Women 4 Ethical Artificial Intelligence

The Women 4 Ethical Artificial Intelligence conference was held on October 30th at UNESCO HQ in Paris, France. The event was held in presence, but luckily for me (who found out about it only after it had happened), I was able to find the video recording of the event online.  

The event aimed at exploring real-world solutions to reach a more ethical AI ecosystem. The discussion was focused on the topic of Applying a Gender Lens in AI RAM (Rwading Assessment Methodologies) and was moderated by Maricagrazia Squicciarini, Chief of the Executive Office, Social and Human Sciences, Unesco. The other panelists were lead experts of  RAM from Uruguay, Mexico, Morocco, and Chile:

• Carolina Aguerre, Professor at Universidad Catolica del Uruguay
• Elena Estavillo Flores, Founder and CEO, Centro-i para la Sociedad del Futuro, W4EAI Working Group Co-Chair
• Saida Belouali, Professor at Mohammed First University 
• Julio Alberto Pertuze Salas, Former Under-Secretary for Economics and Small Enterprises

 

What is RAM?

Mariagrazia Squicciarini opens up the discussion by stating that 193 member states of UNESCO have agreed to use RAM, which is a tool “used to help get a picture of where we are, what we have, or don’t have on how to use AI ethically”.  RAM stands for Readiness Assessment Methodologies and it’s “A tool designed to help countries understand how prepared they are to apply AI ethically and responsibly for all their citizens”. It includes a range of quantitative and qualitative questions that aim to gather information that will later show how prepared is a country to face AI’s ecosystem, which includes the legal, regulatory, educational, social, cultural, scientific, and technological dimensions.

Its implementation is unique and adapted to each country’s circumstances, depending as well on the budget that is allocated for its research. Other than the completion of the questionnaire, RAM offers a comprehensive overview of a country’s status, “summarising where the country stands on each dimension, detailing ongoing activities, summing up the state of the art, and providing concrete policy recommendations on how to address governance gaps”. This helps UNESCO understand which national measures need to be tailored and applied for each specific case, to help achieve a more ethical AI ecosystem. 

 

The case of Uruguay – Carolina Aguerre

In Uruguay, Professor Carolina ran the RAM in the previous year (taking in context that the current AI national strategy was under revision), and what she found was that there is a strong lack of data which inhibits the creation of proper measures that would look forward to a more inclusive AI ecosystem. RAM encompasses over nearly 200 indicators and one of its core socio-cultural dimensions which it addresses is the gender one – this, she says, brings the kind of granularity needed to bring further perspective on how to address AI governance and women’s participation in the governance of AI.

She found that even though in terms of digital usage, men and women scored almost the same (91% of women, and 89% of men using the internet in the past 3 months), this was not the case when she addressed the presence of women in STEM careers in the country, where she found a 10% gender gap (45% women vs 55% men). Uruguay ranks 67th in the global gender gap index, and RAM showed that there have been few indicators that point towards a strategy of inclusion and participation of women in AI and ICTs. 

 

The case of Mexico  – Elena Estavillo Flores

Similar to Carolina, Elena faced difficulties in conducting the RAM due to the lack of official governmental interest or focus on AI. However, she gained support from the Senate, bringing together a more informal group of academia and sparking public interest in the topic. As in Uruguay, they found that in terms of digital usage, there is not much difference depending on the gender dimension, but what she pointed out, is that this is a high-level, general approach.

When looking more closely, she found that more men use the internet for work, single-parent households led by women (who are also facing higher poverty rates) spend less on mobile broadband, and rural populations had significantly less access than urban ones. Lack of data was another challenge, as gender metrics were not systematically measured but rather disaggregated and inconsistent, they worked on data that was gathered from different sources, making it harder to evaluate and compare. For instance, in early ages, only 27% of high-performing girls in science and math aspired careers in STEM, compared to 43% of boys, but this percentage did not translate in the future, where only 32% of STEM graduates are women (min 12).

The RAM process emphasized the need for actionable proposals, including training stakeholders on bias and gender perspectives in AI, reducing digital violence against women, and addressing the feminization of poverty in AI-related labor contexts. Ultimately, Mexico’s experience highlighted the gap between legal frameworks and real-world implementation, underscoring the need for transparency, public data, and a shift from rhetoric to actionable outcomes.

We need to move from discourse, resource to actions (Elena Esvaillo Flores – min 16).

 

Something that caught my attention was a note made by Mariagrazia between the panelists’ discourse. She mentioned the Leaky STEM pipeline phenomenon, and pointed out that all of the statistics we’ve heard so far are the entry point – but how many of these women are actually able to break the glass ceiling and reach high-ranked positions? An analysis ran by UNESCO showed that even simply the language used in some job postings could exclude women because it can be perceived as “hostile”, whereas gender-neutral language, or “female-friendly” language (as she calls it), doesn’t pose any barrier for anyone, and men and women enter the same way. She points out that this is something that could easily be facilitated and is not costly to put into use through prescriptive suggestions of language use, which could have a huge impact on the participation in job market (especially in AI). 

We are here not complaining just for the sake of complaining, here we’re trying to underline that everybody benefits by the time women are more included (Maricagrazia Squicciarini – min 18).

 

The case of Morocco – Saida Belouali

Morocco’s RAM, led by Professor Saida Belouali was the one that surprised me the most. She mentioned that even she was surprised to find out the results of RAM. Before going into RAM’s findings, she wanted to stress that there are a lot of women and girls who are very involved in the world of technology. For example, in her IT and engineering class, 90% of her students are girls, and “that is not something that you would find in statistics, but it is the reality”. 

Thanks to RAM, she found that Morocco has the highest levels of female engineers in the world – 42%, while France and Japan are only at 26% and 14%. Moreover, her discourse strongly emphasized the importance of the inclusion of the local, reminding me of Paulo Freire’s work “Pedagogy of the Oppressed”, which really moved me at the time of reading.

Throughout this conference I felt like I was mostly presented numbers and statistics, which even though claim to represent reality, they didn’t really help me paint the picture of future solutions realistically. The discourse of lack of data seems to always be present, but the emphasis on the local and cultural aspect was something that was exciting to me. I agree with her in her declaration that without a strong cultural, social, and political support base, we wouldn’t able to achieve digital parity in technology. But to really change something, you need to go small, local, close as well. 

 

When one talks about gender it is undoubtedly a complexed debate, when we want to come to pure entire parity here we have to talk about a cultural revolution, and its not merely that will come from technology. We can use this technology to maybe transforms the mentalities, peoples’ approaches. We could use this tech to change the way we conceive our worlds (Saida Belouali)

 

Our brains translate into these codes. The more diverse and the more inclusive [our brains], the more inclusive the results will be (Maricagrazia Squicciarini).

 

 

The case of Chile – Julio Alberto Pertuze Salas

Chile was the very first country to finalise the RAM, and have leveraged this exercise to a massive extent since then. Presented by Julio Alberto Pertuze Salas, RAM’s results showed that also in Chile there is virtually no gender gap in terms of access to technology and internet usage. The difference starts to arise when looking at the performance between men and women in STEM fields, with a difference of more tham 10% between the two (38,1% and 22,7%).

Another common point that was brought up is that there is little data that could be used, for example, universities are not required to run/disclose information on whether publications or professors have parities of gender. Neverthteless, he kept an open view and saw AI technology not only as a challenge but also as a tool, so through a Chat GPT-run exercise, he was able to find that out of 3704 published papers, 77% were of men and 18% of women. 

 

Solutions

To conclude, more (open) data is needed to achieve better decision-making. However, data is never neutral and this should not be taken for granted. As Elena and Saida have stressed, inequalities are systematically produced. To reach an ethical AI ecosystem we must first change the socio-cultural environment in which AI is being produced. Even the languages in which these models are being trained, can have a huge impact on what biases and cultural, social, and gender norms are being translated as “neutral” in these technologies.

Moreover, Saida’s statement really stuck with me “The main challenge is making sure is that the gender concept really finds its place in the local context.” (min 47). When talking about the case in Morocco, she raised the awareness of a different insight than from her peers, when coming from international organisations, “there is a form of resistance against the debunking of biases, perceived of a dynamic that is imported”. With this statement we understand the importance of the local approach and cultural sensibility that is needed to achieve a more ethical AI ecosystem. 

 

Final words 

Mariagrazia concluded the convention by asking all lead experts to share a final sentence on the question “Where would they start with?”:

Elena Estavillo Flores: Bring women to the table. Open the space and bring them to the table. 

Saida Belouali: I would like to talk about multiculturalism once more. I think we should listen to local voices. I think that there are local voices that can transform the reality that they know best. So I think we should bring women to the table, but we should also be careful about the fact that they could be different, and difference can be interesting. 

Carolina Aguerre: In the name of diversity which is what many of the policymakers around are concerned, I would emphasize the positive discrimination in terms of data, in terms of their rapprochement for women. So, women in data, women for data in AI, and women at the table.

Julio Alberto Pertuze Salas: Chile is a small country so we’re not necessarily shaping the technologies that are shaping the world, so I would like to bring the attention on the usage of those technologies by different economics sectors and scientific disciplines to try to narrow down where the gender gaps might be increasing or reducing due to the use of these technologies.

 

Feature image: mikemacmarketing, CC BY 2.0 , via Wikimedia Commons