controversy with strong
opposition from proponents of an open Internet. In this section, we discuss the implications of
our measurements in the context of these concerns and discuss possible ways to leverage the
positive side of this platform.

4.1 Gatekeeping by Facebook

One of the main points of opposition of the Free Basics program has been why do services
need Facebooks approval to get enlisted in the Free Basics program? . In our experience of
deploying Free Basics services, Facebooks feedback has been strictly technical. Our overall
experience has been very positive, as discussed in Section 4. One point to note is both our
services lack controversial content. Bugle News aggregates news and Learn Basics provides text
books. There have been reports expressing censor- ship concerns about the program, and also
about governments trying to exploit Facebooks role as the gatekeeper for surveillance. Thus it
is not clear whether our positive experience with Facebooks gate-keeping can be generalized to
all services.

4.2 Data privacy from Facebook

A point of concern against the Free Basics program is all data will flow through Facebooks
proxies . Facebook is open about its Free Basics architecture, that requests will flow through
its proxies. It states its privacy policy as: a dual certificate policy to encrypt traffic separately
between the mobile client and Facebook proxy and again between the proxy and the web server.
Traffic will be decrypted at proxies, to only inspect domain names and traffic volumes, and store
any cookies in an encrypted and unreadable for- mat. . Our deployment experience validates
Facebooks advertised architecture of a proxy network. Requests came to our web servers from
IP addresses that we could map to Facebook data centers, using traceroute/whois utilities . An

19
interesting observation we made in the context of privacy was the age and gender demographic
information given at the Facebooks developers portal for both Bugle News and Learn Basics.
This is useful information for any web service developer, to monitor its user base. But how
Face- book computes these numbers, though neither the Free Ba- sics website nor our services
explicitly ask for user information, is an open question. Some transparency in this regard will
be good to ensure users knowledge about collecting and sharing information about them.

4.3 Leveraging positives, if any

In the presence of all the concerns about Free Basics, one might ask whether it has any positive
aspect that can be leveraged. Analyzing the currently deployed Free Basics services (Section 3),
we find some excellent services on health, education, social awareness, news and other topics.
These provide a wide range of useful information ranging from how to combat Zika, Malaria and
other viruses, to sharing stories of woman empowerment, to providing free textbooks and online
tutorials. Many services are offered only in a specific country with locally relevant information,
often local languages. Such services aimed at information dissemination often do not need data
privacy from Facebook. They can be simple textual sites, without a need for advanced features
like Javascript or rich multimedia that are not allowed for Free Basics websites. Such services
can also be deployed in a way that their content consumes small amounts of data and is relatively
insensitive to bandwidth caps, which we found Free Basics services to be subjected to in our
prior work . We found that even a simple RSS feed aggregator service on Free Basics can get
49.2K unique visitors within 4 months from 45+ developing countries, as we experienced with
Bugle News. Thus creative thinking on how to harness this platform with services that are
not affected by the down- sides of the program might help advance many research objectives
targeting a Free Basics user population.

20
Chapter 5

Awaaz:An ICTD Service

To leverage the Free Basics platform and reach out to its unique user population, we launched
our third web service called Awaaz: my voice in Free Basics. We submitted this service to
Facebook on March 5, 2017 and it became live on March 13, 2017. In this section, we discuss the
goals and design for this service and user inputs obtained so far, to highlight how Free Basics
can be used to create rich user generated content in the developing regions.

5.1 Service Goals

The goal of launching Awaaz is crowd-sourcing infor- mation about civic issues in developing
countries. There are many such examples of citizen journalism mobile apps and web services
[1, 7], where the user generated inputs are use- ful to understand pain points and possibly
connect issues to relevant civic authorities. We have seen good hits for our preliminary services
Buglenews (a news aggregator service) and Learnbasics (a textbook aggregator service). With
that confidence we launch Awaaz, which is our first service to actually create some valuable
content, harnessing the Free Basics platform.

5.2 Technical Design

To support lower end feature phones prevalent in the developing countries, Facebook imposes
certain technical restrictions in Free Basics web services. For example, Javascript or automatic
location detection in the mobile browser (needs HTML5 support) cannot be used. Incorporating
third party modules within Free Basics services is also difficult, if the third party service is not
part of Free Basics, causing users to be charged for accessing it. For example, showing a Google
Map to take map-based user input is not feasible, as that needs Javascript support and Google
content is not included in Free Basics. This walled garden of Free Basics, along with the technical
restrictions, make the technical design of Awaaz non-trivial. Figure 5.1 shows a screenshot of the
issue reporting page of Awaaz. As civic issue reporting has to be associated with a geographic

21
location to have any meaning or potential of redressal, we use a drop-down list of municipal
blocks in a particular city that users can select from. They can optionally give the full address
of the location that has the issue, manually typing it in a text box. Users can select from seven
categories of issues from a second drop-down list: roads, transportation, electricity, water, health
and sanitation, school and education and others. They need to describe the issue in a text box
and solve a captcha, before they can click the Report button. They can also optionally upload
a small image (size restriction imposed by Facebook), that demonstrates the reported problem.
When users visit the Awaaz web service for the first time, they need to enter their name, city
and optionally their phone number. The backend server associates a browser cookie, and stores
these information for the user. This allows to load the appropriate drop-down list of municipal
blocks for the particular city, on subsequent visits by the same user. Awaaz also includes a link
where users can see issues reported so far, categorized by topics or by location. Each issue in
this feed has an option to add comments, so that users can reinforce, negate or discuss issues
reported by others

5.3 Future Enhancements

To evaluate users interactions with the UI, we have launched Awaaz only for 10 most populous
cities in South Africa (SA). The official language of SA is English, which helps us in prototyping
and seeing user interactions more easily. Some enhancements planned for the service in near
future are:(i) identify and correct UI design issues, (ii) launch the service in more cities in SA and
gradually to other Free Basics countries, and (iii) add support for more languages. Potentially,
this can create a unique opportunity to compare civic issues across 60+ developing countries.
The collected user interactions data will also be valuable from an HCI research point of view.
Finally, connecting the service to the relevant urban authorities will be needed, as done by
similar services in the past so that the reported issues have a potential of getting redressed.

5.4 User interactions so far

In one and a half months of Awaaz deployment in SA, with support for ten cities, we have seen
334 unique visitors who came to our website using Free Basics. 106 unique users among the
visitors registered with their name, city and phone numbers in the landing page. A total of 9
issues have been reported (Figure 5.2). These include civic issues in the categories of roads and
electricity and also reports of some serious criminal offences. The level of English comprehension
has been better compared to a similar form we have for Buglenews URI suggestions. For example,
even without any check, all entered phone numbers are in the appropriate format. This can either
be due to English being the official language in SA, or due to the seriousness of the people who
come to read and report civic issues unlike more casual visitors who come to read news snippets.
The average time spent by a user in the landing page to register is x minutes and to report an
issue is y minutes. To understand UI issues in more detail, we plan to do a user survey in near

22
future using the registered phone numbers.

23
Figure 5.1: Screenshot of Awaaz issue reporting page. Users can select a municipality in South Africa
from a drop-down list, a category for issues from another drop-down list, give a title and description of
the issue, give a full address for the issue (optional), upload a small image (optional), solve a captcha
and submit the issue.

24
Figure 5.2: Screenshot of Awaaz issue reporting page. Users can select a municipality in South Africa
from a drop-down list, a category for issues from another drop-down list, give a title and description of
the issue, give a full address for the issue (optional), upload a small image (optional), solve a captcha
and submit the issue.

25
Chapter 6

Conclusion

Facebooks Free Basics has generated tremendous controversy polarizing people into either be-
lieving what Facebook preaches about the greatness of this program, or the concerns raised
by the proponents of an open Internet against the program. The authors in the previous work
empirically evaluated the concerns raised against Free Basics, specifically regarding data privacy
from Facebook, net neutrality violation and how effectively the program is reaching its target
population. This paper looks at Free Basics from a different perspective, focusing less on the
controversies that surround it. It seeks to understand that in presence of the many issues which
are completely or partially valid about the program, how can one harness this platform in lo-
cations where it is already available? How are global and local developers currently using this
platform? How transparent or opaque is Facebooks gate-keeping procedure allowing or barring
the entry of a particular service in the program? We present measurement based empirical
analysis for these different questions. Our experience shows that ICTD services targeted at
developing region populations can potentially benefit from such a platform, that increases the
out-reach of the world wide web. Such services aimed at information collection or dissemination
might not need data privacy from Facebook and also not use a lot of bandwidth. They can there-
fore potentially tolerate the negatives of the Free Basics architecture, analyzed in the previous
work. Our experience report of deploying three different Free Basics services will hopefully give
more transparency into this platform and encourage other ICTD researchers and practitioners
to consider whether the platform can benefit their cause and research agenda.

26
Bibliography

[1] Internet.org by Facebook. http://internet.org.

[2] Where weve launched - internet.org.. https://info.internet.org/en/story/

where-weve-launched/.

[3] Internet users by country (2016) - internet live stats. http://www.internetlivestats.com/

internet-users-by-country/.

[4] World databank. http://databank.worldbank.org/data/.

[5] Facebooks free basics service has been banned in india. http://www.theverge.com/2016/
2/8/10913398/free-basics-india-regulator-ruling.

[6] Prohibition of discriminatory tariffs for data services regulations, 2016. http://www.trai.
gov.in/WriteReadData/WhatsNew/Documents/Regulation Data Service.pdf.

[7] Technical guidelines - free basics. https://developers.facebook.com/docs/internet-org/

platform-technical-guidelines.

[8] How to submit - free basics. https://developers.facebook.com/docs/internet-org/

how-to-submit.

[9] IMC 16: Proceedings of the 2016 ACM on Internet Measurement Conference, (New York,
NY, USA), ACM, 2016.

[10] Free basics by facebook - android apps on google play. https://play.google.com/store/

apps/details?id=org.internet&hl=en.

27