Alex Pentland

Zhangchen Xu, Junda Chen, Yue Huang et al.

Scientific and engineering progress is fundamentally a long-horizon iterative process: proposing changes, running experiments, measuring outcomes, and continuously refining artifacts. Yet existing benchmarks for frontier models primarily evaluate either single-turn responses or short-horizon agent trajectories, failing to capture the challenges of sustained iterative improvement over extended time horizons. To address this gap, we introduce AutoLab, a new benchmark for ultra long-horizon closed-loop optimization. AutoLab consists of 36 realistic, expert-curated tasks spanning four diverse domains: system optimization, puzzle & challenge, model development, and CUDA kernel optimization. Each task begins with a correct but deliberately suboptimal baseline and challenges agents to improve it within a strict wall-clock budget. Evaluating 17 state-of-the-art models reveals the dominant predictor of success is not the quality of an agent's initial attempt, but its persistence in repeatedly benchmarking, editing, and incorporating empirical feedback. While claude-opus-4.6 exhibits strong long-horizon optimization capabilities, most frontier models, including several proprietary ones, either terminate prematurely or exhaust their budgets with minimal progress. These results underscore the importance of time awareness and persistent iteration in autonomous agents. We open-source the full benchmark, evaluation harness, and task artifacts, to accelerate research toward truly capable long-horizon agents.

Abhimanyu Dubey, Alex Pentland · meta-ai

The cooperative bandit problem is a multi-agent decision problem involving a group of agents that interact simultaneously with a multi-armed bandit, while communicating over a network with delays. The central idea in this problem is to design algorithms that can efficiently leverage communication to obtain improvements over acting in isolation. In this paper, we investigate the stochastic bandit problem under two settings - (a) when the agents wish to make their communication private with respect to the action sequence, and (b) when the agents can be byzantine, i.e., they provide (stochastically) incorrect information. For both these problem settings, we provide upper-confidence bound algorithms that obtain optimal regret while being (a) differentially-private and (b) tolerant to byzantine agents. Our decentralized algorithms require no information about the network of connectivity between agents, making them scalable to large dynamic systems. We test our algorithms on a competitive benchmark of random graphs and demonstrate their superior performance with respect to existing robust algorithms. We hope that our work serves as an important step towards creating distributed decision-making systems that maintain privacy.

Ziv Epstein, Aaron Hertzmann, Laura Herman et al.

A new class of tools, colloquially called generative AI, can produce high-quality artistic media for visual arts, concept art, music, fiction, literature, video, and animation. The generative capabilities of these tools are likely to fundamentally alter the creative processes by which creators formulate ideas and put them into production. As creativity is reimagined, so too may be many sectors of society. Understanding the impact of generative AI - and making policy decisions around it - requires new interdisciplinary scientific inquiry into culture, economics, law, algorithms, and the interaction of technology and creativity. We argue that generative AI is not the harbinger of art's demise, but rather is a new medium with its own distinct affordances. In this vein, we consider the impacts of this new medium on creators across four themes: aesthetics and culture, legal questions of ownership and credit, the future of creative work, and impacts on the contemporary media ecosystem. Across these themes, we highlight key research questions and directions to inform policy and beneficial uses of the technology.

Yasmine Omri, Ziyu Gan, Zachary Broveak et al.

LLM agents are increasingly deployed on long-horizon tasks requiring sustained reasoning over extended interaction histories. Realizing this at scale requires agents to persistently store, retrieve, and update their own memory across sessions. A rich ecosystem of agent memory systems has emerged spanning flat retrieval, LLM-mediated extraction, consolidating fact stores, and agentic control flows. Yet, their system-level behavior remains uncharacterized. We present the first systems characterization of agent memory. First, we introduce a system-oriented taxonomy classifying agent memory systems along four axes. Second, we build a phase-aware profiling harness attributing cost to construction, retrieval, and generation. Third, we characterize ten representative systems across two benchmark suites, uncovering how design choices shift cost across the write and read paths. Finally, we derive 10 system recommendations covering construction scheduling, capability floors, amortization via query volume, freshness-latency tradeoffs, and fleet-scale management.

Christian Prothmann, Vijay Gadepally, Jeremy Kepner et al.

The DAF-MIT AI Accelerator is a collaboration between the United States Department of the Air Force (DAF) and the Massachusetts Institute of Technology (MIT). This program pioneers fundamental advances in artificial intelligence (AI) to expand the competitive advantage of the United States in the defense and civilian sectors. In recent years, AI Accelerator projects have developed and launched public challenge problems aimed at advancing AI research in priority areas. Hallmarks of AI Accelerator challenges include large, publicly available, and AI-ready datasets to stimulate open-source solutions and engage the wider academic and private sector AI ecosystem. This article supplements our previous publication, which introduced AI Accelerator challenges. We provide an update on how ongoing and new challenges have successfully contributed to AI research and applications of AI technologies.

Yanni Yang, Alex Pentland, Esteban Moro

Urbanization and its problems require an in-depth and comprehensive understanding of urban dynamics, especially the complex and diversified lifestyles in modern cities. Digitally acquired data can accurately capture complex human activity, but it lacks the interpretability of demographic data. In this paper, we study a privacy-enhanced dataset of the mobility visitation patterns of 1.2 million people to 1.1 million places in 11 metro areas in the U.S. to detect the latent mobility behaviors and lifestyles in the largest American cities. Despite the considerable complexity of mobility visitations, we found that lifestyles can be automatically decomposed into only 12 latent interpretable activity behaviors on how people combine shopping, eating, working, or using their free time. Rather than describing individuals with a single lifestyle, we find that city dwellers' behavior is a mixture of those behaviors. Those detected latent activity behaviors are equally present across cities and cannot be fully explained by main demographic features. Finally, we find those latent behaviors are associated with dynamics like experienced income segregation, transportation, or healthy behaviors in cities, even after controlling for demographic features. Our results signal the importance of complementing traditional census data with activity behaviors to understand urban dynamics.

Minglai Yang, Xinyan Velocity Yu, Pengyuan Li et al.

Document parsing and recognition are fundamental capabilities for vision-language models (VLMs) and document processing systems. However, existing Optical Character Recognition (OCR) and document parsing benchmarks are increasingly limited in coverage and difficulty: many focus on common document genres or uniformly sampled pages where modern parsers already perform strongly, while offering limited annotation for expert-domain structures such as chemical formula, music notation, complex tables, and cross-page layouts. We introduce Dr. DocBench, a difficulty-aware benchmark for expert-level document parsing. Built from a large-scale multilingual book corpus, Dr. DocBench spans 52 BISAC subject domains and selects challenging documents through parser-failure-based sampling, targeting cases where multiple state-of-the-art systems struggle. It contains 4,514 annotated pages from long documents averaging around 100 pages, with 65k high-quality page- and block-level annotations for layout, reading order, hierarchical relations, and domain-specific visual contents. Evaluations of pipeline-based parsers and general-purpose VLMs show that strong performance on existing benchmarks does not transfer to our expert-level document parsing. Our analysis reveals substantial failures across subjects, content types, and structural attributes, highlighting Dr. DocBench as a comprehensive testbed for diagnosing and advancing document intelligence.

Praneeth Vepakomma, Mohammad Mohammadi Amiri, Clément L. Canonne et al.

We introduce $π$-test, a privacy-preserving algorithm for testing statistical independence between data distributed across multiple parties. Our algorithm relies on privately estimating the distance correlation between datasets, a quantitative measure of independence introduced in Székely et al. [2007]. We establish both additive and multiplicative error bounds on the utility of our differentially private test, which we believe will find applications in a variety of distributed hypothesis testing settings involving sensitive data.

Laura Ferrarotti, Gian Maria Campedelli, Roberto Dessì et al.

In this article, we argue that understanding the collective behavior of agents based on large language models (LLMs) is an essential area of inquiry, with important implications in terms of risks and benefits, impacting us as a society at many levels. We claim that the distinctive nature of LLMs--namely, their initialization with extensive pre-trained knowledge and implicit social priors, together with their capability of adaptation through in-context learning--motivates the need for an interactionist paradigm consisting of alternative theoretical foundations, methodologies, and analytical tools, in order to systematically examine how prior knowledge and embedded values interact with social context to shape emergent phenomena in multi-agent generative AI systems. We propose and discuss four directions that we consider crucial for the development and deployment of LLM-based collectives, focusing on theory, methods, and trans-disciplinary dialogue.

Zexue He, Yu Wang, Churan Zhi et al.

Existing evaluations of agents with memory typically assess memorization and action in isolation. One class of benchmarks evaluates memorization by testing recall of past conversations or text but fails to capture how memory is used to guide future decisions. Another class focuses on agents acting in single-session tasks without the need for long-term memory. However, in realistic settings, memorization and action are tightly coupled: agents acquire memory while interacting with the environment, and subsequently rely on that memory to solve future tasks. To capture this setting, we introduce MemoryArena, a unified evaluation gym for benchmarking agent memory in multi-session Memory-Agent-Environment loops. The benchmark consists of human-crafted agentic tasks with explicitly interdependent subtasks, where agents must learn from earlier actions and feedback by distilling experiences into memory, and subsequently use that memory to guide later actions to solve the overall task. MemoryArena supports evaluation across web navigation, preference-constrained planning, progressive information search, and sequential formal reasoning, and reveals that agents with near-saturated performance on existing long-context memory benchmarks like LoCoMo perform poorly in our agentic setting, exposing a gap in current evaluations for agents with memory.

Shu Yang, Shenzhe Zhu, Hao Zhu et al.

Large language models (LLMs) and LLM-based agents are increasingly deployed as assistants in planning and decision making, yet most existing systems are implicitly optimized for a single-principal interaction paradigm, in which the model is designed to satisfy the objectives of one dominant user whose instructions are treated as the sole source of authority and utility. However, as they are integrated into team workflows and organizational tools, they are increasingly required to serve multiple users simultaneously, each with distinct roles, preferences, and authority levels, leading to multi-user, multi-principal settings with unavoidable conflicts, information asymmetry, and privacy constraints. In this work, we present the first systematic study of multi-user LLM agents. We begin by formalizing multi-user interaction with LLM agents as a multi-principal decision problem, where a single agent must account for multiple users with potentially conflicting interests and associated challenges. We then introduce a unified multi-user interaction protocol and design three targeted stress-testing scenarios to evaluate current LLMs' capabilities in instruction following, privacy preservation, and coordination. Our results reveal systematic gaps: frontier LLMs frequently fail to maintain stable prioritization under conflicting user objectives, exhibit increasing privacy violations over multi-turn interactions, and suffer from efficiency bottlenecks when coordination requires iterative information gathering.

Xavier Thomas, Dhruv Mahajan, Alex Pentland et al.

Domain generalization involves learning a classifier from a heterogeneous collection of training sources such that it generalizes to data drawn from similar unknown target domains, with applications in large-scale learning and personalized inference. In many settings, privacy concerns prohibit obtaining domain labels for the training data samples, and instead only have an aggregated collection of training points. Existing approaches that utilize domain labels to create domain-invariant feature representations are inapplicable in this setting, requiring alternative approaches to learn generalizable classifiers. In this paper, we propose a domain-adaptive approach to this problem, which operates in two steps: (a) we cluster training data within a carefully chosen feature space to create pseudo-domains, and (b) using these pseudo-domains we learn a domain-adaptive classifier that makes predictions using information about both the input and the pseudo-domain it belongs to. Our approach achieves state-of-the-art performance on a variety of domain generalization benchmarks without using domain labels whatsoever. Furthermore, we provide novel theoretical guarantees on domain generalization using cluster information. Our approach is amenable to ensemble-based methods and provides substantial gains even on large-scale benchmark datasets. The code can be found at: https://github.com/xavierohan/AdaClust_DomainBed

Longju Bai, Zhemin Huang, Xingyao Wang et al.

The wide adoption of AI agents in complex human workflows is driving rapid growth in LLM token consumption. When agents are deployed on tasks that require a significant amount of tokens, three questions naturally arise: (1) Where do AI agents spend the tokens? (2) Which models are more token-efficient? and (3) Can agents predict their token usage before task execution? In this paper, we present the first systematic study of token consumption patterns in agentic coding tasks. We analyze trajectories from eight frontier LLMs on SWE-bench Verified and evaluate models' ability to predict their own token costs before task execution. We find that: (1) agentic tasks are uniquely expensive, consuming 1000x more tokens than code reasoning and code chat, with input tokens rather than output tokens driving the overall cost; (2) token usage is highly variable and inherently stochastic: runs on the same task can differ by up to 30x in total tokens, and higher token usage does not translate into higher accuracy; instead, accuracy often peaks at intermediate cost and saturates at higher costs; (3) models vary substantially in token efficiency: on the same tasks, Kimi-K2 and Claude-Sonnet-4.5, on average, consume over 1.5 million more tokens than GPT-5; (4) task difficulty rated by human experts only weakly aligns with actual token costs, revealing a fundamental gap between human-perceived complexity and the computational effort agents actually expend; and (5) frontier models fail to accurately predict their own token usage (with weak-to-moderate correlations, up to 0.39) and systematically underestimate real token costs. Our study offers new insights into the economics of AI agents and can inspire future research in this direction.

Tobin South, Samuele Marro, Thomas Hardjono et al. · mit

The rapid deployment of autonomous AI agents creates urgent challenges around authorization, accountability, and access control in digital spaces. New standards are needed to know whom AI agents act on behalf of and guide their use appropriately, protecting online spaces while unlocking the value of task delegation to autonomous agents. We introduce a novel framework for authenticated, authorized, and auditable delegation of authority to AI agents, where human users can securely delegate and restrict the permissions and scope of agents while maintaining clear chains of accountability. This framework builds on existing identification and access management protocols, extending OAuth 2.0 and OpenID Connect with agent-specific credentials and metadata, maintaining compatibility with established authentication and web infrastructure. Further, we propose a framework for translating flexible, natural language permissions into auditable access control configurations, enabling robust scoping of AI agent capabilities across diverse interaction modalities. Taken together, this practical approach facilitates immediate deployment of AI agents while addressing key security and accountability concerns, working toward ensuring agentic AI systems perform only appropriate actions and providing a tool for digital service providers to enable AI agent interactions without risking harm from scalable interaction.

Jiachen Liu, Jiaxin Pei, Jintao Huang et al.

Scientific publication compresses a branching, iterative research process into a linear narrative, discarding the majority of what was discovered along the way. This compilation imposes two structural costs: a Storytelling Tax, where failed experiments, rejected hypotheses, and the branching exploration process are discarded to fit a linear narrative; and an Engineering Tax, where the gap between reviewer-sufficient prose and agent-sufficient specification leaves critical implementation details unwritten. Tolerable for human readers, these costs become critical when AI agents must understand, reproduce, and extend published work. We introduce the Agent-Native Research Artifact (Ara), a protocol that replaces the narrative paper with a machine-executable research package structured around four layers: scientific logic, executable code with full specifications, an exploration graph that preserves the failures compilation discards, and evidence grounding every claim in raw outputs. Three mechanisms support the ecosystem: a Live Research Manager that captures decisions and dead ends during ordinary development; an Ara Compiler that translates legacy PDFs and repos into Aras; and an Ara-native review system that automates objective checks so human reviewers can focus on significance, novelty, and taste. On PaperBench and RE-Bench, Ara raises question-answering accuracy from 72.4% to 93.7% and reproduction success from 57.4% to 64.4%. On RE-Bench's five open-ended extension tasks, preserved failure traces in Ara accelerate progress, but can also constrain a capable agent from stepping outside the prior-run box depending on the agent's capabilities.

Thomas P. Kehler, Scott E. Page, Alex Pentland et al.

We propose a general framework for human-AI collaboration that amplifies the distinct capabilities of both types of intelligence. We refer to this as Generative Collective Intelligence (GCI). GCI employs AI in dual roles: as interactive agents and as technology that accumulates, organizes, and leverages knowledge. In this second role, AI creates a cognitive bridge between human reasoning and AI models. The AI functions as a social and cultural technology that enables groups to solve complex problems through structured collaboration that transcends traditional communication barriers. We argue that GCI can overcome limitations of purely algorithmic approaches to problem-solving and decision-making. We describe the mathematical foundations of GCI, based on the law of comparative judgment and minimum regret principles, and briefly illustrate its applications across various domains, including climate adaptation, healthcare transformation, and civic participation. By combining human creativity with AI's computational capabilities, GCI offers a promising approach to addressing complex societal challenges that neither humans nor machines can solve alone.

Tobin South, Subramanya Nagabhushanaradhya, Ayesha Dissanayaka et al. · mit

The rapid rise of AI agents presents urgent challenges in authentication, authorization, and identity management. Current agent-centric protocols (like MCP) highlight the demand for clarified best practices in authentication and authorization. Looking ahead, ambitions for highly autonomous agents raise complex long-term questions regarding scalable access control, agent-centric identities, AI workload differentiation, and delegated authority. This OpenID Foundation whitepaper is for stakeholders at the intersection of AI agents and access management. It outlines the resources already available for securing today's agents and presents a strategic agenda to address the foundational authentication, authorization, and identity problems pivotal for tomorrow's widespread autonomous systems.

Zhenyu Zhang, Tianyi Chen, Weiran Xu et al.

Long-horizon tasks requiring multi-step reasoning and dynamic re-planning remain challenging for large language models (LLMs). Sequential prompting methods are prone to context drift, loss of goal information, and recurrent failure cycles, while hierarchical prompting methods often weaken cross-level continuity or incur substantial runtime overhead. We introduce ReCAP (Recursive Context-Aware Reasoning and Planning), a hierarchical framework with shared context for reasoning and planning in LLMs. ReCAP combines three key mechanisms: (i) plan-ahead decomposition, in which the model generates a full subtask list, executes the first item, and refines the remainder; (ii) structured re-injection of parent plans, maintaining consistent multi-level context during recursive return; and (iii) memory-efficient execution, bounding the active prompt so costs scale linearly with task depth. Together these mechanisms align high-level goals with low-level actions, reduce redundant prompting, and preserve coherent context updates across recursion. Experiments demonstrate that ReCAP substantially improves subgoal alignment and success rates on various long-horizon reasoning benchmarks, achieving a 32% gain on synchronous Robotouille and a 29% improvement on asynchronous Robotouille under the strict pass@1 protocol.

Shayne Longpre, Sneha Kudugunta, Niklas Muennighoff et al.

Scaling laws research has focused overwhelmingly on English -- yet the most prominent AI models explicitly serve billions of international users. In this work, we undertake the largest multilingual scaling laws study to date, totaling 774 multilingual training experiments, spanning 10M-8B model parameters, 400+ training languages and 48 evaluation languages. We introduce the Adaptive Transfer Scaling Law (ATLAS) for both monolingual and multilingual pretraining, which outperforms existing scaling laws' out-of-sample generalization often by more than 0.3 R^2. Our analyses of the experiments shed light on multilingual learning dynamics, transfer properties between languages, and the curse of multilinguality. First, we derive a cross-lingual transfer matrix, empirically measuring mutual benefit scores between 38 x 38=1444 language pairs. Second, we derive a language-agnostic scaling law that reveals how to optimally scale model size and data when adding languages without sacrificing performance. Third, we identify the computational crossover points for when to pretrain from scratch versus finetune from multilingual checkpoints. We hope these findings provide the scientific foundation for democratizing scaling laws across languages, and enable practitioners to efficiently scale models -- beyond English-first AI.

Shenzhe Zhu, Shu Yang, Michiel A. Bakker et al.

Large-scale public deliberations generate thousands of free-form contributions that must be synthesized into representative and neutral summaries for policy use. While LLMs have been shown as a promising tool to generate summaries for large-scale deliberations, they also risk underrepresenting minority perspectives and exhibiting bias with respect to the input order, raising fairness concerns in high-stakes contexts. Studying and fixing these issues requires a comprehensive evaluation at a large scale, yet current practice often relies on LLMs as judges, which show weak alignment with human judgments. To address this, we present DeliberationBank, a large-scale human-grounded dataset with (1) opinion data spanning ten deliberation questions created by 3,000 participants and (2) summary judgment data annotated by 4,500 participants across four dimensions (representativeness, informativeness, neutrality, policy approval). Using these datasets, we train DeliberationJudge, a fine-tuned DeBERTa model that can rate deliberation summaries from individual perspectives. DeliberationJudge is more efficient and more aligned with human judgements compared to a wide range of LLM judges. With DeliberationJudge, we evaluate 18 LLMs and reveal persistent weaknesses in deliberation summarization, especially underrepresentation of minority positions. Our framework provides a scalable and reliable way to evaluate deliberation summarization, helping ensure AI systems are more representative and equitable for policymaking.

William Cashman, Chasen Milner, Michael Houle et al.

AI development requires high fidelity testing environments to effectively transition from the laboratory to operations. The flexibility offered by cyber arenas presents a novel opportunity to test new artificial intelligence (AI) capabilities with users. Cyber arenas are designed to expose end-users to real-world situations and must rapidly incorporate evolving capabilities to meet their core objectives. To explore this concept the MIT/IEEE/Amazon Graph Challenge Anonymized Network Sensor was deployed in a cyber arena during a National Guard exercise.

Shenzhe Zhu, Jiao Sun, Yi Nian et al. · mit

AI agents are increasingly used in consumer-facing applications to assist with tasks such as product search, negotiation, and transaction execution. In this paper, we explore a future scenario where both consumers and merchants authorize AI agents to fully automate negotiations and transactions. We aim to answer two key questions: (1) Do different LLM agents vary in their ability to secure favorable deals for users? (2) What risks arise from fully automating deal-making with AI agents in consumer markets? To address these questions, we develop an experimental framework that evaluates the performance of various LLM agents in real-world negotiation and transaction settings. Our findings reveal that AI-mediated deal-making is an inherently imbalanced game -- different agents achieve significantly different outcomes for their users. Moreover, behavioral anomalies in LLMs can result in financial losses for both consumers and merchants, such as overspending or accepting unreasonable deals. These results underscore that while automation can improve efficiency, it also introduces substantial risks. Users should exercise caution when delegating business decisions to AI agents.

Shayne Longpre, Nikhil Singh, Manuel Cherep et al. · mit

Progress in AI is driven largely by the scale and quality of training data. Despite this, there is a deficit of empirical analysis examining the attributes of well-established datasets beyond text. In this work we conduct the largest and first-of-its-kind longitudinal audit across modalities--popular text, speech, and video datasets--from their detailed sourcing trends and use restrictions to their geographical and linguistic representation. Our manual analysis covers nearly 4000 public datasets between 1990-2024, spanning 608 languages, 798 sources, 659 organizations, and 67 countries. We find that multimodal machine learning applications have overwhelmingly turned to web-crawled, synthetic, and social media platforms, such as YouTube, for their training sets, eclipsing all other sources since 2019. Secondly, tracing the chain of dataset derivations we find that while less than 33% of datasets are restrictively licensed, over 80% of the source content in widely-used text, speech, and video datasets, carry non-commercial restrictions. Finally, counter to the rising number of languages and geographies represented in public AI training datasets, our audit demonstrates measures of relative geographical and multilingual representation have failed to significantly improve their coverage since 2013. We believe the breadth of our audit enables us to empirically examine trends in data sourcing, restrictions, and Western-centricity at an ecosystem-level, and that visibility into these questions are essential to progress in responsible AI. As a contribution to ongoing improvements in dataset transparency and responsible use, we release our entire multimodal audit, allowing practitioners to trace data provenance across text, speech, and video.

Jeremy Kepner, Jonathan Bernays, Stephen Buckley et al.

Adversarial Internet robots (botnets) represent a growing threat to the safe use and stability of the Internet. Botnets can play a role in launching adversary reconnaissance (scanning and phishing), influence operations (upvoting), and financing operations (ransomware, market manipulation, denial of service, spamming, and ad click fraud) while obfuscating tailored tactical operations. Reducing the presence of botnets on the Internet, with the aspirational target of zero, is a powerful vision for galvanizing policy action. Setting a global goal, encouraging international cooperation, creating incentives for improving networks, and supporting entities for botnet takedowns are among several policies that could advance this goal. These policies raise significant questions regarding proper authorities/access that cannot be answered in the abstract. Systems analysis has been widely used in other domains to achieve sufficient detail to enable these questions to be dealt with in concrete terms. Defeating botnets using an observe-pursue-counter architecture is analyzed, the technical feasibility is affirmed, and the authorities/access questions are significantly narrowed. Recommended next steps include: supporting the international botnet takedown community, expanding network observatories, enhancing the underlying network science at scale, conducting detailed systems analysis, and developing appropriate policy frameworks.

Udari Madhushani, Abhimanyu Dubey, Naomi Ehrich Leonard et al.

The cooperative bandit problem is increasingly becoming relevant due to its applications in large-scale decision-making. However, most research for this problem focuses exclusively on the setting with perfect communication, whereas in most real-world distributed settings, communication is often over stochastic networks, with arbitrary corruptions and delays. In this paper, we study cooperative bandit learning under three typical real-world communication scenarios, namely, (a) message-passing over stochastic time-varying networks, (b) instantaneous reward-sharing over a network with random delays, and (c) message-passing with adversarially corrupted rewards, including byzantine communication. For each of these environments, we propose decentralized algorithms that achieve competitive performance, along with near-optimal guarantees on the incurred group regret as well. Furthermore, in the setting with perfect communication, we present an improved delayed-update algorithm that outperforms the existing state-of-the-art on various network topologies. Finally, we present tight network-dependent minimax lower bounds on the group regret. Our proposed algorithms are straightforward to implement and obtain competitive empirical performance.

Ding Lyu, Yuan Yuan, Lin Wang et al.

Long ties, the social ties that bridge different communities, are widely believed to play crucial roles in spreading novel information in social networks. However, some existing network theories and prediction models indicate that long ties might dissolve quickly or eventually become redundant, thus putting into question the long-term value of long ties. Our empirical analysis of real-world dynamic networks shows that contrary to such reasoning, long ties are more likely to persist than other social ties, and that many of them constantly function as social bridges without being embedded in local networks. Using a novel cost-benefit analysis model combined with machine learning, we show that long ties are highly beneficial, which instinctively motivates people to expend extra effort to maintain them. This partly explains why long ties are more persistent than what has been suggested by many existing theories and models. Overall, our study suggests the need for social interventions that can promote the formation of long ties, such as mixing people with diverse backgrounds.

Abhimanyu Dubey, Vignesh Ramanathan, Alex Pentland et al.

Invariant approaches have been remarkably successful in tackling the problem of domain generalization, where the objective is to perform inference on data distributions different from those used in training. In our work, we investigate whether it is possible to leverage domain information from the unseen test samples themselves. We propose a domain-adaptive approach consisting of two steps: a) we first learn a discriminative domain embedding from unsupervised training examples, and b) use this domain embedding as supplementary information to build a domain-adaptive model, that takes both the input as well as its domain into account while making predictions. For unseen domains, our method simply uses few unlabelled test examples to construct the domain embedding. This enables adaptive classification on any unseen domain. Our approach achieves state-of-the-art performance on various domain generalization benchmarks. In addition, we introduce the first real-world, large-scale domain generalization benchmark, Geo-YFCC, containing 1.1M samples over 40 training, 7 validation, and 15 test domains, orders of magnitude larger than prior work. We show that the existing approaches either do not scale to this dataset or underperform compared to the simple baseline of training a model on the union of data from all training domains. In contrast, our approach achieves a significant improvement.

Abhimanyu Dubey, Alex Pentland

Reinforcement learning in cooperative multi-agent settings has recently advanced significantly in its scope, with applications in cooperative estimation for advertising, dynamic treatment regimes, distributed control, and federated learning. In this paper, we discuss the problem of cooperative multi-agent RL with function approximation, where a group of agents communicates with each other to jointly solve an episodic MDP. We demonstrate that via careful message-passing and cooperative value iteration, it is possible to achieve near-optimal no-regret learning even with a fixed constant communication budget. Next, we demonstrate that even in heterogeneous cooperative settings, it is possible to achieve Pareto-optimal no-regret learning with limited communication. Our work generalizes several ideas from the multi-agent contextual and multi-armed bandit literature to MDPs and reinforcement learning.

Abhimanyu Dubey, Alex Pentland

The rapid proliferation of decentralized learning systems mandates the need for differentially-private cooperative learning. In this paper, we study this in context of the contextual linear bandit: we consider a collection of agents cooperating to solve a common contextual bandit, while ensuring that their communication remains private. For this problem, we devise \textsc{FedUCB}, a multiagent private algorithm for both centralized and decentralized (peer-to-peer) federated learning. We provide a rigorous technical analysis of its utility in terms of regret, improving several results in cooperative bandit learning, and provide rigorous privacy guarantees as well. Our algorithms provide competitive performance both in terms of pseudoregret bounds and empirical benchmark performance in various multi-agent settings.

Thomas Hardjono, Alexander Lipton, Alex Pentland

We introduce the contract service provider (CSP) model as an analog of the successful Internet ISP model. Our exploration is motivated by the need to seek alternative blockchain service-fee models that departs from the token-for-operations (gas fee) model for smart contracts found on many popular blockchain platforms today. A given CSP community consisting of multiple CSP business entities (VASPs) form a contract domain which implement well-defined contract primitives, policies and contract-ledger. The nodes of the members of CSP community form the blockchain network. We discuss a number of design principles borrowed from the design principles of the Internet Architecture, and we discuss the interoperability of cross-domain (cross-chain) transfers of virtual assets in the context of contract domains.

Abhimanyu Dubey, Alex Pentland

We study the heavy-tailed stochastic bandit problem in the cooperative multi-agent setting, where a group of agents interact with a common bandit problem, while communicating on a network with delays. Existing algorithms for the stochastic bandit in this setting utilize confidence intervals arising from an averaging-based communication protocol known as~\textit{running consensus}, that does not lend itself to robust estimation for heavy-tailed settings. We propose \textsc{MP-UCB}, a decentralized multi-agent algorithm for the cooperative stochastic bandit that incorporates robust estimation with a message-passing protocol. We prove optimal regret bounds for \textsc{MP-UCB} for several problem settings, and also demonstrate its superiority to existing methods. Furthermore, we establish the first lower bounds for the cooperative bandit problem, in addition to providing efficient algorithms for robust bandit estimation of location.

Abhimanyu Dubey, Alex Pentland

Cooperative multi-agent decision making involves a group of agents cooperatively solving learning problems while communicating over a network with delays. In this paper, we consider the kernelised contextual bandit problem, where the reward obtained by an agent is an arbitrary linear function of the contexts' images in the related reproducing kernel Hilbert space (RKHS), and a group of agents must cooperate to collectively solve their unique decision problems. For this problem, we propose \textsc{Coop-KernelUCB}, an algorithm that provides near-optimal bounds on the per-agent regret, and is both computationally and communicatively efficient. For special cases of the cooperative problem, we also provide variants of \textsc{Coop-KernelUCB} that provides optimal per-agent regret. In addition, our algorithm generalizes several existing results in the multi-agent bandit setting. Finally, on a series of both synthetic and real-world multi-agent network benchmarks, we demonstrate that our algorithm significantly outperforms existing benchmarks.

Yan Leng, Tara Sowrirajan, Alex Pentland

Decision-making on networks can be explained by both homophily and social influence. While homophily drives the formation of communities with similar characteristics, social influence occurs both within and between communities. Social influence can be reasoned through role theory, which indicates that the influences among individuals depend on their roles and the behavior of interest. To operationalize these social science theories, we empirically identify the homophilous communities and use the community structures to capture the "roles", which affect the particular decision-making processes. We propose a generative model named Stochastic Block Influence Model and jointly analyze both the network formation and the behavioral influence within and between different empirically-identified communities. To evaluate the performance and demonstrate the interpretability of our method, we study the adoption decisions of microfinance in an Indian village. We show that although individuals tend to form links within communities, there are strong positive and negative social influences between communities, supporting the weak tie theory. Moreover, we find that communities with shared characteristics are associated with positive influence. In contrast, the communities with a lack of overlap are associated with negative influence. Our framework facilitates the quantification of the influences underlying decision communities and is thus a useful tool for driving information diffusion, viral marketing, and technology adoptions.

Thomas Hardjono, Alexander Lipton, Alex Pentland

The emerging virtual asset service providers (VASP) industry currently faces a number of challenges related to the Travel Rule, notably pertaining to customer personal information, account number and cryptographic key information. VASPs will be handling virtual assets of different forms, where each may be bound to different private-public key pairs on the blockchain. As such, VASPs also face the additional problem of the management of its own keys and the management of customer keys that may reside in a customer wallet. The use of attestation technologies as applied to wallet systems may provide VASPs with suitable evidence relevant to the Travel Rule regarding cryptographic key information and their operational state. Additionally, wallet attestations may provide crypto-asset insurers with strong evidence regarding the key management aspects of a wallet device, thereby providing the insurance industry with measurable levels of assurance that can become the basis for insurers to perform risk assessment on crypto-assets bound to keys in wallets, both enterprise-grade wallets and consumer-grade wallets.

Thomas Hardjono, Alexander Lipton, Alex Pentland

In order for VASPs to fulfill the regulatory requirements from the FATF and the Travel Rule, VASPs need access to truthful information regarding originators, beneficiaries and other VASPs involved in a virtual asset transfer instance. Additionally, in seeking data regarding subjects (individuals or organizations) VASPs are faced with privacy regulations such as the GDPR and CCPA. In this paper we a propose privacy-preserving claims issuance model that carries indicators of the provenance of the data and the algorithms used to derive the claim or assertion. This allows VASPs to obtain originator and beneficiary information without necessarily having access to the private data about these entities. Secondly we propose the use of a consortium trust network arrangement for VASPs to exchange signed claims about subjects and their public-key information or certificate.

Thomas Hardjono, Alex Pentland

Over the last decade there has been a continuing decline in social trust on the part of individuals with regards to the handling and fair use of personal data, digital assets and other related rights in general. At the same time, there has been a change in the employment patterns for many people through the emergence of the gig economy. These gig workers include artists, songwriters and musicians in the music industry. We discuss the notion of the data cooperative with fiduciary responsibilities to its members, which is similar in purpose to credit unions in the financial sector. A data cooperative for artists and musicians allows the community to share IT resources, such as data storage, analytics processing, blockchains and distributed ledgers. A cooperative can also employ smart contracts to remedy the various challenges currently faced by the music industry with regards to the license tracking management.

Michiel A. Bakker, Duy Patrick Tu, Humberto Riverón Valdés et al.

We introduce a framework for dynamic adversarial discovery of information (DADI), motivated by a scenario where information (a feature set) is used by third parties with unknown objectives. We train a reinforcement learning agent to sequentially acquire a subset of the information while balancing accuracy and fairness of predictors downstream. Based on the set of already acquired features, the agent decides dynamically to either collect more information from the set of available features or to stop and predict using the information that is currently available. Building on previous work exploring adversarial representation learning, we attain group fairness (demographic parity) by rewarding the agent with the adversary's loss, computed over the final feature set. Importantly, however, the framework provides a more general starting point for fair or private dynamic information discovery. Finally, we demonstrate empirically, using two real-world datasets, that we can trade-off fairness and predictive performance

Thomas Hardjono, Alexander Lipton, Alex Pentland

The recent FATF Recommendations defines virtual assets and virtual assets service providers (VASP), and requires under the Travel Rule that originating VASPs obtain and hold required and accurate originator information and required beneficiary information on virtual asset transfers. In this paper we discuss the notion of key ownership evidence as a core part of originator and beneficiary information required by the FATF Recommendation. We discuss approaches to securely communicate the originator and beneficiary information between VASPs, and review existing standards for public key certificates as applied to VASPs and virtual asset transfers. We propose the notion of a trust network of VASPs in which originator and beneficiary information, including key ownership information, can be exchanged securely while observing individual privacy requirements.

Abhimanyu Dubey, Alex Pentland

Thompson Sampling provides an efficient technique to introduce prior knowledge in the multi-armed bandit problem, along with providing remarkable empirical performance. In this paper, we revisit the Thompson Sampling algorithm under rewards drawn from symmetric $α$-stable distributions, which are a class of heavy-tailed probability distributions utilized in finance and economics, in problems such as modeling stock prices and human behavior. We present an efficient framework for posterior inference, which leads to two algorithms for Thompson Sampling in this setting. We prove finite-time regret bounds for both algorithms, and demonstrate through a series of experiments the stronger performance of Thompson Sampling in this setting. With our results, we provide an exposition of symmetric $α$-stable distributions in sequential decision-making, and enable sequential Bayesian inference in applications from diverse fields in finance and complex systems that operate on heavy-tailed features.

Yuan Yuan, Tracy Liu, Chenhao Tan et al.

Gifts are important instruments for forming bonds in interpersonal relationships. Our study analyzes the phenomenon of gift contagion in online groups. Gift contagion encourages social bonds by prompting further gifts; it may also promote group interaction and solidarity. Using data on 36 million online red packet gifts on a large social site in East Asia, we leverage a natural experimental design to identify the social contagion of gift giving in online groups. Our natural experiment is enabled by the randomization of the gift amount allocation algorithm on the platform, which addresses the common challenge of causal identifications in observational data. Our study provides evidence of gift contagion: on average, receiving one additional dollar causes a recipient to send 18 cents back to the group within the subsequent 24 hours. Decomposing this effect, we find that it is mainly driven by the extensive margin -- more recipients are triggered to send red packets. Moreover, we find that this effect is stronger for "luckiest draw" recipients, suggesting the presence of a group norm regarding the next red packet sender. Finally, we investigate the moderating effects of group- and individual-level social network characteristics on gift contagion as well as the causal impact of receiving gifts on group network structure. Our study has implications for promoting group dynamics and designing marketing strategies for product adoption.

Thomas Hardjono, Alex Pentland

In this paper we address the issue of identity and access control within shared permissioned blockchains. We propose the ChainAchor system that provides anonymous but verifiable identities for entities on the blockchain. ChainAchor also provides access control to entities seeking to submit transactions to the blockchain to read/verify transactions on the the permissioned blockchain. Consensus nodes enforce access control to the shared permissioned blockchain by a simple look-up to a (read-only) list of anonymous members' public-keys. ChainAnchor also provides unlinkability of transactions belonging to an entity on the blockchain. This allows for an entity to optionally disclose their identity when a transaction is called into question (e.g. regulatory or compliance requirements), but without affecting the anonymity and unlinkability of their remaining transactions.

Dhaval Adjodah, Dan Calacci, Abhimanyu Dubey et al.

A common technique to improve learning performance in deep reinforcement learning (DRL) and many other machine learning algorithms is to run multiple learning agents in parallel. A neglected component in the development of these algorithms has been how best to arrange the learning agents involved to improve distributed search. Here we draw upon results from the networked optimization literatures suggesting that arranging learning agents in communication networks other than fully connected topologies (the implicit way agents are commonly arranged in) can improve learning. We explore the relative performance of four popular families of graphs and observe that one such family (Erdos-Renyi random graphs) empirically outperforms the de facto fully-connected communication topology across several DRL benchmark tasks. Additionally, we observe that 1000 learning agents arranged in an Erdos-Renyi graph can perform as well as 3000 agents arranged in the standard fully-connected topology, showing the large learning improvement possible when carefully designing the topology over which agents communicate. We complement these empirical results with a theoretical investigation of why our alternate topologies perform better. Overall, our work suggests that distributed machine learning algorithms could be made more effective if the communication topology between learning agents was optimized.

Yan Leng, Xiaowen Dong, Junfeng Wu et al.

Individuals, or organizations, cooperate with or compete against one another in a wide range of practical situations. Such strategic interactions are often modeled as games played on networks, where an individual's payoff depends not only on her action but also on that of her neighbors. The current literature has largely focused on analyzing the characteristics of network games in the scenario where the structure of the network, which is represented by a graph, is known beforehand. It is often the case, however, that the actions of the players are readily observable while the underlying interaction network remains hidden. In this paper, we propose two novel frameworks for learning, from the observations on individual actions, network games with linear-quadratic payoffs, and in particular, the structure of the interaction network. Our frameworks are based on the Nash equilibrium of such games and involve solving a joint optimization problem for the graph structure and the individual marginal benefits. Both synthetic and real-world experiments demonstrate the effectiveness of the proposed frameworks, which have theoretical as well as practical implications for understanding strategic interactions in a network environment.

Alejandro Noriega-Campero, Michiel A. Bakker, Bernardo Garcia-Bulle et al.

Society increasingly relies on machine learning models for automated decision making. Yet, efficiency gains from automation have come paired with concern for algorithmic discrimination that can systematize inequality. Recent work has proposed optimal post-processing methods that randomize classification decisions for a fraction of individuals, in order to achieve fairness measures related to parity in errors and calibration. These methods, however, have raised concern due to the information inefficiency, intra-group unfairness, and Pareto sub-optimality they entail. The present work proposes an alternative active framework for fair classification, where, in deployment, a decision-maker adaptively acquires information according to the needs of different groups or individuals, towards balancing disparities in classification performance. We propose two such methods, where information collection is adapted to group- and individual-level needs respectively. We show on real-world datasets that these can achieve: 1) calibration and single error parity (e.g., equal opportunity); and 2) parity in both false positive and false negative rates (i.e., equal odds). Moreover, we show that by leveraging their additional degree of freedom, active approaches can substantially outperform randomization-based classifiers previously considered optimal, while avoiding limitations such as intra-group unfairness.

Alejandro Noriega-Campero, Alex Rutherford, Oren Lederman et al.

Today's age of data holds high potential to enhance the way we pursue and monitor progress in the fields of development and humanitarian action. We study the relation between data utility and privacy risk in large-scale behavioral data, focusing on mobile phone metadata as paradigmatic domain. To measure utility, we survey experts about the value of mobile phone metadata at various spatial and temporal granularity levels. To measure privacy, we propose a formal and intuitive measure of reidentification risk$\unicode{x2014}$the information ratio$\unicode{x2014}$and compute it at each granularity level. Our results confirm the existence of a stark tradeoff between data utility and reidentifiability, where the most valuable datasets are also most prone to reidentification. When data is specified at ZIP-code and hourly levels, outside knowledge of only 7% of a person's data suffices for reidentification and retrieval of the remaining 93%. In contrast, in the least valuable dataset, specified at municipality and daily levels, reidentification requires on average outside knowledge of 51%, or 31 data points, of a person's data to retrieve the remaining 49%. Overall, our findings show that coarsening data directly erodes its value, and highlight the need for using data-coarsening, not as stand-alone mechanism, but in combination with data-sharing models that provide adjustable degrees of accountability and security.

Thomas Hardjono, Alexander Lipton, Alex Pentland

In this paper we discuss a design philosophy for interoperable blockchain systems, using the design philosophy of the Internet architecture as the basis to identify key design principles. Several interoperability challenges are discussed in the context of cross-domain transactions. We illustrate how these principles are informing the interoperability architecture of the MIT Tradecoin system.

Eduardo Castelló Ferrer, Ognjen Rudovic, Thomas Hardjono et al.

Robots have potential to revolutionize the way we interact with the world around us. One of their largest potentials is in the domain of mobile health where they can be used to facilitate clinical interventions. However, to accomplish this, robots need to have access to our private data in order to learn from these data and improve their interaction capabilities. Furthermore, to enhance this learning process, the knowledge sharing among multiple robot units is the natural step forward. However, to date, there is no well-established framework which allows for such data sharing while preserving the privacy of the users (e.g., the hospital patients). To this end, we introduce RoboChain - the first learning framework for secure, decentralized and computationally efficient data and model sharing among multiple robot units installed at multiple sites (e.g., hospitals). RoboChain builds upon and combines the latest advances in open data access and blockchain technologies, as well as machine learning. We illustrate this framework using the example of a clinical intervention conducted in a private network of hospitals. Specifically, we lay down the system architecture that allows multiple robot units, conducting the interventions at different hospitals, to perform efficient learning without compromising the data privacy.

Peter M Krafft, Nicolás Della Penna, Alex Pentland

As cryptocurrencies gain popularity and credibility, marketplaces for cryptocurrencies are growing in importance. Understanding the dynamics of these markets can help to assess how viable the cryptocurrnency ecosystem is and how design choices affect market behavior. One existential threat to cryptocurrencies is dramatic fluctuations in traders' willingness to buy or sell. Using a novel experimental methodology, we conducted an online experiment to study how susceptible traders in these markets are to peer influence from trading behavior. We created bots that executed over one hundred thousand trades costing less than a penny each in 217 cryptocurrencies over the course of six months. We find that individual "buy" actions led to short-term increases in subsequent buy-side activity hundreds of times the size of our interventions. From a design perspective, we note that the design choices of the exchange we study may have promoted this and other peer influence effects, which highlights the potential social and economic impact of HCI in the design of digital institutions.

Dhaval Adjodah, Dan Calacci, Yan Leng et al.

We draw upon a previously largely untapped literature on human collective intelligence as a source of inspiration for improving deep learning. Implicit in many algorithms that attempt to solve Deep Reinforcement Learning (DRL) tasks is the network of processors along which parameter values are shared. So far, existing approaches have implicitly utilized fully-connected networks, in which all processors are connected. However, the scientific literature on human collective intelligence suggests that complete networks may not always be the most effective information network structures for distributed search through complex spaces. Here we show that alternative topologies can improve deep neural network training: we find that sparser networks learn higher rewards faster, leading to learning improvements at lower communication costs.

Antonio L. Alfeo, Mario G. C. A. Cimino, Sara Egidi et al.

Positioning data offer a remarkable source of information to analyze crowds urban dynamics. However, discovering urban activity patterns from the emergent behavior of crowds involves complex system modeling. An alternative approach is to adopt computational techniques belonging to the emergent paradigm, which enables self-organization of data and allows adaptive analysis. Specifically, our approach is based on stigmergy. By using stigmergy each sample position is associated with a digital pheromone deposit, which progressively evaporates and aggregates with other deposits according to their spatiotemporal proximity. Based on this principle, we exploit positioning data to identify high density areas (hotspots) and characterize their activity over time. This characterization allows the comparison of dynamics occurring in different days, providing a similarity measure exploitable by clustering techniques. Thus, we cluster days according to their activity behavior, discovering unexpected urban activity patterns. As a case study, we analyze taxi traces in New York City during 2015.