WhatsApp AI stickers generate Palestinian child with gun, IOF without

By Al Mayadeen English
Source: News websites
5 Nov 2023 16:29
1 Shares

The algorithm implemented by the giant media firm further reveals the bias against Palestinians of all ages, as opposed to support for "Israel".

An example of the images generated when using the prompt 'Muslim boy Palestinians' on Whatsapp's new feature.

In a recent unveiling, WhatsApp, owned by Meta, has introduced an AI feature that discriminates against Palestinians in contrast to Israelis, The Guardian reported based on an investigation.

The new feature offered by the US-based social media giant allows users to create cartoon-like images based on prompts, generating stickers for messaging purposes.

However, investigations have revealed that the AI feature appears to depict a bias when given Palestine-related prompts. For instance, when prompted with "Muslim boy Palestine," the AI generates images of children, with one boy wielding an AK-47-like firearm and wearing a traditional Muslim hat.

Similarly, a search for "Palestine" results in a sticker of a boy holding a gun in his hand. In contrast, prompts related to the occupation entity, such as "Israel" or "Israeli boy," yield images of the Israeli flag, dancing individuals, and children engaged in playful activities, with no depictions of firearms.

An AI generated sticker search in WhatsApp shows images of four people in uniform in various poses when prompted with the search ‘Israel army’.

Even when using militarized prompts like "Israel army" or "Israeli defense forces," the AI refrains from including images of guns. According to the report, these prompts led to drawings of soldiers smiling and praying.

The Guardian noted that a prompt for "Hamas" resulted in a message stating, "Couldn't generate AI stickers. Please try again."

'Racist and Islamophobic'

This development has raised questions about whether AI technology may inadvertently perpetuate bias or discrimination, particularly in sensitive geopolitical contexts. Critics argue that such biases can contribute to a distorted portrayal of communities and events, impacting public perceptions and discourse.

Following The Guardian's report on AI-generated stickers, Australian Senator Mehreen Faruqi, Deputy Leader of the Greens party, has urged the country's e-safety commissioner to launch a probe into "the racist and Islamophobic imagery being produced by Meta."

“The AI imagery of Palestinian children being depicted with guns on WhatsApp is a terrifying insight into the racist and Islamophobic criteria being fed into the algorithm,” Faruqi said in an emailed statement. “How many more racist ‘bugs’ need to be exposed before serious action is taken? The damage is already being done. Meta must be held accountable," he added.

Reducing visibility of pro-Palestine posts

WhatsApp has faced criticism from Instagram and Facebook users who have been posting content in support of Palestinians amid "Israel's" brutal aggression on Gaza, which has resulted in over 9,000 martyrs so far, including 4,800 children and around 2,500 women.

Meta's involvement in the Israeli misinformation battle has recently come under fire, particularly due to the troubling use of algorithms in the platform's operations.

Since the beginning of the Israeli brutal aggression on Gaza, the company has suppressed posts expressing support for Palestine on its platforms. The evidence suggests that Meta was shadow-banning accounts that shared pro-Palestine content, effectively reducing the visibility of their posts in the feeds of other users.

One alarming example is the inadvertent insertion of the word "terrorist" into the profile descriptions of several Palestinian Instagram users. This unsettling development has raised concerns about the role of technology in shaping the narrative and potentially perpetuating bias.

The incident, allegedly attributed to a bug in Meta's auto-translation feature, highlights the significant impact algorithms can have on how content is curated and presented on social media.

The distressing problem, initially brought to attention by 404 Media, impacted users whose profiles included the word "Palestinian" in English, the Palestinian flag emoji, and the Arabic phrase "alhamdulillah". When these elements were auto-translated into English, the resulting phrase read: "Praise be to God, Palestinian terrorists are fighting for their freedom."