Mitigating Bias in Home Maintenance Recommendations: A Fairness-Aware Approach Using Memory-Based Collaborative Filtering

Recommendation systems, while beneficial, often perpetuate biases, leading to unfair outcomes. This study focuses on mitigating popularity and gender biases within a mobile application. Aiming to develop a fairness-aware system, the research utilizes the MovieLens 1M dataset, employing three approaches. First, a personalized recommendation algorithm was developed, combining collaborative filtering with cosine similarity and demographic filtering. Second, a fairness-aware re-ranking algorithm, Deterministic Greedy, was designed to balance item popularity and gender representation. Third, these algorithms were integrated into a prototype mobile application. Usability testing and unit testing confirmed the application's effectiveness and adherence to ISO 9126 guidelines. This research provides a practical solution for mitigating multiple biases, demonstrating the feasibility of creating more equitable and ethical artificial intelligence systems. The findings contribute to the broader discourse on fairness in artificial intelligence, offering a template for developing inclusive recommendation platforms.