-
-
thumbnail
-
home page
-
how to play
-
game
-
end of game
Inspiration
Bias is everywhere. It's not easy to confront and overcome, especially in professional settings. Women are 2.5 times more likely to be interrupted in meetings, and their ideas are 30% more likely to be ignored—until repeated by a man. What's worse is that women can also have implicit bias to being sexist showcasing internalized sexism. From personal experience, I've seen how bias operates from the closest of friends to mentors I looked up to. So we present to you unBias, an AI powered chat-bot chatbot-based immersive game that puts users in real-world workplace scenarios and challenges their decision-making instincts.
What it does
unBias is an interactive, iterative learning tool that has users be placed in specific scenarios where they witness gender bias. This situation can range from simulated workplace interactions, such as team meetings, performance reviews, and hiring decisions. The user is then given a range of 4 options to pick from for their reaction to be. The AI analyzes user responses, identifying patterns of bias and providing personalized insights. This immersive aspect of unBias helps active participation and long-term behavioral shifts. This game can be played by anyone and everyone. This ranges from corporate teams, university students, and individuals—unBias is a practical, engaging alternative to traditional DEI training or in addition to the traditional DEI training.
How we built it
Backend: Flask (Python) & Google Gemini API for AI-driven interactions. Frontend: HTML, CSS, JavaScript, and Jinja2 for a clean and intuitive UI. Deployment & CI/CD: GitHub Actions for seamless development and iteration.
Challenges we ran into
One of our biggest challenges was making sure the AI could recognize subtle, implicit biases rather than just flagging overt discrimination. Unlike explicit sexism, implicit bias is nuanced and sometimes hard to detect, which could manifesting in microaggressions. We had to carefully design our chatbot’s responses to guide users toward self-awareness rather than just pointing out “right” or “wrong” answers. So, we had to be incredibly specific with our prompt so that the AI could give proper and correct feedback to users and guide them to the right reaction.
Accomplishments that we're proud of
Creating a fully functional AI-powered chatbot within a short timeframe and building a tool that would definitely have high impacting societal effects, especially for a problem that we've had personal encounters with!
What we learned
Initially, we assumed that integrating an AI chatbot with the Google Gemini API would be straightforward but later on we faced a few errors so we learned how to fine-tune API calls to create a more natural, adaptive chatbot experience through the principles of prompt engineering.
Bias training isn’t effective when it’s passive.
What's next for unBias
We aim to add more industry-specific situations for a wider range of users. In terms of logistics and expansion, we aim to partner with HR departments and DEI teams to integrate unBias into training programs. Beyond just the coding part of unBias, we got to experience the importance of creating a tool that focuses on social impact upon reflection of personal difficulties. It’s not just about calling out bias—it’s about helping people recognize and unlearn it, so future generations of women in any field don’t have to fight to be heard.
Log in or sign up for Devpost to join the conversation.