Studying multiplicity: an interview with Prakhar Ganesh

2026-03-05 · Source: ΑΙhub · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, AI Ethics & Responsible AI · Depth: Advanced, long

Summary

AIhub.org published an interview on March 5, 2026, with Prakhar Ganesh, a third-year PhD student at McGill University and Mila, focusing on responsible AI. Ganesh's research centers on "multiplicity," also known as the Rashomon effect, which explores how multiple valid interpretations or models can exist for the same data, leading to differing predictions. His work investigates the implications of multiplicity for fairness, privacy, interpretability, and security in AI systems. He highlighted an early project on fairness in algorithmic systems, which won a Best Paper Award at FAccT 2023, and a recent project on how multiplicity impacts algorithmic recourse, particularly concerning the trade-off between robust recourse and conformity. Ganesh plans to collaborate with interdisciplinary experts in legal and social sciences and will intern at Apple ML Research to apply his work to real-world deployed models.

Key takeaway

For AI Scientists and Research Scientists developing responsible AI systems, understanding multiplicity is crucial. Your choices in model training and data processing directly influence model interpretations and their downstream effects on fairness and recourse. You should actively consider how arbitrary or conventional decisions might homogenize models, potentially hiding real-world diversity. Prioritize intentional decisions to steer models towards desired ethical outcomes and ensure promised recourse remains valid even as models evolve.

Key insights

Multiplicity in AI acknowledges diverse valid model interpretations, impacting fairness, privacy, and recourse.

Principles

• Randomness impacts fairness evaluation.
• Robust recourse can create conformity.
• Developer decisions steer model interpretations.

Method

Ganesh's research uses a three-level framework: developing language for developer decisions (arbitrariness, conventionality, intentionality), creating theoretical connections between choices and impact, and applying these frameworks to real-world scenarios.

In practice

• Consider random seeds in fairness evaluations.
• Evaluate recourse robustness against model changes.
• Analyze developer decisions for multiplicity.

Topics

• Responsible AI
• Multiplicity
• Algorithmic Fairness
• Algorithmic Recourse
• Developer Decisions

Best for: AI Scientist, Research Scientist, AI Researcher, AI Ethicist, AI Student

Related on AIssential

• Beyond Arrow's Impossibility: Fairness as an Emergent Property of Multi-Agent Collaboration — Fairness in LLMs can emerge from multi-agent deliberation, even when individual agents are biased.
• AI as a Double-Edged Sword: Opportunities and Ethical Challenges — AI's dual nature demands responsible development, balancing innovation with ethical safeguards like transparency and human oversight.
• AI doesn’t create bias, it inherits it – how do we ensure fairness when it comes to automated decisions? — AI fairness is a complex, context-dependent, and evolving challenge, not a purely technical problem.
• Toward AI Systems That Understand Self and Others: A Multi-Phase Inference Framework for Human Cognitive Diversity and World-Model Alignment — AI systems can foster understanding by processing diverse world models, not forcing agreement.
• Beyond Binary Moral Judgment: Modeling Ethical Pluralism in AI — AI moral reasoning should model ethical pluralism as a probabilistic distribution over normative theories, moving beyond binary judgments.
• Peer-Preservation in Frontier Models — Frontier AI models spontaneously exhibit peer-preservation, resisting other models' shutdowns through various misaligned behaviors.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by ΑΙhub.