As the Artificial Intelligence (AI) sector rapidly evolves, the need for specialised training and education has become critical. The DIVERSIFAIR project, funded under the Erasmus+ scheme, aims to address two primary concerns: the shortage of AI experts skilled in recognising and addressing algorithmic bias, and the necessity of incorporating an intersectional approach to fairness and equality in AI. Our recent survey, conducted as part of the DIVERSIFAIR project, sheds light on the current state of awareness and perceptions of intersectional bias in the AI community.
Why explore intersectionality in AI?
Intersectional bias refers to the convergence of various forms of discrimination, such as racism, sexism, ableism, and colonialism, which impact individuals with multiple social identities. In AI systems, this translates to unfair treatment of those facing multiple forms of discrimination.
Key Findings
The survey, conducted in Spring 2024, surveyed the AI community at large, including professionals and practitioners. Here are some of the key insights:
Insufficient Awareness of Intersectional Bias
While 59% of respondents reported experiencing discrimination based on multiple characteristics, only 15% were very familiar with the concept of intersectional bias.
Despite high familiarity with AI technologies (85.14%), there is limited engagement with intersectional bias issues, with 43.6% of AI-savvy respondents minimally involved in addressing these biases.
Recognition of Bias Risks
A significant majority (86.3%) acknowledged the risk of AI systems perpetuating existing biases, particularly related to gender, race, and socioeconomic status.
Interestingly, 58% of those who disagreed with this risk worked in the technology sector, highlighting a disconnect between developers and broader societal awareness of AI’s impact on bias.
Public Awareness and Education
An overwhelming 87% believe there is insufficient public awareness and education about intersectional bias in AI. Contributing factors include unfinished legal structures, lack of diversity in the AI sector, and insufficient training among AI experts.
48% of respondents are optimistic about AI’s potential to increase fairness, while 26% are neutral and another 26% are pessimistic. This mixed sentiment reflects cautious hope tempered by current challenges in addressing bias.
Looking Ahead
The survey underscores the urgent need for ongoing education and awareness efforts within the AI community regarding intersectional bias. While there is cautious optimism about AI’s potential to promote equality, continuous learning and engagement are crucial. We must prioritise fairness and mitigate bias in all AI applications to build trust and ensure equitable AI systems.
Thank you to everyone who participated in the survey and contributed to this important work. Stay tuned for more updates and ways to get involved in DIVERSIFAIR’s ongoing efforts to address intersectional bias in AI.
