Background

Background:

Violence against women (VAW) is one of the most widespread violations of human rights. It can include physical, sexual, psychological and economic abuse, and it cuts across boundaries of age, race, culture, wealth and geography. It takes place in the home, on the streets, in schools, the workplace, in farm fields, refugee camps, during conflicts and crises.

The growing reach of the Internet, the rapid spread of mobile information and communications technologies (ICTs) and the wide diffusion of social media have presented new opportunities and enabled various efforts to address VAW. However, the online space has been used to inflict harm on women and to perpetrate VAW. Online and ICT facilitated -VAW is emerging as a global problem with serious implications for societies and economies around the world. This situation has been compounded with the COVID-19 crisis, which has initiated a shift to the online space and has fueled both online misogyny and ICT-facilitated violence.

In 2021, UN women Regional Office for Arab States (ROAS) conducted a comprehensive study project to reveal the prevalence, impact and consequences of online violence against women (online VAW) in the Arab states and deepen the understanding of the barriers that women survivors face to report this type of violence and access services.[1] Some of the study findings included that almost half of women in the Arab states do not feel safe online. It also revealed the significant threats online violence imposes on women in the Arab countries, especially women activists. Increasingly, the region witnessed several VAW incidents linked to women’s online presence, and it often represents a serious threat to women’s physical safety and metal wellbeing.[2] In addition to the shocking physical and mental consequences of online violence, the study also highlighted that online violence hampers women’s full participation in the society and contributes to silencing their voices in the Arab states, as over 1 in 5 women (22 per cent) who experienced online violence deleted or deactivated their accounts.

In Libya, the web-survey was completed by 1362 respondents, including 507 Libyan women. The survey indicated that 16% of women in Libya experience online violence, of which 62% were women in 2020, and 46% had experienced it more than once. Another survey that was directed to women activists has demonstrated that the situation is worse for activists in Libya. 36.40% had experienced online violence at least once, while 57.60% of women activists do not feel safe online. Ultimately, the online violence against women activists and in politics are a direct attack on women’s visibility and full participation in public life.

Online Violence acts as a weapon aimed at silencing women’s voices, hindering the peaceful exercise of their rights as well as an obstacle to achieving gender equality.[3] As noted by the UN Special Rapporteur on violence against women, its causes and consequences, “Online violence against women not only violates a woman’s right to live free from violence and to participate online but also undermines democratic exercise and good governance, and as such creates a democratic deficit.”[4] Therefore, shedding the light on online violence is essential for supplementing Libya’s efforts to transition into a sustainable and stable democracy underpinned by the rule of law and respect for human rights. Moreover, despite the substantial implications of online violence on survivors, the regional web-survey showed that 39 % of women and 53% of men in Libya believe that "online violence is not a serious issue as long as it remains online." 43% of women and 55% of men believe that "women who display their own photos and videos should accept that the material can be used against them." A major consequence of online violence is a society where women no longer feel safe either online or offline, given the trivialization of online violence acts and the widespread impunity for perpetrators of gender-based violence.[5]  

Social media is usually considered to be the digital tribunes where people express their thoughts and opinions. However, this freedom comes at a price, especially when it enables spreading abusive language and hate speech against women. What worsen the situation is that the internet intermediary companies’ human and automated reviewers struggle to comprehend the varied dialects used across the Middle East and North Africa.[6] Yet, research has proved that Misogyny automatic detection systems can assist in the prohibition of anti-women Arabic toxic content. Developing such systems is hindered by the lack of the Arabic misogyny benchmark datasets[7] and very little datasets publicly-available such as Levantine Hate Speech and Abusive (L-HSAB) Twitter dataset[8].

UN Women Libya is seeking to engage a consultant to conduct a big data analysis and text mining on online VAW in Libya to shed the light on the situation and contribute to creating an Arabic misogyny benchmark datasets for Libya. The work aims to a). shed the light on the prevalence and magnitude of online VAW in Libya, especially activist women and women in politics; b). document and analyze the incidents of online VAW in Libya during the period of 2021-2022; and understand the common patterns, drives and types of online VAW.

[1] The key findings of the regional research study on online violence in the Arab states is available at https://arabstates.unwomen.org/en/digital-library/publications/2021/11/violence-against-women-in-the-online-space

[2] This has notably been the case in Egypt where women were being detained due to their activities on TikTok, find out more here https://www.bbc.com/news/world-middle-east-57566506

[3] Lawyers for Justice in Libya (LFJL) report “We will not be silenced -online violence against women in Libya” https://www.libyanjustice.org/news/urgent-action-needed-to-address-shocking-levels-of-online-violence-against-libyan-women

[4] UN Human Rights Council, ‘Report of the Special Rapporteur on violence against women, its causes and consequences on online violence against women and girls from a human rights perspective,’ A/HRC/38/47 (18 June 2018), para.29. https://digitallibrary.un.org/record/1641160?ln=en

[5] UN Human Rights Council, ‘Report of the Special Rapporteur on violence against women

[6] The Wired article https://wired.me/business/facebook-is-everywhere-but-its-arabic-moderation-is-nowhere-close/

[7] Let-Mi: An Arabic Levantine Twitter Dataset for Misogynistic Language https://aclanthology.org/2021.wanlp-1.16.pdf

[8] L-HSAB: A Levantine Twitter Dataset for Hate Speech and Abusive Language https://aclanthology.org/W19-3512/

Duties and Responsibilities

The overall objective of this assignment is to retrieve, detect and analyze online posts on the online violence incidents and the experience of women in Libya, focusing on certain social media platforms that include mainly Facebook (FB) and Twitter. This consultancy aims to provide evidence on the magnitude, nature, and patterns of online VAW in Libya to demonstrate the impact of this life-threatening phenomena on their lives, especially Activist women and women in politics. The work will shed the light on this by documenting and analyzing the incidents of online VAW in Libya during the period of 2021-2022; and revealing the common patterns, drives and types of online VAW.

The consultant will work under the guidance of the Libya Head of Programmes and ROAS EVAW team to:

1. Prepare a list of key words either using a publicly available latest, reliable, Arabic datasets or manually annotating twitter/ FB posts that would capture online violence and misogynistic speech in Arabic language and in the Libyan dialect that will be used as a training dataset for detecting and classifying misogynistic posts.
2. Data acquisition:
   1. Gather all the geo-tagged online posts over 2021-2022 for Libya using elevated data access APIs or using TWINT package for Twitter metadata and any other possible methods for Facebook and Twitter platforms.
   2. Include demographic data such as gender, location, age, etc in the datasets or provide options to incorporate disaggregated data.
   3. Build a clean labelled dataset.
3. Features and Model Selection: Develop a classification model based on established literature (detection/ prediction) of the online posts to identify the online violence content targeted towards women (via machine learning model, and manually first using a small subset[1] to enhance the accuracy of the model - e.g., Precision/Recall). It is preferred that the model is robust and inclusive of heterogenous features but is not overfitted. One may use readily available and reliable Python/ GitHub packages where necessary.
4. Interpretation: Conduct a baseline and an in-depth levels analysis of the data, presenting trend analysis and demographic analysis. Other interpretations such as which features best explain model performance, what are the data &/or model limitations, etc.
5. Provide a report with both numerical and qualitative analysis of the data accompanied with graphs and visuals that captures the key findings from the data.
6. Create a presentation that summarizes the key findings targeting the global and national audience.

The consultant is expected to hold several follow up meetings with the ROAS EVAW team and UN Women Libya office.

* Consultant are expected to be equipped with all materials and equipment needed to perform their tasks for example laptop, software etc..

[1] Referred to as a “training dataset”. Depending on the total number of the posts and Tweets, the rule of thumb is an average of 30% of the total. Based on the manual classification, the machine learning model is then trained to recognize certain types of online violence patterns using the small subset, resulting with an algorithm that will be applied to the remaining broader dataset. Based on the accuracy, consultant would improve the algorithm by adding more features leading to improved detection ability.

Competencies

Core Values:

• Respect for Diversity
• Integrity
• Professionalism

Core Competencies:

• Awareness and Sensitivity Regarding Gender Issues
• Accountability
• Creative Problem Solving
• Effective Communication
• Inclusive Collaboration
• Stakeholder Engagement
• Leading by Example

Please visit this link for more information on UN Women’s Core Values and Competencies:  https://www.unwomen.org/sites/default/files/Headquarters/Attachments/Sections/About%20Us/Employment/UN-Women-values-and-competencies-framework-en.pdf

Functional Competencies

• Proof record of experience in Arabic Natural Language Processing, Text mining or other related relevant experience;
• Strong understanding of and commitment to gender equality and women's empowerment and its policy implications;
• Excellent written communication skills in English and Arabic, including editing;
• Ability to consolidate information from multiple sources;
• Ability to prepare strategic information for decision makers;
• Ability to work independently as well as good team player;
• Excellent time management and ability to produce outputs as per agreed deadlines;
• Demonstrate knowledge of the Libyan context and the Libyan dialect is preferable.

Required Skills and Experience

Education:

• An academic degree in Natural Language Processing, Text mining, or other related relevant academic degree.

Experience:

• At least 5 years professional experience producing Natural Language Processing and using Twitter and FB APIs.
• Experience and understanding of Arabic language bias, gender-based violence, and online violence against women on social platforms (especially FB and Twitter).
• Demonstrated experience in producing reports relating to big data analysis and Gender Based Violence is an asset

Language Requirements:

• Fluency Arabic and English is required;
• Knowledge of Libyan dialect is a strong asset.

Application:

All applications must include (as an attachment) the completed UN Women Personal History form (P-11) which can be downloaded from http://www.unwomen.org/about-us/employment.

Kindly note that the system will only allow one attachment. Applications without the completed UN Women P-11 form will be treated as incomplete and will not be considered for further assessment.

At UN Women, we are committed to creating a diverse and inclusive environment of mutual respect. UN Women recruits, employs, trains, compensates, and promotes regardless of race, religion, color, sex, gender identity, sexual orientation, age, ability, national origin, or any other basis covered by ap-propriate law. All employment is decided on the basis of qualifications, competence, integrity and or-ganizational need.

If you need any reasonable accommodation to support your participation in the recruitment and selec-tion process, please include this information in your application.

UN Women has a zero-tolerance policy on conduct that is incompatible with the aims and objectives of the United Nations and UN Women, including sexual exploitation and abuse, sexual harassment, abuse of authority and discrimination.All selected candidates will be expected to adhere to UN Wom-en’s policies and procedures and the standards of conduct expected of UN Women personnel and will therefore undergo rigorous reference and background checks. (Background checks will include the veri-fication of academic credential(s) and employment history. Selected candidates may be required to provide additional information to conduct a background check.)