Recommender systems use algorithms to suggest items to users. The most two common approaches are
- content-based filtering: recommend items to you based on features of user and items, which largely falls under supervised learning
- collaborative filtering: recommend items to you based on ratings of users who gave similar ratings as you, which largely falls under unsupervised learning
Utility matrix
Two main entities in recommender system is users and items. Each user has a different degree of preference for each item, usually measured by user rate for that item.
Utility matrix summarizes all these ratings, including unknown ratings. recommender system aims to predict user ratings and give suggestions.
Ethical considerations
| Problems | Amelioration |
|---|---|
| To maximize profits, some companies using recommender system to recommend the most popular, profitable products, instead of products of user preference. | - Be transparent with users - Analyze the trade-off |
| Maximizing user engagement has led to large SMS, video sharing sites to amplify misleading information, hate, toxicity, etc | Filter out problematic content (by AI or human) |