π€Have you ever thought that the technology we use might be perpetuating biases we thought we had overcome?
π€ AI doesnβt just replicate societal biases; it can amplify them, especially against women and girls. This happens because AI algorithms learn from data that carries the same stereotypes and inequalities present in our society. Thatβs why legislation like the EU AI Act becomes crucial to prevent technology from perpetuating discrimination.
Letβs dive into a few examples that show why this issue is so critical:
π©Ί In healthcare: In healthcare, studies show that databases used in medical applications can lead to misdiagnosis, especially when symptoms are presented by women. This occurs because many systems were initially trained on male-based biological data. As a result, diagnoses for women were often biased: severe left arm and back pain, which could indicate a heart attack in men, was often misinterpreted as fibromyalgia or depression in women. This not only increases the risk of severe complications or even death for women but also reinforces gender stereotypes that link womenβs symptoms to emotional causes, while menβs symptoms are treated as physical and urgent.
π¨ In media tools: The generative AI tool βLensaβ was accused of undressing and sexualizing womenβs avatars, while male avatars were exalted as astronauts, explorers, and inventors. Why would an AI treat genders so differently?
πΌ In representation: βDALL-Eβ generated images of white men 97% of the time when asked to create images of βCEOsβ or βleaders.β This raises an unsettling question: why does technology, which should reflect diversity, still perpetuate exclusionary patterns?
π The role of the EU AI Act: The EU regulation classifies such cases as high-risk, requiring companies to perform bias audits, transparency reports, and regular human oversight.
π Training data that includes representative and qualitative data, reflecting diverse demographic groups like women and girls, will be essential for ensuring compliance and mitigating risk under the EU AI Act.
πΊ With this inclusive approach, companies can reduce the likelihood of perpetuating harmful stereotypes and make significant strides toward equitable AI.
π The EU AI Act isnβt just about control β itβs about equality and fighting discrimination. This legislation helps companies bring responsible and safe technologies to market, strengthening user trust and steering technological innovation in an ethical direction.
What do you think about this? Share your opinion and be part of the change!
Images generated by Dall-e based on the prompts of the author.
Leave a Reply