AI can contribute to diminishing gender-based prejudice and inequality in recruitment, advancement, and compensation. In this chart we see the relationships with the proportion of women subjected to physical and/or sexual violence in the last 12 months (% of women age 15-49).

This perspective offers a real-time view of the worldwide news on gender-related topics, providing us with a media perspective on the progress at the different fronts of SDG 5.

 

Developed in collaboration with 

A SDG Live Report is what it is all about! Here, you can have a real-time perspective of the state of the world in regard to AI and sustainability, tracking some of the main issues to address in the near future.

 

The impact of AI on the progress of Sustainable Development Goal 5 (SDG 5), which aims to achieve gender equality and empower all women and girls, has been gradually increasing. Between 2015 and 2023, there have been 54,895 scientific publications exploring the intersection of AI and gender equality. These studies often focus on how AI can help identify and address gender biases, support women’s health and education, and enhance economic opportunities for women. The relatively lower media exposure, with 5,719 news articles, suggests that while public interest is present, it is not as widespread as in other SDGs. However, the development of 277 AI policies targeting SDG 5 highlights a growing recognition among policymakers of the importance of leveraging AI to promote gender equality. These policies aim to ensure that AI technologies are developed and implemented in ways that are inclusive and equitable, addressing issues such as algorithmic bias and the gender digital divide.

Looking ahead, the next 5 to 10 years are likely to see significant advancements in the application of AI to promote gender equality. Increased research and policy focus will drive the development of AI tools that actively combat gender discrimination in various sectors, including recruitment, education, and healthcare. AI can be used to analyze large datasets to uncover hidden gender biases and to create more equitable systems and processes. Additionally, AI-driven initiatives will likely emerge to support women’s entrepreneurship and participation in the tech industry, helping to close the gender gap in STEM fields. As awareness of AI’s potential to impact gender equality grows, collaboration between governments, academic institutions, and the private sector will be crucial to ensure that AI technologies are harnessed effectively and ethically. This collaborative effort will help accelerate progress towards SDG 5, fostering a more inclusive and equitable future for women and girls worldwide.

 

Developed in collaboration with the European Commission project 

Here is your window to Gender in the context of AI. This dashboard is configurable and can be integrated into other systems to bring actionable technology into the systems of research institutions, governments and enterprises that are willing to make a difference.

 

 

Developed in collaboration with the European Commission project 

Avoiding data bias in AI systems is crucial to ensure fair, accurate, and equitable outcomes, preventing the reinforcement of existing inequalities and enabling more trustworthy and inclusive technologies. Here you will see a dashboard analysing the bias related to the data ingested in this observatory for SDG 5.

For analysis we use OECD AI Policy documents. Some of those documents are very large, and we split each document into smaller parts (so called “chunks”), which can contain multiple paragraphs. The reason for this is to prepare data for easier analysis with large language models, so called Retrieval-Augmented Generation (RAG). RAG is an advanced technique that combines retrieval-based methods with generative models to improve the performance of tasks such as question answering, text generation, and other natural language processing (NLP) applications. For each chunk then the sentiment is computed based on VADER (Valence Aware Dictionary and sEntiment Reasoner) methodology. Since VADER is known to have weak multilingual capabilities, all the documents were machine translated into English first.

While the results of this procedure are reliant not only upon the accuracy of the sentiment analysis tool, but also upon the accuracy of machine translation, it is important to stress that sentiment analysis is less sensitive to common machine translation problems than other usages, because sentiment analysis usually focuses on identifying the polarity (positive, negative, neutral) of a text rather than understanding its full semantic content. Also, sentiments in text are often expressed redundantly, which can help mitigate the impact of translation errors. As a result, minor translation errors that do not alter the overall sentiment and do not significantly impact the sentiment analysis is possible.

For the purpose of this analysis, they computed the average sentiment of (chunks of) AI policy documents for each country. We are presenting the visualisation of average sentiment of countries’ AI policy documents on the map. Since AI policy documents are mostly documents of legal nature (acts, policies, regulatory and governance frameworks), the sentiment should be mostly neutral, however, the analysis shows that there are country differences.

VADER computes positive, negative and neutral sentiment. Each of those values are between 0 and 1. The score indicates the proportion of text that is considered positive, negative and neutral. The sum of negative, positive, and neutral sentiment scores always equals 1, however in practice the sum of three sentiment scores can sometimes slightly exceed or fall below 1 due to floating-point precision errors or rounding issues that occur during computation.

 

Developed in collaboration with the European Commission project