"Fake news detection" AI is more likely to fail in the Global South, new study shows

The article discusses the challenges faced by Indian journalists in detecting political deepfakes during the largest democratic election in history. It highlights biases in AI detection tools, which predominantly train on datasets from the Global North, resulting in lower accuracy for content in regional Indian languages. The researchers emphasize that the algorithms reflect cultural imperialism, sidelining perspectives from the Global South. A recent paper critiques popular AI models for 'fake news detection,' revealing their ineffectiveness in diverse cultural contexts, thus necessitating a reevaluation of these technologies to ensure equitable media representation.

As opposed to investigating deepfake detection models, the study looks more broadly at AI models used for "fake news detection." This highlights systemic biases in technology.

Global North hegemony - and thus, cultural imperialism - [is] substantively percolating into AI algorithms built to mitigate fake news, undermining global perspectives.

Read at Nieman Lab