DeBiasMe for bias-aware AI use

DeBiasMe

DeBiasMe for bias-aware AI use

AI Literacy Tool for human and AI biases, building better critical thinking skills for all.

London, UK

GBR

Nonprofit

AI tools are changing how students learn, but they prioritize convenience over critical reflection, negatively impacting their critical thinking skill development. This creates critical educational challenges: while nearly a third of children aged 8 or younger are using AI for learning, they can develop problematic "human bias" – becoming overly reliant on AI outputs or easily trusting information that confirms existing thoughts. What's more concerning is how students outsource higher-level thinking to AI, having it analyze, evaluate, and create for them. Research shows this weakens critical thinking skills at a time when effective AI collaboration is becoming essential. This creates an AI literacy gap, widening existing educational inequalities that disproportionately affect students with limited AI competency and insufficient learning supports. Current K-12 AI education focuses primarily on technical skills while overlooking how learners' biases can shape AI interactions. Addressing this literacy gap requires AI literacy tool focusing on metacognitive skills - the ability to monitor, evaluate, and regulate one’s own learning and decision-making processes - equitable educational outcomes, while empowering students’ learning and development in the AI age.

DeBiasMe reimagines traditional LLM-based chatbot interfaces by introducing strategic pause points ("frictions") in the student-AI interaction flow. This bi-directional approach addresses both how students formulate questions (input stage) and interpret AI responses (output stage). Our design solution creates intentional moments for reflection on human and AI biases, helping students develop critical thinking skills and metacognitive abilities. 1. Input-Stage Friction (Prompt Refinement Tool): Before students submit questions to AI, our tool uses visual cues to highlight potential biases in their questions. If a student asks, "Why are video games better than books?", our tool shows a "one-sided thinking" and suggests alternative phrasing like "How do books and video games affect learning differently?" Real-time feedback helps learners recognize when they're starting with assumptions rather than open questions. 2. Output-Stage Friction (Bias Visualization Map): After receiving an AI response, students interact with a visual map showing potential biases in the AI's answer. The map shows visual connections between different parts of the AI's response and potential biases through interactive network diagrams mapping different bias types to specific text segments in AI outputs. This helps students break down AI responses and evaluate their quality, reliability, and limitations. Product demo: https://youtu.be/E4oszAKsoTU

DeBiasMe empowers students to develop self-awareness about their thinking patterns when using AI, addressing critical gaps in current AI literacy efforts while supporting equitable educational outcomes. Our solution helps learners develop the metacognitive skills essential for academic success and lifelong learning in our AI-integrated world, benefiting multiple stakeholders: Diverse Learners: DeBiasMe benefits learners with differing levels of competence with AI, particularly those from underserved communities with limited digital access or language barriers. Our planned co-design session can further ensure the tool meets diverse children's needs through a participatory and community engagement approach. Educators/Educational Institutions: We provide teachers with practical tools/strategies to promote higher-level thinking (analysis, evaluation, creation) while maintaining AI benefits. The visualization helps educators explain abstract bias concepts concretely and can be applied to existing curricula across subjects (history, literature, science, etc). Our solution transforms passive AI consumption into active, reflective collaboration, aligning AI use with pedagogical approaches. Scientific/Creative Collaboration: AI safety researchers and data scientists researching human-AI bias interactions can gain valuable insights from our implementation of the tool. Creative collaborations with artists can explore bias-related themes through experimental visualizations. Policy Implications: Our frictional design approach influences broader conversations about thoughtful AI integration in education.