Snow White or snowflake? New AI tool analyses fairy tales for gender bias

Are Snow White, Cinderella and Sleeping Beauty about to be cancelled? A new AI tool analyses classic fairy tales looking for gender bias and stereotypes.

Researchers from Northeastern University, University of California Los Angeles and IBM Research created a framework that can analyse problematic storylines such as helpless princesses needing to be rescued.

Of the 33,577 events in existing fairy tales, 69% were attributed to male characters and 31% to female characters.

The events associated with female characters were often connected to domestic tasks like grooming, cleaning, cooking and sewing, while those for male characters were connected to failure, success or aggression.

The researchers hope that the AI-driven 'spellcheck-like' tool will enable writers and publishers to create more inclusive stories for children.

Read more: Five easy ways to help girls crush gender bias

Dakuo Wang, an associate professor at Northeastern said: "If in the future I have a baby girl, I don't want her to feel discouraged to take on those tasks or conquer those challenges - say, someone will come save me or it's not supposed to be something I would do as a girl.

"If we can develop a technology to automatically detect or flag those kinds of gender biases and stereotypes, then it can at least serve as a guardrail or safety net not just for ancient fairy tales but the new stories being written and created every day today."

The work started as part of the team's ongoing research into how AI can help build language learning skills for young children.

They recruited a group of educational experts to comb through the stories and create a list of questions and answers that would help prove whether a child was learning from these stories.

Watch: 17 times Disney totally changed the plot from the original fairy tales

Read more: A 1988 warning about climate change was mostly right

The end result was 10,000 question-answer pairs and the realisation that all of these stories, no matter where they came from, had "stubborn and profound" gender stereotypes in them.

The princess eats a poisoned apple, gets imprisoned, kidnapped or cursed or dies and has no way to change her situation.

Meanwhile, male characters were killing dragons, breaking curses and saving the princess.

The researchers focused on "temporal narrative event chains," the specific combination, and order, of events and actions a character experiences or takes.

Wang said: "It's actually the experience and the action that defines who this person is, and those actions influence our readers about what [they] should do or shouldn't do to mimic that fictional character."

"Someone is being saved and then getting married and then living happily ever after; some others killed the monster, saved the princess and lived happily ever after.

"It's not the 'lived happily ever after' part or 'get married' part that are different. It's actually the events happening before these events in a chain that make a difference."

By automating this process, Wang said he hopes the tool will find use among people outside the research community who are actually creating these stories.

In the process, they can start preventing stories from passing down these outdated, harmful ideas to the next generation.

"With our tool, they can simply upload their first draft into a tool like this and it should generate some score or meter that indicates.

"Here are the things you may or may not want to check. If this intention is not what you would want to express, then maybe you should think about a rewrite. Here are some suggestions."

Moving forward, Wang and the team plan on expanding their work to look at other forms of bias.