Aarti Singh
Professor Carnegie Mellon University
- Pittsburgh PA
Aarti Singh's research focuses on design of AI agents that complement human decision making.
Biography
Areas of Expertise
Media Appearances
Most local governments don't have AI guidelines. Allegheny County instituted a policy.
Pittsburgh Post-Gazette online
2025-08-24
Aarti Singh, a CMU professor who has researched AI for about 20 years, currently is director of the university’s AI Institute for Societal Decision Making. She said it’s good that the county has produced a policy.
Personalities of Pittsburgh: Aarti Singh, CMU's AI Institute for Societal Decision Making
Pittsburgh Business Times online
2023-07-17
Aarti Singh is leading innovations in AI at Carnegie Mellon University.
Our Region's Business - NSF AI Institute for Societal Decision Making CMU
WPXI tv
2023-07-16
[no quote available]
'WTAE Listens': AI; the technology of tomorrow
WTAE tv
2023-06-25
Artificial intelligence is a topic we're hearing about all the time lately. But what is AI, really, and how is it changing the world around us? On WTAE Listens, we're decoding this new technology.
RealLIST Connectors 2023: Meet 20 (more) Pittsburghers leading the tech community into the future
Technical.ly online
2023-06-20
Aarti Singh is a professor in Carnegie Mellon University’s Machine Learning Department and, as of May, the head of the new AI Institute for Societal Decision Making. As the institute’s co-director, she’ll be responsible for overseeing its mission to bring social sciences and tech research together to figure out how humans and the technology can better interact: “We need to have social scientists and AI researchers collaborate to come up with solutions that will leverage AI capability while ensuring social acceptance,” Singh previously told Technical.ly Over the years, Singh has received awards such as the Faculty Early Career Development Award from the National Science Foundation.
Carnegie Mellon University receives $20M funding to establish AI institute
Coingeek online
2023-05-26
“We need to develop AI technology that works for the people,” said Aarti Singh, a Machine Learning professor tapped to be the institute’s first director. “It’s actually built on data that is vetted, algorithms that are vetted, with feedback from all the stakeholders and participatory design.”
Carnegie Mellon artificial intelligence institute to receive $20M in federal funding
TribLIVE.com online
2023-05-11
Aarti Singh, a professor in the Machine Learning Department of CMU’s School of Computer Science, will serve as the institute’s director. She said it’s important to “have social scientists and AI researchers collaborate to come up with solutions that will leverage AI capability while ensuring social acceptance.”
CMU won $20M to create a new institute focused on ‘human-centric’ AI solutions
Technical.ly online
2023-05-09
The new institute’s co-director, Aarti Singh, told Technical.ly the institute is going to bring social sciences and tech research together to figure out how humans and the technology can better interact. “For [the] maximal impact of these technologies, we need to have social scientists and AI researchers collaborate to come up with solutions that will leverage AI capability while ensuring social acceptance,” Singh said.
Carnegie Mellon leads NSF AI Institute for Societal Decision Making
EurekAlert! online
2023-05-04
"The best applications of artificial intelligence in societal domains will come when we not only advance AI for decision-making, but also better understand human decision-making, and when we can bring the two together," said Aarti Singh, a professor in the Machine Learning Department of CMU's School of Computer Science, who will serve as the institute's director.
Media
Education
Princeton University
Postdoc
Applied and Computational Math
University of Wisconsin, Madison
Ph.D.
Electrical Engineering
University of Wisconsin, Madison
M.S.
Electrical Engineering
University of Delhi, India
B.E.
Electronics & Communication Engineering
Links
Articles
Online Social Welfare Function-based Resource Allocation
International Conference on Machine Learning, ICML'262026
In many real-world settings, a centralized decision-maker must repeatedly allocate finite resources to a population over multiple time steps. Individuals who receive a resource derive some stochastic utility; to characterize the population-level effects of an allocation, the expected individual utilities are then aggregated using a social welfare function (SWF). We formalize this setting and present a general confidence sequence framework for SWF-based online learning and inference, valid for any monotonic, concave, and Lipschitz-continuous SWF. Our key insight is that monotonicity alone suffices to lift confidence sequences from individual utilities to anytime-valid bounds on optimal welfare.
Expanding the Capabilities of Reinforcement Learning via Text Feedback
International Conference on Machine Learning, ICML'262026
The success of RL for LLM post-training stems from an unreasonably uninformative source: a single bit of information per rollout as binary reward or preference label. At the other extreme, distillation offers dense supervision but requires demonstrations, which are costly and difficult to scale. We study text feedback as an intermediate signal: richer than scalar rewards, yet cheaper than complete demonstrations. Textual feedback is a natural mode of human interaction and is already abundant in many real-world settings, where users, annotators, and automated judges routinely critique LLM outputs. Towards leveraging text feedback at scale, we formalize a multi-turn RL setup, RL from Text Feedback (RLTF), where text feedback is available during training but not at inference.
Align AI to Dynamic Human-AI Workflows
Pre-print2026
Current alignment approaches typically focus on emulating human behavior using static representations of human preferences, failing to capture the dynamic, context-dependent nature of real-world human-AI interactions. In this paper, we argue for a shift from static and emulative to interactive and complementary alignment, where preferences emerge through interaction and alignment is defined not by satisfying preferences alone. We first formalize this gap by contrasting existing alignment with a trajectory-level view in which human and model behavior co-evolve over time. Because these interaction dynamics have not been adequately captured within existing ML formulations, we ground this perspective in insights from an interdisciplinary workshop.
Developing and evaluating a chatbot to support maternal health care
Pre-print2026
The ability to provide trustworthy maternal health information using phone-based chatbots can have a significant impact, particularly in low-resource settings where users have low health literacy and limited access to care. However, deploying such systems is technically challenging: user queries are short, underspecified, and code-mixed across languages, answers require regional context-specific grounding, and partial or missing symptom context makes safe routing decisions difficult.
We present a chatbot for maternal health in India developed through a partnership between academic researchers, a health tech company, a public health nonprofit, and a hospital.
AI+HW 2035: Shaping the Next Decade
Preprint2026
Artificial intelligence (AI) and hardware (HW) are advancing at unprecedented rates, yet their trajectories have become inseparably intertwined. The global research community lacks a cohesive, long-term vision to strategically coordinate the development of AI and HW. This fragmentation constrains progress toward holistic, sustainable, and adaptive AI systems capable of learning, reasoning, and operating efficiently across cloud, edge, and physical environments. The future of AI depends not only on scaling intelligence, but on scaling efficiency, achieving exponential gains in intelligence per joule, rather than unbounded compute consumption.
Frontiers of Statistics in Science and Engineering: 2035 and Beyond
The National Academies Press, National Academies of Sciences, Engineering, and Medicine2026
From advances in artificial intelligence and blockchain to precision agriculture, statistical innovation drives progress and bolsters U.S. competitiveness. Data underpins science in every field and sector, and statistics transforms that data into insight. This report explores key developments in statistics, highlights critical future directions, and presents investment priorities to secure U.S. economic and technological advantage over the next decade.
Toward a Science of Human-AI Teaming for Decision-making: A complementarity framework.
Proceedings of National Academy of Sciences (PNAS) Nexus2026
As artificial intelligence (AI) becomes embedded in critical decisions involving health, safety, finance, and governance, the key challenge is no longer whether humans and AI will collaborate, but rather how to structure this collaboration to achieve true complementarity. Human–AI complementarity refers to the conditions under which human–AI teams outperform either humans alone or AI systems alone. This paper advances the science of human–AI teaming for decision making by integrating insights from cognitive science, AI, human factors, organizational behavior, and ethics. We propose a framework grounded in collective intelligence and anchored in the foundational cognitive processes–reasoning, memory, and attention–to understand and engineer effective human–AI teams.
Investigating the Role of AI in Emergency Management: Use Cases, Challenges, and Opportunities
ACM Conference on Fairness, Accountability, and Transparency, FAccT'262026
Though natural disasters and other hazards remain a persistent threat, Emergency Managers (EMs) face substantial barriers to effectively completing their life-saving job functions. Pervasive staffing challenges and diminishing funding sources complicate a profession tasked with timely, critical decision-making under uncertainty. While technology presents a significant opportunity to complement EMs' workflows, commercial products are extremely costly and typically serve narrow use cases. To understand how EMs use and envision using technology, and in particular AI, we conducted in-depth interviews with 32 professionals spanning different geographic regions within the US.
Projection Optimization: A General Framework for Multi-Objective and Multi-Group RLHF
International Conference on Machine Learning, ICML'252025
Reinforcement Learning with Human Feedback (RLHF) is a widely used fine-tuning approach that aligns machine learning model, particularly Language Model (LM) with human preferences. There are typically multiple objectives driving the preference, hence humans find it easier to express per-objective comparisons rather than a global preference between two choices. Multi-Objective RLHF (MORLHF) aims to use per-objective preference feedback and achieve Pareto optimality among these objectives by aggregating them into a single unified objective for optimization. However, nearly all prior works rely on linear aggregation, which rules out policies that favor specific objectives such as the worst one. The only existing approach using non-linear aggregation is computationally expensive due to its reward-based nature and the need for retraining whenever the aggregation parameters change.
Learning Social Welfare Functions
Neural Information Processing Systems, NeurIPS'242024
Is it possible to understand or imitate a policy maker's rationale by looking at past decisions they made? We formalize this question as the problem of learning social welfare functions belonging to the well-studied family of power mean functions. We focus on two learning tasks; in the first, the input is vectors of utilities of an action (decision or policy) for individuals in a group and their associated social welfare as judged by a policy maker, whereas in the second, the input is pairwise comparisons between the welfares associated with a given pair of utility vectors. We show that power mean functions are learnable with polynomial sample complexity in both cases, even if the comparisons are social welfare information is noisy. Finally, we design practical algorithms for these tasks and evaluate their performance.
Two-Sample Testing on Ranked Preference Data and the Role of Modeling Assumptions
Journal of Machine Learning Research2022
A number of applications require two-sample testing on ranked preference data. For instance, in crowdsourcing, there is a long-standing question of whether pairwise-comparison data provided by people is distributed identically to ratings-converted-to-comparisons. Other applications include sports data analysis and peer grading. In this paper, we design two-sample tests for pairwise-comparison data and ranking data. For our two-sample test for pairwise-comparison data, we establish an upper bound on the sample complexity required to correctly test whether the distributions of the two sets of samples are identical.
PeerReview4All: fair and accurate reviewer assignment in peer review
Journal of Machine Learning Research2021
We consider the problem of automated assignment of papers to reviewers in conference peer review, with a focus on fairness and statistical accuracy. Our fairness objective is to maximize the review quality of the most disadvantaged paper, in contrast to the commonly used objective of maximizing the total quality over all papers. We design an assignment algorithm based on an incremental max-ow procedure that we prove is near-optimally fair. Our statistical accuracy objective is to ensure correct recovery of the papers that should be accepted. We provide a sharp minimax analysis of the accuracy of the peer-review process for a popular objective-score model as well as for a novel subjective-score model that we propose in the paper.
Best Arm Identification under Additive Transfer Bandits
55th Asilomar Conference on Signals, Systems, and Computers2021
We consider a variant of the best arm identification (BAI) problem in multi-armed bandits (MAB) in which there are two sets of arms (source and target), and the objective is to determine the best target arm while only pulling source arms. In this paper, we study the setting when, despite the means being unknown, there is a known additive relationship between the source and target MAB instances. We show how our framework covers a range of previously studied pure exploration problems and additionally captures new problems. We propose and theoretically analyze an LUCB-style algorithm to identify an e-optimal target arm with high probability.
Prior and Prejudice: The Novice Reviewers' Bias against Resubmissions in Conference Peer Review
Proceedings of the ACM on Human-Computer Interaction2021
Modern machine learning and computer science conferences are experiencing a surge in the number of submissions that challenges the quality of peer review as the number of competent reviewers is growing at a much slower rate. To curb this trend and reduce the burden on reviewers, several conferences have started encouraging or even requiring authors to declare the previous submission history of their papers. Such initiatives have been met with skepticism among authors, who raise the concern about a potential bias in reviewers' recommendations induced by this information.
Local signal adaptivity: Provable feature learning in neural networks beyond kernels
Advances in Neural Information Processing Systems2021
Neural networks have been shown to outperform kernel methods in practice (including neural tangent kernels). Most theoretical explanations of this performance gap focus on learning a complex hypothesis class; in some cases, it is unclear whether this hypothesis class captures realistic data. In this work, we propose a related, but alternative, explanation for this performance gap in the image classification setting, based on finding a sparse signal in the presence of noise. Specifically, we prove that, for a simple data distribution with sparse signal amidst high-variance noise, a simple convolutional neural network trained using stochastic gradient descent learns to threshold out the noise and find the signal.
