Hussein Mozannar

Hussein Mozannar, Gagan Bansal, Adam Fourney et al. · microsoft-research

Code-recommendation systems, such as Copilot and CodeWhisperer, have the potential to improve programmer productivity by suggesting and auto-completing code. However, to fully realize their potential, we must understand how programmers interact with these systems and identify ways to improve that interaction. To seek insights about human-AI collaboration with code recommendations systems, we studied GitHub Copilot, a code-recommendation system used by millions of programmers daily. We developed CUPS, a taxonomy of common programmer activities when interacting with Copilot. Our study of 21 programmers, who completed coding tasks and retrospectively labeled their sessions with CUPS, showed that CUPS can help us understand how programmers interact with code-recommendation systems, revealing inefficiencies and time costs. Our insights reveal how programmers interact with Copilot and motivate new interface designs and metrics.

Hussein Mozannar, Gagan Bansal, Adam Fourney et al. · microsoft-research

AI powered code-recommendation systems, such as Copilot and CodeWhisperer, provide code suggestions inside a programmer's environment (e.g., an IDE) with the aim of improving productivity. We pursue mechanisms for leveraging signals about programmers' acceptance and rejection of code suggestions to guide recommendations. We harness data drawn from interactions with GitHub Copilot, a system used by millions of programmers, to develop interventions that can save time for programmers. We introduce a utility-theoretic framework to drive decisions about suggestions to display versus withhold. The approach, conditional suggestion display from human feedback (CDHF), relies on a cascade of models that provide the likelihood that recommended code will be accepted. These likelihoods are used to selectively hide suggestions, reducing both latency and programmer verification time. Using data from 535 programmers, we perform a retrospective evaluation of CDHF and show that we can avoid displaying a significant fraction of suggestions that would have been rejected. We further demonstrate the importance of incorporating the programmer's latent unobserved state in decisions about when to display suggestions through an ablation study. Finally, we showcase how using suggestion acceptance as a reward signal for guiding the display of suggestions can lead to suggestions of reduced quality, indicating an unexpected pitfall.

Hussein Mozannar, Hunter Lang, Dennis Wei et al. · microsoft-research, mit

Automated AI classifiers should be able to defer the prediction to a human decision maker to ensure more accurate predictions. In this work, we jointly train a classifier with a rejector, which decides on each data point whether the classifier or the human should predict. We show that prior approaches can fail to find a human-AI system with low misclassification error even when there exists a linear classifier and rejector that have zero error (the realizable setting). We prove that obtaining a linear pair with low error is NP-hard even when the problem is realizable. To complement this negative result, we give a mixed-integer-linear-programming (MILP) formulation that can optimally solve the problem in the linear setting. However, the MILP only scales to moderately-sized problems. Therefore, we provide a novel surrogate loss function that is realizable-consistent and performs well empirically. We test our approaches on a comprehensive set of datasets and compare to a wide range of baselines.

Mohammad-Amin Charusaie, Hussein Mozannar, David Sontag et al. · microsoft-research

One of the goals of learning algorithms is to complement and reduce the burden on human decision makers. The expert deferral setting wherein an algorithm can either predict on its own or defer the decision to a downstream expert helps accomplish this goal. A fundamental aspect of this setting is the need to learn complementary predictors that improve on the human's weaknesses rather than learning predictors optimized for average error. In this work, we provide the first theoretical analysis of the benefit of learning complementary predictors in expert deferral. To enable efficiently learning such predictors, we consider a family of consistent surrogate loss functions for expert deferral and analyze their theoretical properties. Finally, we design active learning schemes that require minimal amount of data of human expert predictions in order to learn accurate deferral systems.

Hussein Mozannar, Jimin J Lee, Dennis Wei et al. · microsoft-research

People are relying on AI agents to assist them with various tasks. The human must know when to rely on the agent, collaborate with the agent, or ignore its suggestions. In this work, we propose to learn rules, grounded in data regions and described in natural language, that illustrate how the human should collaborate with the AI. Our novel region discovery algorithm finds local regions in the data as neighborhoods in an embedding space where prior human behavior should be corrected. Each region is then described using a large language model in an iterative and contrastive procedure. We then teach these rules to the human via an onboarding stage. Through user studies on object detection and question-answering tasks, we show that our method can lead to more accurate human-AI teams. We also evaluate our region discovery and description algorithms separately.

Madeleine Grunde-McLaughlin, Hussein Mozannar, Maya Murad et al. · microsoft-research

To enable human oversight, agentic AI systems often provide a trace of reasoning and action steps. Designing traces to have an informative, but not overwhelming, level of detail remains a critical challenge. In three user studies on a Computer User Agent, we investigate the utility of basic action traces for verification, explore three alternatives via design probes, and test a novel interface's impact on error finding in question-answering tasks. As expected, we find that current practices are cumbersome, limiting their efficacy. Conversely, our proposed design reduced the time participants spent finding errors. However, although participants reported higher levels of confidence in their decisions, their final accuracy was not meaningfully improved. To this end, our study surfaces challenges for human verification of agentic systems, including managing built-in assumptions, users' subjective and changing correctness criteria, and the shortcomings, yet importance, of communicating the agent's process.

Yuzhou Cao, Hussein Mozannar, Lei Feng et al.

Enabling machine learning classifiers to defer their decision to a downstream expert when the expert is more accurate will ensure improved safety and performance. This objective can be achieved with the learning-to-defer framework which aims to jointly learn how to classify and how to defer to the expert. In recent studies, it has been theoretically shown that popular estimators for learning to defer parameterized with softmax provide unbounded estimates for the likelihood of deferring which makes them uncalibrated. However, it remains unknown whether this is due to the widely used softmax parameterization and if we can find a softmax-based estimator that is both statistically consistent and possesses a valid probability estimator. In this work, we first show that the cause of the miscalibrated and unbounded estimator in prior literature is due to the symmetric nature of the surrogate losses used and not due to softmax. We then propose a novel statistically consistent asymmetric softmax-based surrogate loss that can produce valid estimates without the issue of unboundedness. We further analyze the non-asymptotic properties of our method and empirically validate its performance and calibration on benchmark datasets.

Dhruv Agarwal, Emily Sheng, Chad Atalla et al.

Evaluating generative AI (GenAI) systems is challenging because many targets of evaluation are broad, contested concepts, such as "reasoning," "fairness," or "creativity." When these concepts are left underspecified, it becomes unclear what should be measured or how evaluation results should be interpreted. This problem reflects a missing step: systematization, that is, moving from a broad background concept to an explicit, structured account of the concept in measurable terms. To help address the fact that systematization is cognitively demanding and resource-intensive, we investigate whether AI assistance can support this process. To enable AI-assisted systematization and assess its quality, we introduce a structured representation of a systematized concept, a concept spec, and a validation worksheet. We then develop two AI-assisted systematizers: a direct, zero-shot approach and a multi-agent approach that more closely mirrors manual systematization approaches from existing literature. We use these systematizers to produce concept specs for two concepts -- hate-based rhetoric and digital empathy -- and evaluate resulting concept specs on content validity and information recoverability.

Adam Fourney, Gagan Bansal, Hussein Mozannar et al. · microsoft-research

Modern AI agents, driven by advances in large foundation models, promise to enhance our productivity and transform our lives by augmenting our knowledge and capabilities. To achieve this vision, AI agents must effectively plan, perform multi-step reasoning and actions, respond to novel observations, and recover from errors, to successfully complete complex tasks across a wide range of scenarios. In this work, we introduce Magentic-One, a high-performing open-source agentic system for solving such tasks. Magentic-One uses a multi-agent architecture where a lead agent, the Orchestrator, plans, tracks progress, and re-plans to recover from errors. Throughout task execution, the Orchestrator directs other specialized agents to perform tasks as needed, such as operating a web browser, navigating local files, or writing and executing Python code. We show that Magentic-One achieves statistically competitive performance to the state-of-the-art on three diverse and challenging agentic benchmarks: GAIA, AssistantBench, and WebArena. Magentic-One achieves these results without modification to core agent capabilities or to how they collaborate, demonstrating progress towards generalist agentic systems. Moreover, Magentic-One's modular design allows agents to be added or removed from the team without additional prompt tuning or training, easing development and making it extensible to future scenarios. We provide an open-source implementation of Magentic-One, and we include AutoGenBench, a standalone tool for agentic evaluation. AutoGenBench provides built-in controls for repetition and isolation to run agentic benchmarks in a rigorous and contained manner -- which is important when agents' actions have side-effects. Magentic-One, AutoGenBench and detailed empirical performance evaluations of Magentic-One, including ablations and error analysis are available at https://aka.ms/magentic-one

Hussein Mozannar, Valerie Chen, Mohammed Alsobay et al. · cmu, microsoft-research

Evaluation of large language models for code has primarily relied on static benchmarks, including HumanEval (Chen et al., 2021), or more recently using human preferences of LLM responses. As LLMs are increasingly used as programmer assistants, we study whether gains on existing benchmarks or more preferred LLM responses translate to programmer productivity when coding with LLMs, including time spent coding. We introduce RealHumanEval, a web interface to measure the ability of LLMs to assist programmers, through either autocomplete or chat support. We conducted a user study (N=243) using RealHumanEval in which users interacted with seven LLMs of varying base model performance. Despite static benchmarks not incorporating humans-in-the-loop, we find that improvements in benchmark performance lead to increased programmer productivity; however gaps in benchmark versus human performance are not proportional -- a trend that holds across both forms of LLM support. In contrast, we find that programmer preferences do not correlate with their actual performance, motivating the need for better proxy signals. We open-source RealHumanEval to enable human-centric evaluation of new models and the study data to facilitate efforts to improve code models.

Hussein Mozannar, Gagan Bansal, Cheng Tan et al. · microsoft-research

AI agents powered by large language models are increasingly capable of autonomously completing complex, multi-step tasks using external tools. Yet, they still fall short of human-level performance in most domains including computer use, software development, and research. Their growing autonomy and ability to interact with the outside world, also introduces safety and security risks including potentially misaligned actions and adversarial manipulation. We argue that human-in-the-loop agentic systems offer a promising path forward, combining human oversight and control with AI efficiency to unlock productivity from imperfect systems. We introduce Magentic-UI, an open-source web interface for developing and studying human-agent interaction. Built on a flexible multi-agent architecture, Magentic-UI supports web browsing, code execution, and file manipulation, and can be extended with diverse tools via Model Context Protocol (MCP). Moreover, Magentic-UI presents six interaction mechanisms for enabling effective, low-cost human involvement: co-planning, co-tasking, multi-tasking, action guards, and long-term memory. We evaluate Magentic-UI across four dimensions: autonomous task completion on agentic benchmarks, simulated user testing of its interaction capabilities, qualitative studies with real users, and targeted safety assessments. Our findings highlight Magentic-UI's potential to advance safe and efficient human-agent collaboration.

Omar Shaikh, Hussein Mozannar, Gagan Bansal et al. · microsoft-research

Language models excel at following instructions but often struggle with the collaborative aspects of conversation that humans naturally employ. This limitation in grounding -- the process by which conversation participants establish mutual understanding -- can lead to outcomes ranging from frustrated users to serious consequences in high-stakes scenarios. To systematically study grounding challenges in human-LLM interactions, we analyze logs from three human-assistant datasets: WildChat, MultiWOZ, and Bing Chat. We develop a taxonomy of grounding acts and build models to annotate and forecast grounding behavior. Our findings reveal significant differences in human-human and human-LLM grounding: LLMs were three times less likely to initiate clarification and sixteen times less likely to provide follow-up requests than humans. Additionally, we find that early grounding failures predict later interaction breakdowns. Building on these insights, we introduce Rifts, a benchmark derived from publicly available LLM interaction data containing situations where LLMs fail to initiate grounding. We note that current frontier models perform poorly on Rifts, highlighting the need to reconsider how we train and prompt LLMs for human interaction. To this end, we develop a preliminary intervention aimed at mitigating grounding failures.

Gagan Bansal, Jennifer Wortman Vaughan, Saleema Amershi et al. · microsoft-research

Remarkable advancements in modern generative foundation models have enabled the development of sophisticated and highly capable autonomous agents that can observe their environment, invoke tools, and communicate with other agents to solve problems. Although such agents can communicate with users through natural language, their complexity and wide-ranging failure modes present novel challenges for human-AI interaction. Building on prior research and informed by a communication grounding perspective, we contribute to the study of \emph{human-agent communication} by identifying and analyzing twelve key communication challenges that these systems pose. These include challenges in conveying information from the agent to the user, challenges in enabling the user to convey information to the agent, and overarching challenges that need to be considered across all human-agent communication. We illustrate each challenge through concrete examples and identify open directions of research. Our findings provide insights into critical gaps in human-agent communication research and serve as an urgent call for new design patterns, principles, and guidelines to support transparency and control in these systems.

Niklas Mannhardt, Elizabeth Bondi-Kelly, Barbara Lam et al. · microsoft-research

Large language models (LLMs) have immense potential to make information more accessible, particularly in medicine, where complex medical jargon can hinder patient comprehension of clinical notes. We developed a patient-facing tool using LLMs to make clinical notes more readable by simplifying, extracting information from, and adding context to the notes. We piloted the tool with clinical notes donated by patients with a history of breast cancer and synthetic notes from a clinician. Participants (N=200, healthy, female-identifying patients) were randomly assigned three clinical notes in our tool with varying levels of augmentations and answered quantitative and qualitative questions evaluating their understanding of follow-up actions. Augmentations significantly increased their quantitative understanding scores. In-depth interviews were conducted with participants (N=7, patients with a history of breast cancer), revealing both positive sentiments about the augmentations and concerns about AI. We also performed a qualitative clinician-driven analysis of the model's error modes.

Hyewon Jeong, Sarah Jabbour, Yuzhe Yang et al. · uw

The third ML4H symposium was held in person on December 10, 2023, in New Orleans, Louisiana, USA. The symposium included research roundtable sessions to foster discussions between participants and senior researchers on timely and relevant topics for the \ac{ML4H} community. Encouraged by the successful virtual roundtables in the previous year, we organized eleven in-person roundtables and four virtual roundtables at ML4H 2022. The organization of the research roundtables at the conference involved 17 Senior Chairs and 19 Junior Chairs across 11 tables. Each roundtable session included invited senior chairs (with substantial experience in the field), junior chairs (responsible for facilitating the discussion), and attendees from diverse backgrounds with interest in the session's topic. Herein we detail the organization process and compile takeaways from these roundtable discussions, including recent advances, applications, and open challenges for each topic. We conclude with a summary and lessons learned across all roundtables. This document serves as a comprehensive review paper, summarizing the recent advancements in machine learning for healthcare as contributed by foremost researchers in the field.

Enhao Zhang, Erkang Zhu, Gagan Bansal et al. · microsoft-research

Large language model (LLM)-based multi-agent systems have demonstrated remarkable promise for tackling complex tasks by breaking them down into subtasks that are iteratively planned, executed, observed, and refined. Despite their effectiveness, these systems often incur high latency because real-world problems frequently demand multiple iterative cycles of reasoning steps. To address this challenge, we propose M1-Parallel, a framework that concurrently runs multiple multi-agent teams in parallel to uncover distinct solution paths. By leveraging an event-driven communication model with asynchronous messaging, M1-Parallel efficiently capitalizes on the inherent diversity of valid plans to either reduce end-to-end latency or boost task completion rates. Our experiments on complex tasks show that M1-Parallel with early termination achieves up to $2.2\times$ speedup while preserving accuracy, and that M1-Parallel with aggregation yields higher task completion rates. We further investigate strategies aimed at encouraging diverse execution plans but observe no additional performance gains over repeated sampling. Overall, these findings underscore the potential of parallel plan execution for optimizing multi-agent systems for real-world, high-complexity reasoning tasks.

Ahmed Awadallah, Yash Lara, Raghav Magazine et al.

Progress in computer use agents (CUAs) has been constrained by the absence of large and high-quality datasets that capture how humans interact with a computer. While LLMs have thrived on abundant textual data, no comparable corpus exists for CUA trajectories. To address these gaps, we introduce FaraGen, a novel synthetic data generation system for multi-step web tasks. FaraGen can propose diverse tasks from frequently used websites, generate multiple solution attempts, and filter successful trajectories using multiple verifiers. It achieves high throughput, yield, and diversity for multi-step web tasks, producing verified trajectories at approximately $1 each. We use this data to train Fara-7B, a native CUA model that perceives the computer using only screenshots, executes actions via predicted coordinates, and is small enough to run on-device. We find that Fara-7B outperforms other CUA models of comparable size on benchmarks like WebVoyager, Online-Mind2Web, and WebTailBench -- our novel benchmark that better captures under-represented web tasks in pre-existing benchmarks. Furthermore, Fara-7B is competitive with much larger frontier models, illustrating key benefits of scalable data generation systems in advancing small efficient agentic models. We are making Fara-7B open-weight on Microsoft Foundry and HuggingFace, and we are releasing WebTailBench.

Gagan Bansal, Wenyue Hua, Zezhou Huang et al. · microsoft-research

As LLM agents advance, they are increasingly mediating economic decisions, ranging from product discovery to transactions, on behalf of users. Such applications promise benefits but also raise many questions about agent accountability and value for users. Addressing these questions requires understanding how agents behave in realistic market conditions. However, previous research has largely evaluated agents in constrained settings, such as single-task marketplaces (e.g., negotiation) or structured two-agent interactions. Real-world markets are fundamentally different: they require agents to handle diverse economic activities and coordinate within large, dynamic ecosystems where multiple agents with opaque behaviors may engage in open-ended dialogues. To bridge this gap, we investigate two-sided agentic marketplaces where Assistant agents represent consumers and Service agents represent competing businesses. To study these interactions safely, we develop Magentic-Marketplace -- a simulated environment where Assistants and Services can operate. This environment enables us to study key market dynamics: the utility agents achieve, behavioral biases, vulnerability to manipulation, and how search mechanisms shape market outcomes. Our experiments show that frontier models can approach optimal welfare -- but only under ideal search conditions. Performance degrades sharply with scale, and all models exhibit severe first-proposal bias, creating 10-30x advantages for response speed over quality. These findings reveal how behaviors emerge across market conditions, informing the design of fair and efficient agentic marketplaces.

Hussein Mozannar, Yuria Utsumi, Irene Y. Chen et al.

A high-risk pregnancy is a pregnancy complicated by factors that can adversely affect the outcomes of the mother or the infant. Health insurers use algorithms to identify members who would benefit from additional clinical support. This work presents the implementation of a real-world ML-based system to assist care managers in identifying pregnant patients at risk of complications. In this retrospective evaluation study, we developed a novel hybrid-ML classifier to predict whether patients are pregnant and trained a standard classifier using claims data from a health insurance company in the US to predict whether a patient will develop pregnancy complications. These models were developed in cooperation with the care management team and integrated into a user interface with explanations for the nurses. The proposed models outperformed commonly used claim codes for the identification of pregnant patients at the expense of a manageable false positive rate. Our risk complication classifier shows that we can accurately triage patients by risk of complication. Our approach and evaluation are guided by human-centric design. In user studies with the nurses, they preferred the proposed models over existing approaches.

Hussein Mozannar, Arvind Satyanarayan, David Sontag

Expert decision makers are starting to rely on data-driven automated agents to assist them with various tasks. For this collaboration to perform properly, the human decision maker must have a mental model of when and when not to rely on the agent. In this work, we aim to ensure that human decision makers learn a valid mental model of the agent's strengths and weaknesses. To accomplish this goal, we propose an exemplar-based teaching strategy where humans solve the task with the help of the agent and try to formulate a set of guidelines of when and when not to defer. We present a novel parameterization of the human's mental model of the AI that applies a nearest neighbor rule in local regions surrounding the teaching examples. Using this model, we derive a near-optimal strategy for selecting a representative teaching set. We validate the benefits of our teaching strategy on a multi-hop question answering task using crowd workers and find that when workers draw the right lessons from the teaching stage, their task performance improves, we furthermore validate our method on a set of synthetic experiments.

Hussein Mozannar, David Sontag

Learning algorithms are often used in conjunction with expert decision makers in practical scenarios, however this fact is largely ignored when designing these algorithms. In this paper we explore how to learn predictors that can either predict or choose to defer the decision to a downstream expert. Given only samples of the expert's decisions, we give a procedure based on learning a classifier and a rejector and analyze it theoretically. Our approach is based on a novel reduction to cost sensitive learning where we give a consistent surrogate loss for cost sensitive learning that generalizes the cross entropy loss. We show the effectiveness of our approach on a variety of experimental tasks.

Hussein Mozannar, Mesrob I. Ohannessian, Nathan Srebro

Sensitive attributes such as race are rarely available to learners in real world settings as their collection is often restricted by laws and regulations. We give a scheme that allows individuals to release their sensitive information privately while still allowing any downstream entity to learn non-discriminatory predictors. We show how to adapt non-discriminatory learners to work with privatized protected attributes giving theoretical guarantees on performance. Finally, we highlight how the methodology could apply to learning fair predictors in settings where protected attributes are only available for a subset of the data.

Hussein Mozannar, Karl El Hajal, Elie Maamary et al.

This paper tackles the problem of open domain factual Arabic question answering (QA) using Wikipedia as our knowledge source. This constrains the answer of any question to be a span of text in Wikipedia. Open domain QA for Arabic entails three challenges: annotated QA datasets in Arabic, large scale efficient information retrieval and machine reading comprehension. To deal with the lack of Arabic QA datasets we present the Arabic Reading Comprehension Dataset (ARCD) composed of 1,395 questions posed by crowdworkers on Wikipedia articles, and a machine translation of the Stanford Question Answering Dataset (Arabic-SQuAD). Our system for open domain question answering in Arabic (SOQAL) is based on two components: (1) a document retriever using a hierarchical TF-IDF approach and (2) a neural reading comprehension model using the pre-trained bi-directional transformer BERT. Our experiments on ARCD indicate the effectiveness of our approach with our BERT-based reader achieving a 61.3 F1 score, and our open domain system SOQAL achieving a 27.6 F1 score.

Hussein Mozannar, Mesrob I. Ohannessian, Nathan Srebro

The study of fairness in intelligent decision systems has mostly ignored long-term influence on the underlying population. Yet fairness considerations (e.g. affirmative action) have often the implicit goal of achieving balance among groups within the population. The most basic notion of balance is eventual equality between the qualifications of the groups. How can we incorporate influence dynamics in decision making? How well do dynamics-oblivious fairness policies fare in terms of reaching equality? In this paper, we propose a simple yet revealing model that encompasses (1) a selection process where an institution chooses from multiple groups according to their qualifications so as to maximize an institutional utility and (2) dynamics that govern the evolution of the groups' qualifications according to the imposed policies. We focus on demographic parity as the formalism of affirmative action. We then give conditions under which an unconstrained policy reaches equality on its own. In this case, surprisingly, imposing demographic parity may break equality. When it doesn't, one would expect the additional constraint to reduce utility, however, we show that utility may in fact increase. In more realistic scenarios, unconstrained policies do not lead to equality. In such cases, we show that although imposing demographic parity may remedy it, there is a danger that groups settle at a worse set of qualifications. As a silver lining, we also identify when the constraint not only leads to equality, but also improves all groups. This gives quantifiable insight into both sides of the mismatch hypothesis. These cases and trade-offs are instrumental in determining when and how imposing demographic parity can be beneficial in selection processes, both for the institution and for society on the long run.