Ashudeep Singh

Areas of Interest: AI Safety & Alignment, Responsible AI, Large Language Models, Machine Learning, Information Retrieval.

I am a Principal Applied Scientist at Microsoft AI, focusing on agentic AI for search systems. My work develops safe, human-aligned multimodal LLMs that learn from feedback while ensuring trustworthy deployment at scale.

Previously, I worked at Pinterest Labs, pioneering bias evaluation frameworks and red-teaming methodologies for LLMs and image generation models in production. I also established Pinterest’s fairness evaluation and mitigation frameworks across recommendation systems. Prior to Pinterest, I completed my PhD at Cornell University under Thorsten Joachims, establishing foundational work in fairness for ranking and recommender systems. I also completed research internships at Google Brain, Facebook Research, and Microsoft Research.

My research bridges AI safety theory with practical deployment, publishing at top venues (NeurIPS, ICML, SIGIR, KDD, FAccT) while shipping state-of-the-art responsible AI systems serving hundreds of millions of users. I also serve as Area Chair for NeurIPS and ICML, and co-taught the NeurIPS 2022 tutorial on responsible ML.

news

May 17, 2024	Paper on measuring user impact of diversification in recommendations titled “Inclusive Recommendations and User Engagement: Experimental Evidence from Pinterest” accepted at ACM EC 2024.
Oct 17, 2023	Presented a guest lecture for the Operations Management class (BUAD 311) at USC Marshall School of Business on “Responsible ML for Real-World Search and Recommender Systems: A Multistakeholder Perspective” (slides).
Sep 9, 2023	Ongoing research with Madhav Kumar(MIT) on the impact of diversification in recommender systems on its users was recently accepted to Workshop on Information Systems and Economics (WISE) 2023.
Jun 15, 2023	Our work on interpretability for recommender systems called “RecRec: Algorithmic Recourse for Recommender Systems” was accepted to ACM CIKM 2023. This is joint work with co-authors from University of Washington and University of Maryland. Here is the arXiv link for the paper.
Jun 11, 2023	Our recent work on diversification in search and recommender systems across the Pinterest platform was accepted to ACM FAccT 2023. See you in Chicago! It was recently also published on the Pinterest Engineering blog.
Dec 5, 2022	The slides to the NeurIPS 2022 tutorial on Fair and Socially Responsible ML for Recommendations are public. Download them here. Visit the NeurIPS portal to access the recording video if you are registered.
Sep 28, 2022	Hannah Korevaar (Meta), Manish Raghavan (MIT), and I are presenting a tutorial at NeurIPS 2022 on Fair and Socially Responsible ML for Recommendations.
Sep 29, 2021	Our paper “Fairness in Ranking under Uncertainty” has been accepted to NeurIPS 2021. This is joint work with my advisor Thorsten Joachims (Cornell) and David Kempe from USC. 📄

selected publications

Find more information on the publications page and google scholar.

* contributed equally.

KDD

Fairness of Exposure in Rankings

Ashudeep Singh, and Thorsten Joachims

In In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, London, UK, 2018.

Abstract arXiv

Rankings are ubiquitous in the online world today. As we have transitioned from finding books in libraries to ranking products, jobs, job applicants, opinions and potential romantic partners, there is a substantial precedent that ranking systems have a responsibility not only to their users but also to the items being ranked. To address these often conflicting responsibilities, we propose a conceptual and computational framework that allows the formulation of fairness constraints on rankings in terms of exposure allocation. As part of this framework, we develop efficient algorithms for finding rankings that maximize the utility for the user while provably satisfying a specifiable notion of fairness. Since fairness goals can be application specific, we show how a broad range of fairness constraints can be implemented using our framework, including forms of demographic parity, disparate treatment, and disparate impact constraints. We illustrate the effect of these constraints by providing empirical results on two ranking problems.
NeurIPS

Policy Learning for Fairness in Ranking

Ashudeep Singh, and Thorsten Joachims

In Neural Information Processing Systems (NeurIPS), 2019.

Abstract arXiv Code Website

Conventional Learning-to-Rank (LTR) methods optimize the utility of the rankings to the users, but they are oblivious to their impact on the ranked items. However, there has been a growing understanding that the latter is important to consider for a wide range of ranking applications (e.g. online marketplaces, job placement, admissions). To address this need, we propose a general LTR framework that can optimize a wide range of utility metrics (e.g. NDCG) while satisfying fairness of exposure constraints with respect to the items. This framework expands the class of learnable ranking functions to stochastic ranking policies, which provides a language for rigorously expressing fairness specifications. Furthermore, we provide a new LTR algorithm called Fair-PG-Rank for directly searching the space of fair ranking policies via a policy-gradient approach. Beyond the theoretical evidence in deriving the framework and the algorithm, we provide empirical results on simulated and real-world datasets verifying the effectiveness of the approach in individual and group-fairness settings.
SIGIR Best Paper Award

Controlling Fairness and Bias in Dynamic Learning-to-Rank

Marco Morik*, Ashudeep Singh*, Jessica Hong, and Thorsten Joachims

In Proceedings of ACM SIGIR Conference on Research and Development in Information Retrieval, 2020.

Abstract arXiv Code Slides Website

Rankings are the primary interface through which many online platforms match users to items (e.g. news, products, music, video). In these two-sided markets, not only the users draw utility from the rankings, but the rankings also determine the utility (e.g. exposure, revenue) for the item providers (e.g. publishers, sellers, artists, studios). It has already been noted that myopically optimizing utility to the users – as done by virtually all learning-to-rank algorithms – can be unfair to the item providers. We, therefore, present a learning-to-rank approach for explicitly enforcing merit-based fairness guarantees to groups of items (e.g. articles by the same publisher, tracks by the same artist). In particular, we propose a learning algorithm that ensures notions of amortized group fairness, while simultaneously learning the ranking function from implicit feedback data. The algorithm takes the form of a controller that integrates unbiased estimators for both fairness and utility, dynamically adapting both as more data becomes available. In addition to its rigorous theoretical foundation and convergence guarantees, we find empirically that the algorithm is highly practical and robust.
FAccTRec

Building Healthy Recommendation Sequences for Everyone: A Safe Reinforcement Learning Approach

Ashudeep Singh, Yoni Halpern, Nithum Thain, Konstantina Christakopoulou, Ed H. Chi, Jilin Chen, and Alex Beutel

In FAccTRec Workshop at ACM RecSyS, 2020.

Abstract PDF Code Slides Website

A key consideration in the design of recommender systems is the long term well-being of users. In this work, we formalize this challenge as a multi-objective, safe reinforcement learning problem, balancing positive user feedback and the “healthiness” of user trajectories. We note that in some cases, naively balancing these objectives can lead to unhealthy experiences, even if unlikely, still occurring in a small subset of users leading us to examine a distributional notion of recommendation safety. Thus, we propose a reinforcement learning approach that optimizes for positive feedback from users while simultaneously optimizing for the health of worst-case users to remain high. To empirically validate our method, we develop a novel research simulation environment motivated by a movie recommendation setting that considers exposure to violence as a proxy for unhealthy recommendations. We demonstrate how our method reduces unhealthy recommendations to the most vulnerable users, without sacrificing much user satisfaction.
FAccT

Representation Online Matters: Practical End-to-End Diversification in Search and Recommender Systems

Pedro Silva, Bhawna Juneja, Shloka Desai, Ashudeep Singh, and Nadia Fawaz

In Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency, 2023.

Abstract arXiv

As the use of online platforms continues to grow across all demographics, users often express a desire to feel represented in the content. To improve representation in search results and recommendations, we introduce end-to-end diversification, ensuring that diverse content flows throughout the various stages of these systems, from retrieval to ranking. We develop, experiment, and deploy scalable diversification mechanisms in multiple production surfaces on the Pinterest platform, including Search, Related Products, and New User Homefeed, to improve the representation of different skin tones in beauty and fashion content. Diversification in production systems includes three components: identifying requests that will trigger diversification, ensuring diverse content is retrieved from the large content corpus during the retrieval stage, and finally, balancing the diversity-utility trade-off in a self-adjusting manner in the ranking stage. Our approaches, which evolved from using Strong-OR logical operator to bucketized retrieval at the retrieval stage and from greedy re-rankers to multi-objective optimization using determinantal point processes for the ranking stage, balances diversity and utility while enabling fast iterations and scalable expansion to diversification over multiple dimensions. Our experiments indicate that these approaches significantly improve diversity metrics, with a neutral to a positive impact on utility metrics and improved user satisfaction, both qualitatively and quantitatively, in production.