
The widespread adoption of generative AI risks increasing the reproduction of stereotyped digital content and, therefore, the spread of gender biases to a broader audience. This reflects a growing interest in addressing children’s AI literacy and mitigating the potential negative impact of biased AI-generated content on their well-being. CARE-Kids tackles this challenge by empowering children (aged 8 to 12) to think critically about AI-generated content, educating them to recognise stereotypes in these images and helping them to become aware of their beliefs about themselves and others. CARE-Kids aims to deconstruct stereotypes perpetuated by biased AI algorithms (and society) by fostering children’s active engagement to mitigate its impact on their lives.
We will develop a web app to enhance children’s critical thinking and awareness of gender biases enabling children to interact with AI-generated content and understand gender biases. The project’s innovation lies in raising children’s awareness of gender biases by using AI-generated images with the aim of:
- Conducting research to identify existing gender biases in AI-generated images.
- Improving understanding of how to promote critical thinking and gender bias recognition.
Following a participatory design approach, CARE-Kids will answer to two main questions:
- How does the app enhance children’s critical understanding and recognition of gender biases in AI-generated images?
- In what ways does the app contribute to broader efforts to promote digital equity and meaningful digital inclusion and AI literacy in schools located in marginalised areas?
Through collaborative partnerships and a holistic approach with local communities, we envision creating a lasting impact on children’s digital literacy and societal attitudes by providing a meaningful experience. The interdisciplinary team is composed of a computer scientist expert in AI and a child-computer interaction scholar. The team will cover all the skills needed to achieve the project objectives.
Outcomes
Building on the success of the initial programme, CARE-Kids has been extended to allow further workshops in the two participating schools, further co-design sessions with the teachers and refinement of the prototype and the storytelling generative AI tool.
Phase one outputs include:
- Literature review on gender stereotypes on media for kids, AI literacy, biases on generative AI, critical thinking and AI literacy, generative AI tools for kids
- The CareKids tool: Image Generative AI tool to be used by children
- A questionnaire to assess children gender stereotypes knowledge and endorsement (Wood et al. 2022)
- A dataset of 216 images created by researchers using a standard generative AI tool and classified as classified by 111 adults as feminine/masculine/neutral
- A dataset of 1,789 mages created by children using CareKids tool associated with female/male traits, emotions and hobbies, classified by 4 experts as feminine/masculine/neutral (algorithmic auditing)
- A paper accepted to ACM CHI conference
Team
Dr Elisa Rubegni – Senior Lecturer, Lancaster University, School of Computing and Communication