Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systems

2026-09-27

AI Summary

This study addresses the causal-driven algorithmic harms and endogenous risks introduced by integrating generative models into recommender systems, which remain uncaptured by existing taxonomies. To bridge this gap, this work extends the harm taxonomy for non-conversational recommender systems by constructing novel threat models that account for the inductive priors and input misalignment inherent to generative models. Furthermore, it proposes an expanded classification framework encompassing endogenous harms such as “purification,” alongside a causal inference analytical framework designed to elucidate the formation mechanisms underlying undesirable input-output distributions. Ultimately, this research establishes standardized criteria for harm identification in generative-augmented recommender systems, providing system designers with a robust foundation for proactive risk warning and effective mitigation strategies.

Abstract

In this work, we consider algorithmic harms that may arise as generative models are incorporated into machine learning platforms. We argue that existing harm taxonomies and threat models require extension to (1) address novel causal drivers of well-studied representational and quality-of-service harms; and (2) anticipate and mitigate endogenous harms, such as sanitization, which may arise when system inputs are misaligned with the system designer's objectives, or the generative model's inductive priors. To this end, we introduce an expanded taxonomy of algorithmic harms associated with the use of generative models in non-conversational recommendation systems. In addition, we offer a causal analysis of how problematic subsets of the (input, output) joint distribution can arise, in an effort to inform harms detection and mitigation efforts.

Problem

• Algorithmic Harms
• Generative Models
• Recommendation Systems
• Harm Taxonomy
• Endogenous Harms

Innovation

• Generative Models
• Recommendation Systems
• Algorithmic Harms
• Causal Analysis
• Harm Taxonomy