Education
Inclusive AI: Can Technology Overcome Bias?
Discover how Artificial Intelligence can be an ally for inclusion, avoiding stereotypes and promoting diversity.
Artificial intelligence has the potential to drive inclusion, but it often reflects and perpetuates existing societal stereotypes, excluding millions of people. This article, inspired by insights from SXSW 2025, explores how we can train and develop AIs to be truly inclusive. Learn practical principles and tips to transform technology into a tool for equity for all.
In this chat from Turma do Jiló, Sheila and Laura delve into valuable insights from SXSW 2025 to discuss an urgent topic: how Artificial Intelligence (AI) can and should be a tool for inclusion. Despite its enormous potential, AI often reflects and perpetuates the stereotypes of our society, marginalizing millions of people, especially those with disabilities.
PUBLICIDADE (GAM) — 1250x250
The Challenge of Biased AI and the Importance of Representation
The conversation highlights that AI, from its beginnings, has struggled to recognize diversity. Remember when Face ID systems failed to identify Black people? This demonstrates that the problem is old and is exacerbated by technological advancement.
An alarming statistic, from the lecture “Disabling AI: How to Make Machines See Disabled People,” reveals that 1 billion people worldwide identify as disabled, and more than 3 billion have a family member or friend with a disability. However, their voices and representations are often ignored or distorted by AI. This happens because AI is fed data, and if that data is not inclusive, it reproduces the unconscious biases present in society.
For example, when asking AI to generate images of people with disabilities, they often appear sad, alone, or in somber settings, reinforcing a stereotype that disability is synonymous with tragedy, or, conversely, with forced heroism. This bias affects not only visual representation but also critical processes like recruitment, where algorithms can favor specific profiles, excluding diverse and equally competent talents.
The Power of the Prompt: Training AI for Inclusion
The good news is that we can act! AI reflects our society and the data we provide. Therefore, it is essential to train existing AIs and develop new ones with inclusive databases.
You don't need to be a developer to contribute. As users, we have the power to influence AI through prompts (commands we give to artificial intelligence). Laura exemplifies how she trained ChatGPT to write in plain language, providing clear context and guidelines. Instead of just asking for a paragraph about “people with disabilities in sports,” she first defined what a person with a disability is, explained adapted sports, and the importance of these adaptations. This way, the AI generated responses much more aligned with the principles of inclusion, equity, and accessibility.
This approach is crucial because AI can be a powerful ally. Applications like Be My Eyes help blind people identify obstacles, and assistants like Alexa assist in the daily routines of people with disabilities. The key is to use technology to further inclusion.
4 Essential Principles for More Inclusive AI
From the SXSW 2025 lecture, Turma do Jiló shares four principles to correct and make AI more inclusive:
1. Data is a Choice
The data that feeds AI is crucial. When creating prompts, be detailed and ensure that the information provided promotes diverse and representative images and texts. Our bias can be positive if directed towards inclusion.
2. Design a Better Bias
A bias is not always negative. We can train AI to have a positive bias toward inclusion. Providing data and contexts that enrich AI's understanding of diversity is fundamental, as in the example of ChatGPT and plain language.
3. AI is Never Finished: Constant Testing and Retesting
Technology is constantly evolving. It is essential to test and retest AIs, providing inclusive data and seeking out other platforms if one is not aligned with inclusion. Diversity improves, and AI should too.
4. Choose Your Team Wisely
Whether creating or using AI, the team needs to be inclusive. Developers should include people with disabilities and diverse individuals on their teams. Diversity is not just about quotas; it's a competitive advantage, driving innovation and creativity. Companies with diverse teams are better prepared to navigate periods of crisis and major changes.
The Human Aspect in Technology: Lessons from SXSW
SXSW insights reinforce that, even with impressive technological advancements (like robots with articulation), the human part of technology is non-negotiable. We need to inject humanity into AI, and this begins with the data that feeds it and consideration for all facets of diversity.
The diversity and inclusion (D&I) crisis, especially in the US, is real, but fear is the worst reaction. We need to remain calm, approach D&I from different angles (human rights, mental health), and understand that those who abandon these agendas now will suffer in the future, losing innovation and creativity.
In a world where AI is rapidly advancing, the invitation is: spread this knowledge, question the standards, and help build technology that truly sees and includes each of us. Change begins with our humanity and the human aspect of technology. Let's transform communication and technology so that inclusion becomes the rule, not the exception!
PUBLICIDADE (GAM) — 1250x250
