Artificial intelligence (AI) may not be as effective in detecting signs of depression in the social media posts of Black Americans as it is in those of their white counterparts, according to a recent study.
The study, conducted by researchers in the U.S., revealed that an AI model was over three times less accurate in predicting depression in Black individuals on Meta Platforms Inc.’s Facebook compared to white individuals. These findings were published in the Proceedings of the National Academy of Sciences.
Previous studies had suggested that certain language patterns, such as the use of first-person pronouns, could indicate a higher risk of depression. However, this latest research found that these associations were only relevant for white individuals, highlighting a significant gap in understanding mental health cues across different racial groups.
Lead author Sharath Chandra Guntuku from the Center for Insights to Outcomes at Penn Medicine expressed surprise at the lack of consistency in language associations across racial backgrounds. While social media data cannot serve as a diagnostic tool for depression, it could potentially aid in assessing the risk of depression in individuals or communities.
AI biases have been a recurring issue in various technologies, with Meta’s chief AI scientist, Yann LeCun, emphasizing the impossibility of creating completely unbiased AI systems. This study’s findings add to a growing body of evidence showcasing the need for more inclusive and diverse technology solutions.
The research also underscores the significance of inclusive technology development, such as Google’s Real Tone, which aims to enhance the accuracy of cameras for a diverse range of skin tones.
This study serves as a reminder of the importance of creating technology that is sensitive to the diverse needs and characteristics of all individuals, particularly in critical areas such as mental health assessment. As advancements in AI continue, addressing biases and ensuring inclusivity must remain at the forefront of technological innovation.
