De Nadai's study revealed that AI models, while improved, still show significant bias in evaluating job interview responses, particularly against men with Anglo-Saxon names.
De Nadai challenged the notion of bias-free AI recruiting, emphasizing that even without names, other indicators in language reveal someone's background.
The research utilized three contemporary AI models, analyzing how they rated responses across different interview questions and cultural associations linked to names.
She highlighted the ongoing interest in understanding AI bias by referencing earlier studies predominantly focused on older models like GPT-3.5.
Collection
[
|
...
]