To address these biases, Sony suggests adopting an automated approach based on the existing CIELAB color standard, which would eliminate the need for manual categorization. This approach aims to replace existing skin tone scales that focus only on light versus dark. However, it is emphasized that any new measures must maintain simplicity to be practically effective.
Key takeaways:
- Sony's research has found that AI algorithms tend to face biases when it comes to skin tones, particularly inaccuracies in detecting darker skin tones and certain hues.
- Existing skin tone scales, such as Google's 10-point Monk Skin Tone Scale and the Fitzpatrick scale, primarily focus on the lightness or darkness of skin tone, allowing many biases to go undetected.
- Sony's solution suggests adopting an automated approach based on the existing CIELAB color standard, which would eliminate the need for manual categorization and address the limitations of the Monk Skin Tone Scale.
- Common image datasets disproportionately feature individuals with lighter, redder skin tones, while underrepresenting those with darker, yellower skin tones, leading to reduced accuracy in AI systems.