2025 Diversity Project Presentation

On April 7, 2025, from 3:00 p.m. to 5:00 p.m. at the Big Data Hub Presentation Studio, we showcased the amazing projects our students are working on to advance Diversity, Inclusion, Equity, and Justice in Computing Science Research and Professional Practices.

The event began with a brief introduction to the CSDC Diversity Awards, followed by project presentations. We continued with a break for networking and gourmet snacks, and then introduce the winning project.

Our shortlisted projects for 2025 CS Diversity Award were:

Joint Winner of the 2025 CS Diversity Award

Project Title:  Predictive Farming: AI-Driven Crop Yield Optimization
Team members:  Yasmine Gouin (she/her), Fitzpatrick Laddaran (he/him), Nolan Isaac-Smith (he/him), Chi Nguyen (she/her)
Advisor:  Steven Bergner

Description: Canadian farmers, particularly those in rural communities, face significant challenges due to geographic isolation, systemic inequities, and climate change. These communities often struggle with limited access to essential services and rely heavily on climate-sensitive agriculture. Our intention is to help farmers across Canada make informed decisions on planting, harvesting, and selling by leveraging advanced data analytics and predictive modelling to provide easily accessible and actionable insights on crop yields and pricing. This initiative strengthens the resilience of marginalized rural communities. Beyond supporting farmers, the project has broader implications on Canada’s economy and food supply chain, promoting long-term sustainability and food security.

Joint Winner of the 2025 CS Diversity Award

Project Title: Intelligent Guiding Robot System for Blind and Visually Impaired People Enabling Independent Navigation
Team members:  Qihan Gao, Shaojun Cai, Kwek Bin Chong
Advisors:  Lawrence Kim, David Hsu

Description: Blind and visually impaired individuals continue to face significant barriers to independence due to limitations in current accessibility technologies. Our project addresses this gap by developing an intelligent navigation assistant system for robotic guide platforms. The system integrates a voice interface, a reasoning backbone for planning and decision-making, and modular functions such as object detection, text reading, and environmental description. Deployed on a quadruped robot, the system enables users to navigate complex public environments with greater autonomy. Field tests demonstrate its potential to support blind users in performing everyday tasks more independently. This work exemplifies how advanced AI and robotics can be developed to support inclusive design, ensuring that emerging technologies are accessible and empowering for everyone.

Winner of the People's Choice Award at the 2025 CS Diversity Awards

Project Title: Advancing Equity in Health AI: Improving Dermatological Image Datasets for Fair and Inclusive Healthcare Solutions
Team members: Kumar Abhishek, Aditi Jain
Advisors: Ghassan Hamarneh

Description: Deep learning-based systems for dermatological diagnosis rely on high-quality, unbiased datasets, since flawed data can lead to misdiagnoses and unequal care for underrepresented groups. Datasets like HAM10000 and Fitzpatrick17k are widely used to train AI models, but issues such as data leakage, mislabeling, and duplicates can perpetuate biases, particularly against patients with darker skin tones or rare diseases. Our research uncovered critical issues in thse widely-used datasets, including train-test data leakage, duplicate images with conflicting skin tone labels (varying by up to 4 points on the 6-point Fitzpatrick scale), and mislabeled rare diseases — errors that may disproportionately harm patients with darker skin or uncommon conditions. These issues artificially inflate model performance while masking bias risks. Our work advances EDIJ efforts by directly addressing these hidden biases in AI datasets, which disproportionately affect marginalized groups. By improving dataset quality, we promote equitable healthcare solutions and foster trust in AI technologies.

Project Title: Facilitating Bids for Conversation Entry using ConvoBuddy, A Chatbot Conversation Assistant

Team Members: Sarah Jade Pratt and Kenny Zhang 
Advisor: Xing-Dong Yang

Description: Facilitating smooth entry into ongoing conversations is essential for fostering inclusive and enjoyable social interactions, yet it remains one of the most challenging aspects of conversational dynamics. We present ConvoBuddy, a novel conversational assistant designed to support users actively seeking to join existing conversational groups based upon a preliminary need-finding user study and literature review. The aim of ConvoBuddy is to level the playing field when it comes to navigating informal networking and social scenarios for individuals with social difficulties or anxiety. ConvoBuddy leverages smartphone-based sensors to detect conversational group spatial formations in real time and incorporates entry timing and turn-taking management. By adopting a low-barrier, software-only approach, we aim to make ConvoBuddy easily accessible and readily available to foster social connection among diverse communities.

Project Title: Tensions in Learning About Generative AI: Investigating How Artists Use Tutorials vs. Self-Experimentation for Exploring Image Generation Tools
Team members: Afra Liu, Isabelle Kwan, Taiga Okuma, Jeffery Loverock
Advisors:  Nicholas Vincent, Parmit Chilana

Description:  With the rapid rise of generative AI tools for image creation, there’s an urgent need to ensure more equitable use of these technologies. While these tools promise to democratize creativity, many rely on tutorials that assume technical fluency—leaving out users from non-technical backgrounds. We explore how visual artists and hobbyists learn to use generative AI, focusing on the tension between structured tutorials and self-experimentation. Our survey of 159 artists revealed that many found existing tutorials too complex to follow, especially without a tech background. To address this, we created Peek-Box, an interactive, jargon-free tutorial that explains AI concepts using simple language and visuals. By reducing learning barriers, we aim to promote digital justice and support Equity, Diversity, Inclusion, and Justice (EDIJ), ensuring people with different skills, experiences, and needs can participate fully in AI-powered creative spaces.