Identifying trends around workplace discrimination using social media

7 min read
When we sit down with colleagues from different parts of the UN, we tend to start by explaining the kinds of insights that can be drawn from analysis of big data. Some of that is what people think – often expressed in their public statements on Twitter, Facebook, website fora or blog posts. We call that type of information social data.
The second broad category is what people do – this might include looking at aggregate trends in population movement using de-identified mobile phone data, or whether people are spending more or less money on phone credit – which can give an indication of whether people have more or less disposable income than six months ago. Mobile phone data, financial data, and even postal data can be used to give insights that can be useful for humanitarian response or development work.
But of course there are many actions and transactions that still happen offline. And there are many things that people think, that they would be uncomfortable discussing publicly.  Those issues aren’t always clear cut and different people and cultures have different ideas about what is acceptable to discuss on social media.
Often the best way to explore whether big data analysis could be a helpful tool is to run a feasibility test and look at a sample of digital data.
In a project with the International Labor Organization (ILO) here in Indonesia, that’s exactly what we did – and today I’d like to talk about what did and didn’t work in the spirit of open innovation and #FailFriday.
Ultimately, as is sometimes the case, the signals in the areas that ILO Indonesia really needed insights into were not strong enough. So instead we will explore a different research focus with them, looking at broader labor practices.
But we think it’s still useful to look at which issues people discuss in online conversations – and we hope other researchers will explore similar topics.

Eight Topics for exploration

Our starting point was a collaboration between ILO Indonesia, the Indonesian Government and Pulse Lab Jakarta. Our colleagues at ILO wanted to find out more about some of the discrimination that workers face in Indonesia by looking at what people say online. And they have a number of specific areas of interest.
Following a brainstorming session around common issues related to discriminatory practices or barriers to entering the workplace, ILO selected eight topics for a feasibility study.
Preliminary analyses looking at eight subtopics was conducted exploring the volume of signals available for this research. The subtopics were: sexual harassment in workplace, permission to work,  lack of skills or education,  cost to access employment, perceptions regarding the appropriateness of work, the multiple burdens of working women,  discrimination in job requirements and home-based workers.
We developed taxonomies in the Bahasa Indonesia language. Then using Crimson Hexagon – a tool for processing vast quantities of social media posts to find posts that match the taxonomy – we looked at each of the topics and identified which ones were promising targets for further research and which topics didn’t have sufficient volume to analyse further.
I’ll go through those briefly in turn.

Permission to work

Rationale:  Often women need to seek the permission of their husbands or families to be able to engage in productive employment.
Preliminary result: Promising
There are relevant online signals around the conversation on the need for permission to work from husband and family members. Relating to this topic, some people talk about continuing to work after marriage.
Some taxonomy refinement is required because other kinds of permission are captured, for example, whether teenagers should have a curfew.
Sample of Relevant Tweets:
“Gw pernah jg loh diminta resign kerja sama suami. Dan gw nurut.
Sampe akhrnya gw masuk kerja lagi pun atas ijin beliau.”
“One time my husband asked me to resign my job, and i just follow his suggestion.
After that I got a new job, also with his permission.”

Perceptions regarding the appropriateness of work

Rationale: Many women are constrained in terms of the types of employment they are able to engage in. This varies at different ages and stages in women’s lives. When a woman marries, some husbands prefer that their wife remain at home or in the community, rather than work a distance from home or in the formal economy.
Preliminary result: Promising
More than half of the collected tweets were relevant, including some conversation regarding research stating that women doing long-term night shift work could put them at risk of getting cancer.
Further refinement of the dataset would be required – for example excluding jokes.
Sample of Relevant Tweets:
“Waduh, jgn smpe, Wanita Kerja Malam, Rawan Kanker Rahim !!!”
“Oh please don’t let me get this, woman with night shift work is having high risk of cervical cancer!!!”

Multiple burdens of working women

Rationale: Having children and working can be especially challenging for women in Indonesia, with some finding it difficult to find appropriate childcare in their family, community or otherwise. Other women with domestic responsibilities feel they do not have time to engage in employment due to their duties at home.
Preliminary result: Promising
Some interesting conversations were discovered, such as discussion about double burdens of working women and a widely shared questionnaire asking people about their preference between career women and housewives. There was a mix of differing opinions on the subject, with some women noting that they felt they could juggle both career and family responsibilities.
Further taxonomy revision or dataset refinement could be conducted for this sub-topic.
Sample of Relevant Tweets:
“Pilih mana Perempuan Karir atau Ibu Rumah Tangga”
“Which one you choose, career woman or house wife?”

Discriminatory job requirements

Rationale: Some recruiters request ‘attractive’ women, women of a certain height and weight, women to wear or not wear a jilbab etc. Some recruiters also specify that a certain job can only be carried out by men.
Preliminary result: Promising
A large volume of tweets were captured using the current taxonomy. Most of them were job announcement posts rather than tweets with discussion or opinion.
To improve understanding about what would be considered discriminatory in the specific context of job requirements, a clearer definition would be needed.
Sample of Relevant Tweets:
LOWONGAN KERJA #Bag. Kasir. syarat : 1. Perempuan 2. Menarik 3. Pendidikan min. SMEA / SMA 4. Umur max. 23.
“Job vacancy: Receptionist position, Requirements: 1. Woman, 2. Attractive, 3. Min. Education High School, 4. Max Age 23

Lack of skills or education

Rationale: Lack of formal education or vocational skills training and an inability to pay or enroll in these can impact on women’s ability to work.
Preliminary result: Less promising
The tweets monitored by the current taxonomy are mostly advice or discussions about the skills or education considered necessary or important to get a job. Few Tweets mentioned women’s discrimination or difficulties in accessing or getting a job.
Sample of Relevant Tweets:
“Tenaga perempuan dianggap murah karena dianggap tak punya nilai ekonomi.
Skill nya tidak dihargai krn pekerjaan rumah tangga.”
“Woman labour considered as cheap worker because considered not having economic value. Their skill is not recognized because they usually do domestic work.”

Cost to access employment

Rationale: Some employers or recruiters require women to pay an ‘informal’ fee to obtain a job. This expense can vary from area to area.
Preliminary result: Less promising
There were few relevant tweets on this topic while many captured tweets discussed a particular case or issues related to bribes.
Sample of Relevant Tweets:
“Realitanya yg bkin susah cari kerja zamn skrg adalah “Mau ngelamar kerja aja hrs ada uang “Sogokan” dr pihak2 yg ga brtanggung jawab”
“The reality is very hard to get a job right now, because we need to bribe some irresponsible people to get a job.”

Home-based workers

Rationale: Many home-based workers are vulnerable to exploitation and are discriminated against by employers without written contracts or standard working conditions or protections.
Preliminary result: Not promising
Most Tweets over the past four years were irrelevant but only tens of tweets sharing a couple of news articles about home-based workers are relevant. Broadening the taxonomy by removing some keywords did not improve the size of the dataset.
Sample of Relevant Tweets:
“Pemerintah Perlu Buat Aturan Pekerja Rumahan ”
“Government needs to make a regulation for home-based workers.”

Sexual harassment in workplaces

Rationale: Sexual harassment at work can be an issue in male-dominated working environments.
Preliminary result: Not promising
There were almost no relevant tweets exactly fitting into this topic. Closely relevant tweets included tweets sharing a case of sexual assault, or sexual harassment in other places, such as public transportation, but there were few of these Tweets.
Result: Not promising
Perhaps unsurprisingly there were next to no posts related to sexual harassment in workplaces.
We encountered only two relevant posts to sexual harassment in workplaces while we came across 276 irrelevant posts and 21 posts partially related to general harassment.  Even regarding the two relevant posts, they are not user discussion or opinion but news sharing regarding sexual assault.
The Pulse Lab Jakarta research team tried various data analysis techniques including word cloud and topic analysis, and this also supports the discovery that online discussion related to sexual harassment in workplaces is not common in Indonesia. For example all the top ranked words and topics were unrelated to sexual harassment, but referred instead to more general harassment in workplaces.
The results of the analysis indicate that people in Indonesia tend not to explicitly discuss this topic on social media. PLJ and ILO Indonesia therefore agreed to move onto another topic for investigation with social media monitoring methods.
Sample of Relevant Tweets:
“Malam ini jam 9 di”suatutempat” ada para wanita korban pelecehan seksual oleh atasan di kantor buka suara. Salut dgn wanita2 pemberani ini”
“Tonight at 9, in someplace there some women, who are victim of sexual harassment in the office will open up their voice. Salute for all of those brave woman.”


So as described above this research project will not be continued in its current form.
But the good news for people interested in the area of workplace-related discrimination is that there are increasing volumes of relevant social data, with the number of posts captured by our taxonomy steadily increasing over time. For instance, although there were only a few posts (17 or 0.002% of the total) for the first two years until December 2010, the number of posts per month in 2013 was almost more than 10,000.
Clearly there are some issues that will never be widely discussed via social media. But by testing the taxonomies we are able to benchmark conversation, rather than making assumptions about what we may or may not find online.
If you have undertaken similar research in this area we are keen to hear about it so do get in touch with us
Photocredit: A women working on a loom in Bali, Indonesia, Hadi via Creative Commons
Identifying trends around workplace discrimination using social media 1

Did you enjoy this blog post? Share it with your networks!

News, thoughts and ideas about big data and AI, data privacy and ethics from across the Pulse Lab Network. Read more on the blog.

Scroll to Top