Threats, Abuses, Flirting, and Blackmail:

Gender Inequity in Social Media Voice Forums

Aditya Vashistha

University of Washington

adityav@cs.washington.edu

Abhinav Garg

University of Washington

aagarg@uw.edu

Richard Anderson

University of Washington

Agha Ali Raza

agha.ali.raza@itu.edu.pk

ABSTRACT

these services is almost non-existent. In this paper, we investi-

gate the reasons for the low participation of women in social

media voice forums by examining the use of Sangeet Swara in

India and Baang in Pakistan by marginalized women and men.

Our mixed-methods approach spanning content analysis of

audio posts, quantitative analysis of interactions between

users, and qualitative interviews with users indicate gender

inequity due to deep-rooted patriarchal values. We found that

women on these forums faced systemic discrimination and

encountered abusive content, irts, threats, and harassment.

IVR; Voice forum; Social media; Gender; Women; HCI4D;

India; Pakistan.

Permission to make digital or hard copies of all or part of this work for

personal or classroom use is granted without fee provided that copies are not

made or distributed for prot or commercial advantage and that copies bear

this notice and the full citation on the rst page. Copyrights for components

of this work owned by others than ACM must be honored. Abstracting with

credit is permitted. To copy otherwise, or republish, to post on servers or to

redistribute to lists, requires prior specic permission and/or a fee. Request

permissions from [email protected].

5

,

6

,

8

]

Unfortunately, gender inequality also manifests in adop-

9

]. Even when they own

9

].

Recognizing thesestructural limitations,Human-Computer

calls, literacy barriers by using local language speaking and

listening skills, and socioeconomic barriers by using toll-free

(1-800) lines. These services let users call a toll-free phone

number to record audio posts in their local language, listen

to posts recorded by others, and act on posts (e.g., like, dis-

like, comment, or report posts). They nd applications in

22

24

tremely low-participation of women on these services. For

27

38

larly, Baang [

32

] and Polly [

33

] in Pakistan have only 10% and

11% female contributors, respectively. Ila Dhageyso [

], a

voice forum to connect government ocials and tribal people

in Somaliland, has only 15% female users.

In this work, we examine why the participation of women

is almost non-existent on social media voice forums that are

designed to be inclusive, accessible, and usable for everyone.

To do so, we focus our attention on Sangeet Swara (or simply

Swara) and Baang, two widely popular social media voice

factors aected their participation. Our work also highlights

voice forum users in India and Pakistan.

Our mixed-methods analysis indicate that womenon Swara

and Baang faced systemic discrimination and harassment in

them. Most women lacked agency to retaliate due to deep-

rooted patriarchal values and most men who behaved inappro-

their behavior. Most male users condoned unruly behavior by

provide an equitable and inclusive platform to women.

2 RELATED WORK

dia platforms by low-income communities, rural residents,

17

39

44

46

28

]. Bosch examined the role of social media platforms

16

]. Bidwell et al. examined

].

Several HCI4D researchers have leveraged voice forums

,

4

,

20

21

,

31

,

33

,

40

,

presents the rst in-depth account of how social media voice

forums were used by women in India and Pakistan.

We use a feminist HCI lens to examine the design of Swara

and Baang and to make recommendations for creating ser-

vices that foster gender equity in their use. Bardzell intro-

duced the feminist HCI research agenda and presented a

range of feminist interaction design qualities such as plu-

ralism, participation, advocacy, ecology, embodiment, and

and use of HCI artifacts [

19

,

23

25

sectional HCI lens [

35

,

42

] to examine marginalities within

marginalities in the use of these services.

3 BACKGROUND

We now briey discuss the purpose, design, and deployment

details of two social media voice forums, Swara in India and

Baang in Pakistan, that are the focus of our inquiry.

Since voice forums have thousands of audio posts in local

languages with limited training data and speech recognition

voted on a wide variety of user-generated content including

songs, poems, jokes, among others [

]. Swara relied on users’

votes to lter low-quality content and leveraged micro tasks

completed by users to categorize posts. In an 11-week deploy-

ment in India, Swara received about 25,000 calls, over 5,300

posts, and 140,000 votes from 1,521 people living in 11 states.

The community moderation was 98% accurate in content cat-

egorization, and made meaningful distinctions between high-

quality andlow-quality posts. Mostof the userswere from low-

income families living in rural and peri-urban regions, and

used the service to expand their social network in far-o loca-

communities. Although Swara was a toll-free service that

could be accessed from any phone without requiring airtime

]. Baang and

While low usage of these services among women could be

], and phone sharing where

in India and Pakistan to analyze these posts. The coders were

familiar with local socio-cultural norms, languages, and collo-

quial terms. On average, the coders in India and Pakistan were

32 years old and 27 years old, respectively. They had at least

a bachelor’s degree and were from middle-income families.

audio posts. For each post, a coder noted content type, gender

of the recorder, how the post is related to women, and whether

the recorder did not speak anything (e.g., in a blank post or

for pre-recorded content). When a post had multiple speakers,

the coders were asked to mark the gender of the person who

spoke for the most time. If a recorder referred to specic fe-

male users in the post, the coders marked the post as ‘directed

at female users’. If a recorder referred to women generally in

the post, the coders marked the post as ‘directed at women

in general.’ If a recorder had a conversation with other male

users on topics that followed prior conversations with female

users or on women, the coders marked the post as ‘discussion

when the recorder threatened other users in the post.

one partitionto eachcoder. Collectively, thesix coders listened

tonearly 140hours ofpoststo generatemetadata thatis central

probes, such as the number of female and male contributors,

similarities and dierences in content recorded by women and

men, content of messages directed at women, and interactions

between female and male users, among other things.

To examine how female and male users reacted to non-

inclusive posts such as abuse, irts, threats, or verbal harass-

ment, we mapped each anonymized phone number with the

was used by both male and female users.

alitative Analysis

domly selected users who used these services more than ten

times and recorded at least one post. We conducted semi-

Twenty Baang participants were male and 12 were female.

Table 1 shows how men and women participated on Swara

and Baang. Men on Swara recorded 17 times more posts, and

liked and disliked these posts 80 times and 22 times more

than women. On average, they recorded 1.5 times more posts,

women. We found similar usage patterns on Baang, indicating

that the participation on both services was dominated by men.

Even the most fervent female users were far behind their male

by top 25 female contributors combined were less than the

utors. The median number of posts recorded by these men

(

,Z =

.

,p <

gender (

Content created by users:

Compared to women on Swara, women on Baang recorded

ported posts containing messages to others (52%) and mis-

cellaneous content (16%). In addition to reporting such posts

(15%). These ndings indicate that women did not use this fea-

ture completely and men misused it even to report acceptable

content. Our interviews indicated that many female users

were unaware about the process once a post was reported.

Our content analysis also indicated that a notable number

Figure 4: Distribution of dierent types of posts (on a log

scale) directed at Swara female users.

Figure 4 shows the distribution of Swara posts (on a log

scale) for the three categories. Male Swara users recorded

500 posts (83%) that were directed at female users or that dis-

cussed women in general. About 95% of the posts that called

out female users had abuse, irts, and adulation. Many men

women. About 90% of these posts had male users ghting

with each other to impress other female users or requesting

troublemakers to avoid recording abusive and irty posts.

We found a similar pattern on Baang. While most women

recorded supportive comments for other female users, the

majority of posts from men were abuse, irts, or verbal harass-

ment directed at women. In ten posts, men either supported

or rebuked other male users for abusing women.

These results indicate that although Swara and Baang users

had only 11% and 7% of all posts targeting or discussing

A Fisher’s exact test indicated a signicant eect of gender

On Swara, men irted with women in 166 posts (97%) and

with other men in two posts, and women irted with men in

Men irted with women inseveral ways.For example,many

Some men were forceful in sharing their feelings with

We found two Swara posts where men irted with other

We also analyzed users’ votes to examine their reactions

Although many interview participants stated reporting

we could not nd any quantitative evidence for the claim.

While women did not report any irty posts, men reported

such posts only 22 times. Male and female interview partic-

ipants gave dierent reasonings for why men were irting

with women. Some male participants held movies responsible

for irtatious behavior of men. One of them stated:

On the other hand, four female participants blamed women

for irtatious behavior of men. One of them stated:

To summarize, these results indicate that most of the irty

posts on Swara and Baang were directed at women. Often

Threatening Posts

Swara and Baang had 104 and 36 threatening posts, respec-

tively. All threatening posts on Swara were recorded by men

and the majority of these posts (62%) were directed at women.

We found a signicant dierence between the proportion

of threatening posts directed at women and men (

χ

,N =

,

ening posts were recorded by men and 19% of these posts were

directed at women. A Fisher’s exact test indicated a signi-

man recited the following content as a poem:

We found three main reasons why users were threatening

gestive content, sparking sharp criticism from other male

While no female Swara users threatened men, we found

one-third on Baang were because of irting. Some men con-

stantly harassed women to share their phone numbers. When

these women did not share their number, men felt rejected and

recorded abusive messages directed at these women. Some

men instead of responding to their irty messages. When

other users admonished these men for abusing women, they

Our interviews indicated that a few women on Swara and

Baang were blackmailed by men. As previously outlined, sev-