Epistemic Harms
Algorithmic recommender systems reinforce and amplify anthropocentric bias or desire of some people for animal cruelty as entertainment — leading to greater harm to animals through reinforcement of meat eating from factory farms, cruel uses of animals for entertainment, etc
ENTITY
2 - AI
INTENT
2 - Unintentional
TIMING
2 - Post-deployment
Risk ID
mit679
Domain lineage
6. Socioeconomic and Environmental
6.6 > Environmental harm
Mitigation strategy
1. Redesign Algorithmic Objective Functions to Integrate Ecocentric and Non-Anthropocentric Fairness Constraints: Implement debiasing techniques on training data and incorporate novel objective functions that actively track and minimize the systemic reinforcement and recommendation of content promoting animal cruelty, factory farming, or other environmentally detrimental human-centric behaviors, moving beyond solely human-centered notions of harm and utility. 2. Strengthen Content Moderation Protocols for Amplified Harmful Content: Mandate the classification of the recommendation and amplification of content depicting, glorifying, or normalizing animal cruelty (including non-documentary footage of factory farming or cruel uses of animals for entertainment) as a high-severity systemic risk. This requires enhanced, context-aware AI-assisted detection and immediate suppression from all discovery and recommender surfaces to disrupt the harmful feedback loop. 3. Establish Independent, Ecologically Focused Systemic Risk Assessment and Audit Mandates: Require comprehensive, periodic systemic risk assessment reports that explicitly measure the recommender system's contribution to the reinforcement of anthropocentric biases and potential environmental harms. These reports must be subjected to mandatory, independent third-party audits with a focus on metrics related to non-human welfare and ecological integrity, ensuring public accountability and long-term risk mitigation.