Guide to AI: The cycle of harm

What is the ‘cycle of harm’?

This chart was developed to explain why gender based violence keeps happening and how each part of the problem feeds into the next. It breaks the issue into seven stages and shows where action is most needed to break the cycle.

Data and legislation can only achieve so much. Alongside these practical changes, we also need a cultural shift that challenges misogyny and harmful behaviours.

Datasets are missing women leads to gender bias leads to harmful stereotypes and inequality leads to gender based violence and abuse, leads to ineffective safety measures, leads to poor representation in policy, leads to under-representation in AI development. And so the cycle renews.

Gender bias

The cycle begins with widespread gender bias and the underlying attitudes that shape how women and girls are treated in everyday life.

Harmful stereotypes and inequality

Gender bias continues to cause harmful attitudes towards women and girls

Gender-based violence and abuse

This environment enables gender based violence, including online harm like image-based sexual abuse and digitally manipulated synthetic content.

Ineffective safety measures

While tools like safety devices and safety apps exist, they don’t address the root causes and are reactive, not preventative. They put the onus on the victims to protect themselves instead of reducing the risk of harm in the first place.

Poor representation in policy

Policy decisions often do not fully address current risks to women and girls. This is why we need inclusive policymaking that properly reflects their needs and experiences.

Under-representation in AI development

Women are often missing from tech development, meaning the risks they face aren’t built into design. This is why we need diverse teams shaping future technologies.

Datasets are missing women

We don’t have enough data on women’s experiences. This is why we need better research and evidence to better inform decision making.