Research evaluates energy cost of fairness in recommender systems

A new research paper published on arXiv explores the trade-offs between accuracy, fairness, and energy consumption in recommender systems. The study investigates how different fairness interventions, such as in-processing, graph-level reweighting, and post-processing methods, impact computational costs and environmental footprint. The findings indicate that the energy cost of fairness is not uniform, with post-processing methods shifting costs to inference time, while other methods vary significantly based on the model, dataset, and hardware used. The research calls for a three-way evaluation of recommender systems, considering accuracy, fairness, and computational cost. AI

IMPACT Highlights the need to consider energy efficiency alongside accuracy and fairness in AI system design.

RANK_REASON The cluster contains a research paper published on arXiv detailing a new evaluation methodology for recommender systems. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• Accuracy
• alphaXiv
• arXiv
• CatalyzeX
• CORE Recommender
• DagsHub
• energy consumption
• Gotit.pub
• Hugging Face
• Influence Flower
• Recommender Systems
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →