CLIP4KIDS: Exploring AI Bias & Fairness with Multimodal Learning

About the Project:

CLIP4KIDS is a multimodal, web-based system created using the Python library Gradio. Accessible through standard web browsers, it features three interactive interfaces designed to offer students valuable insights into AI’s capabilities and ethical considerations.

Core Technology: CLIP-ViT-B32 Model

At the core of CLIP4KIDS is OpenAI’s CLIP-ViT-B32 model. This model enables the encoding of inputs in text and image modalities, allowing the computation of cosine similarity vectors. This unique capability quantifies relationships between text and images, facilitating functions like classifications, comparisons, and the visualization of associations.

Tinkerability as a Design Goal

CLIP4Kids is desgined with tinkerability in mind, encouraging users to experiment and explore the system. Key features supporting this goal include:

Immediate Feedback: CLIP4KIDS provides immediate feedback, displaying and updating results in real-time. Users can quickly observe the results of their actions, fostering a responsive and interactive learning experience.
Fluid Experimentation: CLIP4KIDS supports fluid experimentation, making it easy for users to initiate exploration and engage in iterative experimentation by connecting, disconnecting, and reconnecting with their ideas.

Scaffolded Learning Experience

CLIP4KIDS follows a scaffolded approach, as each interface uses a similar but slightly more sophisticated version of the preceding interface, gradually increasing learner responsibility.

Interface 1 (Real Time Word Cloud): Users are introduced to associations in AI word clouds using word and image associations.
Interface 2 (Zero Shot Classification): Users are introduced to potential biases in model predictions based on user-defined labels.
Interface 3 (Smart Door Lock): Users are presented with a fairness task, where they decide how to use a smart door lock including defining positive/negative classes and set acceptance threshold.

Links and Resources:

CLIP4KIDS Installation: Clone the repository and run it in your local machine to see CLIP4KIDS in action.
Classroom Activities: Explore classroom activities designed to facilitate learning about AI bias and fairness using CLIP4KIDS.
Research Instruments: Access our research instruments used in evaluating CLIP4KIDS’ effectiveness in promoting children’s understanding of AI bias and fairness.

Acknowledgements:

This material is based upon work supported under the AI Research Institutes program by National Science Foundation and the Institute of Education Sciences, U.S. Department of Education, through Award # 2229873 - National AI Institute for Exceptional Education. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation, the Institute of Education Sciences, or the U.S. Department of Education.