f<a+i>r Middle East and North Africa Hub

Middle East and North Africa Hub

Arab states rank lowest in the Global Gender Gap Index on multiple fronts including women’s economic participation, and access to economic opportunities (World Economic Forum 2019). The region also has the lowest political participation of women globally (Kumar 2014). The gender gap reflects directly on the gender digital divide in the region. While on average more men access the Internet than women globally, this gap is most pronounced in the MENA region (ITU International Telecommunication Union 2019).

Inequalities exist as well between countries. According to the AI Readiness Index, the UAE, Qatar, and Saudi Arabia are ranked highest in the region in terms of readiness to capitalise on and meet the challenges brought forward by AI. The index reveals a large disparity in AI readiness between countries in the region, some scoring in the top 20 such as the UAE, and others scoring in the lowest tier. In fact, four of the countries are in a state of an ongoing war.

Concern over the use of AI for military and policing purposes is legitimate. The Gulf states, in particular the UAE and Saudi Arabia, are most committed to a national AI agenda, aggressively pushing their AI capabilities and centring their digital transformation around AI. Both have shown a strong commitment to becoming leaders in the field, the UAE went so far as to appoint a Minister of AI and Saudi Arabia has a dedicated government authority for “Data and Artificial Intelligence”. As oil-rich countries, AI presents an opportunity for diversifying these countries’ economies. However, combined with the countries’ track record of political repression and an absent civic space, this technology put at the hands of governments is a grave threat to human rights to which gender equality is inextricably linked.

Civil society organisations in the MENA already face political and societal obstacles, and few engage with topics at the intersection of technology and human rights and social development, let alone the impact of AI in society. While many governments in the region have implemented or drafted a policy around data and AI, not as many CSOs are debating these policies. A stringent civic space, particularly in absolute monarchies, discourages this type of activity.

However, research ecosystems in the MENA have a unique chance to operate independently from socio-political pressures. Developments in the field of Arabic natural language processing (or Arabic NLP) is a critical step to democratise access to AI in the region. Not only is Arabic the fifth most spoken language, but it is also highly complex, comprising a multitude of dialects that know no orthographic conventions (Darwish et al. 2020).

Bias can emerge at the level of data annotation particularly if annotators are indifferent to dialectic differences (Sap et al. 2019). Researchers in the MENA recognise this conundrum and are heeding to its details such as mapping dialectical words into their corresponding Modern Standard Arabic words (Duwairi 2015).

At a high level, bias can also emanate from the content within lexical resources. In recent years researchers at NYU, Abu Dhabi have developed the “Arabic Parallel Gender Corpus” with an intention to complement machine translation applications (particularly those that translate to Arabic) and make them gender-aware (Habash, Bouamor, and Chung 2019). Arabic is a gendered language, and such a corpus allows the development of applications that are able to identify gender-specific terms. This work is a critical first step. To cite a fundamental example, recent research associates languages with high gender biases more likely to associate men with careers and women with family, and countries with a stronger language bias have a lower representation of women in science (Lewis, Molly, and Gary Lupyan. 2019).

Investing in inclusive lexical resources offer a chance to not repeat existing biases. Kohl, a feminist journal on gender and sexuality in the MENA region, has recently inaugurated “conceptual maps of translation” producing a database of feminist and queer-inclusive terms and their corresponding translations in Arabic bearing in mind dialectic nuances (Abbani 2021).

Processes of innovation and development of intelligent systems are necessary to link different sectors to promote infrastructures, capabilities, and policies that incorporate human rights-based principles of gender equality, and include women and girls’ participation. This is essential for  MENA communities to differently thrive in the 21st century.

A more equitable AI future in MENA will need to expand collaboration platforms among multiple social actors, promote situated research, and develop empirical evidence based on use cases anchored in local communities and from a gender perspective. This investment in research, and breaking down of silos between academic ecosystems, industry, and the government seeks to change the trajectory of AI and ADM in MENA by including women, girls, and feminist principles in the creation of the new AI that will affect all in the region.