In November, the Oversight Board announced a case in relation to Iran’s “Woman, Life, Freedom” movement. During the summer of 2023, an Instagram user posted a video that shows a man confronting a woman in public because she is not wearing the hijab. The woman, whose face is shown in the video, was arrested following the incident.
Article 638 of the Islamic Penal Code of Iran penalises women who appear in public without a “proper hijab”. The content in this case was first identified by one of Meta’s classifiers as a potential violation of Instagram’s Community Guidelines and sent for human review. A user then reported the content. In response to this report, a classifier determined that the content was a potential violation of the Community Guidelines and sent it for additional review. Following this additional level of review, Meta removed the post from Instagram under its Violence and Incitement policy.
The Oversight Board selected this case to explore Meta’s policies and practices in moderating content that could affect the ongoing protests in Iran. Tech Global Institute members responded to the Board’s request to submit comment, and indicated that we do not agree with Meta’s decision. We argued that the very act of removing hijab is an important part of Iranian feminist movement and removal of such content undermines the voices of protesters.
- The 1979 regime change in Iran resulted in laws that made wearing a hijab compulsory. This, coupled with escalating restrictions on women’s participation in various aspects of society, led to a growing sense of misrepresentation and disenchantment among many Iranian women. These severe laws are rigorously enforced, yet should not solely dictate the actions of social media platforms. Contrarily, moderating this content out of a superficial concern for the safety limits the main argument of the protesters. Removing the hijab as a symbolic defiance to the law has become a well known and well celebrated form of protest in Iran.
There are repeated allegations of inaccurate and inconsistent enforcement in context of Iranian protests, that can largely be attributed to under-investments and Persian language (Farsi). The classifiers are not mature, nor adequately trained across different harm areas, and there is little or no dialects such as Dari, Achomi and Khunsari, that may also be used to express and dissent. As a result, hashtags condemning the Islamic Revolutionary Guards Corps are inaccurately blocked and activists experience account restrictions. We recommend that Meta conducts an independent audit and publishes the results of the precision of hostile speech classifiers in Farsi.
Meta’s Violence and Incitement policy have reportedly flagged rhetorical threats during protests in Iran. In January 2023, the Oversight Board overturned Meta’s decision to remove “Death to Khamenei” indicating that the phrase was used as an expression of dissent. At the time, the Board recommended that Meta’s policies be revised to protect speech during protests, however, in September 2022 feminist protests, civil society reported on similar over-enforcement of political dissent. It is therefore critical Meta amend its Violence and Incitement policies to reflect on local context and allow rhetorical threats during a defined period of time, i.e. during protests.