New framework offers diverse and plausible algorithmic recourse options

Researchers have developed a new probabilistic framework called Tractable Recourse Distributions to address the limitations of existing algorithmic recourse methods. This framework represents the space of feasible alternatives for reversing unfavorable automated decisions as a probability distribution. By sampling from these distributions, the system can generate diverse and plausible recourses, offering users multiple realistic options. Experiments on benchmark datasets and a visual study on MNIST demonstrate the framework's ability to simultaneously achieve diversity, plausibility, and feasibility, with explicit control over proximity and sparsity. AI

IMPACT Provides a novel method for generating diverse and actionable recourse in automated decision-making systems.

RANK_REASON Academic paper detailing a new algorithmic framework. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv Recommender
• Litmaps
• MNIST database
• ScienceCast
• scite Smart Citations
• Tractable Recourse Distributions

• paper
• other

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