Julie Shah

Yilun Zhou, Marco Tulio Ribeiro, Julie Shah · microsoft-research, uw

Interpretability methods are developed to understand the working mechanisms of black-box models, which is crucial to their responsible deployment. Fulfilling this goal requires both that the explanations generated by these methods are correct and that people can easily and reliably understand them. While the former has been addressed in prior work, the latter is often overlooked, resulting in informal model understanding derived from a handful of local explanations. In this paper, we introduce explanation summary (ExSum), a mathematical framework for quantifying model understanding, and propose metrics for its quality assessment. On two domains, ExSum highlights various limitations in the current practice, helps develop accurate model understanding, and reveals easily overlooked properties of the model. We also connect understandability to other properties of explanations such as human alignment, robustness, and counterfactual minimality and plausibility.

Mycal Tucker, Julie Shah, Roger Levy et al. · mit

Emergent communication research often focuses on optimizing task-specific utility as a driver for communication. However, human languages appear to evolve under pressure to efficiently compress meanings into communication signals by optimizing the Information Bottleneck tradeoff between informativeness and complexity. In this work, we study how trading off these three factors -- utility, informativeness, and complexity -- shapes emergent communication, including compared to human communication. To this end, we propose Vector-Quantized Variational Information Bottleneck (VQ-VIB), a method for training neural agents to compress inputs into discrete signals embedded in a continuous space. We train agents via VQ-VIB and compare their performance to previously proposed neural architectures in grounded environments and in a Lewis reference game. Across all neural architectures and settings, taking into account communicative informativeness benefits communication convergence rates, and penalizing communicative complexity leads to human-like lexicon sizes while maintaining high utility. Additionally, we find that VQ-VIB outperforms other discrete communication methods. This work demonstrates how fundamental principles that are believed to characterize human language evolution may inform emergent communication in artificial agents.

Mycal Tucker, Tiwalayo Eisape, Peng Qian et al. · mit

Recent causal probing literature reveals when language models and syntactic probes use similar representations. Such techniques may yield "false negative" causality results: models may use representations of syntax, but probes may have learned to use redundant encodings of the same syntactic information. We demonstrate that models do encode syntactic information redundantly and introduce a new probe design that guides probes to consider all syntactic information present in embeddings. Using these probes, we find evidence for the use of syntax in models where prior methods did not, allowing us to boost model performance by injecting syntactic information into representations.

Peter Stone, Rodney Brooks, Erik Brynjolfsson et al.

In September 2016, Stanford's "One Hundred Year Study on Artificial Intelligence" project (AI100) issued the first report of its planned long-term periodic assessment of artificial intelligence (AI) and its impact on society. It was written by a panel of 17 study authors, each of whom is deeply rooted in AI research, chaired by Peter Stone of the University of Texas at Austin. The report, entitled "Artificial Intelligence and Life in 2030," examines eight domains of typical urban settings on which AI is likely to have impact over the coming years: transportation, home and service robots, healthcare, education, public safety and security, low-resource communities, employment and workplace, and entertainment. It aims to provide the general public with a scientifically and technologically accurate portrayal of the current state of AI and its potential and to help guide decisions in industry and governments, as well as to inform research and development in the field. The charge for this report was given to the panel by the AI100 Standing Committee, chaired by Barbara Grosz of Harvard University.

Mycal Tucker, Julie Shah · mit

Artificial neural nets can represent and classify many types of data but are often tailored to particular applications -- e.g., for "fair" or "hierarchical" classification. Once an architecture has been selected, it is often difficult for humans to adjust models for a new task; for example, a hierarchical classifier cannot be easily transformed into a fair classifier that shields a protected field. Our contribution in this work is a new neural network architecture, the concept subspace network (CSN), which generalizes existing specialized classifiers to produce a unified model capable of learning a spectrum of multi-concept relationships. We demonstrate that CSNs reproduce state-of-the-art results in fair classification when enforcing concept independence, may be transformed into hierarchical classifiers, or even reconcile fairness and hierarchy within a single classifier. The CSN is inspired by existing prototype-based classifiers that promote interpretability.

Michael L. Littman, Ifeoma Ajunwa, Guy Berger et al.

In September 2021, the "One Hundred Year Study on Artificial Intelligence" project (AI100) issued the second report of its planned long-term periodic assessment of artificial intelligence (AI) and its impact on society. It was written by a panel of 17 study authors, each of whom is deeply rooted in AI research, chaired by Michael Littman of Brown University. The report, entitled "Gathering Strength, Gathering Storms," answers a set of 14 questions probing critical areas of AI development addressing the major risks and dangers of AI, its effects on society, its public perception and the future of the field. The report concludes that AI has made a major leap from the lab to people's lives in recent years, which increases the urgency to understand its potential negative effects. The questions were developed by the AI100 Standing Committee, chaired by Peter Stone of the University of Texas at Austin, consisting of a group of AI leaders with expertise in computer science, sociology, ethics, economics, and other disciplines.

Yanwei Wang, Nadia Figueroa, Shen Li et al.

Learning from demonstration (LfD) has succeeded in tasks featuring a long time horizon. However, when the problem complexity also includes human-in-the-loop perturbations, state-of-the-art approaches do not guarantee the successful reproduction of a task. In this work, we identify the roots of this challenge as the failure of a learned continuous policy to satisfy the discrete plan implicit in the demonstration. By utilizing modes (rather than subgoals) as the discrete abstraction and motion policies with both mode invariance and goal reachability properties, we prove our learned continuous policy can simulate any discrete plan specified by a linear temporal logic (LTL) formula. Consequently, an imitator is robust to both task- and motion-level perturbations and guaranteed to achieve task success. Project page: https://yanweiw.github.io/tli/

Lindsay Sanneman, Mycal Tucker, Julie Shah · mit

Recent advances in artificial intelligence (AI) have underscored the need for explainable AI (XAI) to support human understanding of AI systems. Consideration of human factors that impact explanation efficacy, such as mental workload and human understanding, is central to effective XAI design. Existing work in XAI has demonstrated a tradeoff between understanding and workload induced by different types of explanations. Explaining complex concepts through abstractions (hand-crafted groupings of related problem features) has been shown to effectively address and balance this workload-understanding tradeoff. In this work, we characterize the workload-understanding balance via the Information Bottleneck method: an information-theoretic approach which automatically generates abstractions that maximize informativeness and minimize complexity. In particular, we establish empirical connections between workload and complexity and between understanding and informativeness through human-subject experiments. This empirical link between human factors and information-theoretic concepts provides an important mathematical characterization of the workload-understanding tradeoff which enables user-tailored XAI design.

Andi Peng, Aviv Netanyahu, Mark Ho et al.

Policies often fail due to distribution shift -- changes in the state and reward that occur when a policy is deployed in new environments. Data augmentation can increase robustness by making the model invariant to task-irrelevant changes in the agent's observation. However, designers don't know which concepts are irrelevant a priori, especially when different end users have different preferences about how the task is performed. We propose an interactive framework to leverage feedback directly from the user to identify personalized task-irrelevant concepts. Our key idea is to generate counterfactual demonstrations that allow users to quickly identify possible task-relevant and irrelevant concepts. The knowledge of task-irrelevant concepts is then used to perform data augmentation and thus obtain a policy adapted to personalized user objectives. We present experiments validating our framework on discrete and continuous control tasks with real human users. Our method (1) enables users to better understand agent failure, (2) reduces the number of demonstrations required for fine-tuning, and (3) aligns the agent to individual user task preferences.

Yilun Zhou, Julie Shah

Feature attribution methods are popular for explaining neural network predictions, and they are often evaluated on metrics such as comprehensiveness and sufficiency. In this paper, we highlight an intriguing property of these metrics: their solvability. Concretely, we can define the problem of optimizing an explanation for a metric, which can be solved by beam search. This observation leads to the obvious yet unaddressed question: why do we use explainers (e.g., LIME) not based on solving the target metric, if the metric value represents explanation quality? We present a series of investigations showing strong performance of this beam search explainer and discuss its broader implication: a definition-evaluation duality of interpretability concepts. We implement the explainer and release the Python solvex package for models of text, image and tabular domains.

Andi Peng, Mycal Tucker, Eoin Kenny et al. · mit

Neural networks often learn task-specific latent representations that fail to generalize to novel settings or tasks. Conversely, humans learn discrete representations (i.e., concepts or words) at a variety of abstraction levels (e.g., "bird" vs. "sparrow") and deploy the appropriate abstraction based on task. Inspired by this, we train neural models to generate a spectrum of discrete representations, and control the complexity of the representations (roughly, how many bits are allocated for encoding inputs) by tuning the entropy of the distribution over representations. In finetuning experiments, using only a small number of labeled examples for a new task, we show that (1) tuning the representation to a task-appropriate complexity level supports the highest finetuning performance, and (2) in a human-participant study, users were able to identify the appropriate complexity level for a downstream task using visualizations of discrete representations. Our results indicate a promising direction for rapid model finetuning by leveraging human insight.

Andreea Bobu, Andi Peng, Pulkit Agrawal et al.

To act in the world, robots rely on a representation of salient task aspects: for example, to carry a coffee mug, a robot may consider movement efficiency or mug orientation in its behavior. However, if we want robots to act for and with people, their representations must not be just functional but also reflective of what humans care about, i.e. they must be aligned. We observe that current learning approaches suffer from representation misalignment, where the robot's learned representation does not capture the human's representation. We suggest that because humans are the ultimate evaluator of robot performance, we must explicitly focus our efforts on aligning learned representations with humans, in addition to learning the downstream task. We advocate that current representation learning approaches in robotics should be studied from the perspective of how well they accomplish the objective of representation alignment. We mathematically define the problem, identify its key desiderata, and situate current methods within this formalism. We conclude by suggesting future directions for exploring open challenges.

Yiming Zheng, Serena Booth, Julie Shah et al.

Neural rationale models are popular for interpretable predictions of NLP tasks. In these, a selector extracts segments of the input text, called rationales, and passes these segments to a classifier for prediction. Since the rationale is the only information accessible to the classifier, it is plausibly defined as the explanation. Is such a characterization unconditionally correct? In this paper, we argue to the contrary, with both philosophical perspectives and empirical evidence suggesting that rationale models are, perhaps, less rational and interpretable than expected. We call for more rigorous and comprehensive evaluations of these models to ensure desired properties of interpretability are indeed achieved. The code can be found at https://github.com/yimingz89/Neural-Rationale-Analysis.

Yilun Zhou, Serena Booth, Marco Tulio Ribeiro et al.

Feature attribution methods are popular in interpretable machine learning. These methods compute the attribution of each input feature to represent its importance, but there is no consensus on the definition of "attribution", leading to many competing methods with little systematic evaluation, complicated in particular by the lack of ground truth attribution. To address this, we propose a dataset modification procedure to induce such ground truth. Using this procedure, we evaluate three common methods: saliency maps, rationales, and attentions. We identify several deficiencies and add new perspectives to the growing body of evidence questioning the correctness and reliability of these methods applied on datasets in the wild. We further discuss possible avenues for remedy and recommend new attribution methods to be tested against ground truth before deployment. The code is available at https://github.com/YilunZhou/feature-attribution-evaluation

Serena Booth, Yilun Zhou, Ankit Shah et al.

Post-hoc explanation methods are gaining popularity for interpreting, understanding, and debugging neural networks. Most analyses using such methods explain decisions in response to inputs drawn from the test set. However, the test set may have few examples that trigger some model behaviors, such as high-confidence failures or ambiguous classifications. To address these challenges, we introduce a flexible model inspection framework: Bayes-TrEx. Given a data distribution, Bayes-TrEx finds in-distribution examples with a specified prediction confidence. We demonstrate several use cases of Bayes-TrEx, including revealing highly confident (mis)classifications, visualizing class boundaries via ambiguous examples, understanding novel-class extrapolation behavior, and exposing neural network overconfidence. We use Bayes-TrEx to study classifiers trained on CLEVR, MNIST, and Fashion-MNIST, and we show that this framework enables more flexible holistic model analysis than just inspecting the test set. Code is available at https://github.com/serenabooth/Bayes-TrEx.

Yanwei Wang, Lirui Wang, Yilun Du et al.

Generative policies trained with human demonstrations can autonomously accomplish multimodal, long-horizon tasks. However, during inference, humans are often removed from the policy execution loop, limiting the ability to guide a pre-trained policy towards a specific sub-goal or trajectory shape among multiple predictions. Naive human intervention may inadvertently exacerbate distribution shift, leading to constraint violations or execution failures. To better align policy output with human intent without inducing out-of-distribution errors, we propose an Inference-Time Policy Steering (ITPS) framework that leverages human interactions to bias the generative sampling process, rather than fine-tuning the policy on interaction data. We evaluate ITPS across three simulated and real-world benchmarks, testing three forms of human interaction and associated alignment distance metrics. Among six sampling strategies, our proposed stochastic sampling with diffusion policy achieves the best trade-off between alignment and distribution shift. Videos are available at https://yanweiw.github.io/itps/.

Yanwei Wang, Tsun-Hsuan Wang, Jiayuan Mao et al.

Grounding the common-sense reasoning of Large Language Models (LLMs) in physical domains remains a pivotal yet unsolved problem for embodied AI. Whereas prior works have focused on leveraging LLMs directly for planning in symbolic spaces, this work uses LLMs to guide the search of task structures and constraints implicit in multi-step demonstrations. Specifically, we borrow from manipulation planning literature the concept of mode families, which group robot configurations by specific motion constraints, to serve as an abstraction layer between the high-level language representations of an LLM and the low-level physical trajectories of a robot. By replaying a few human demonstrations with synthetic perturbations, we generate coverage over the demonstrations' state space with additional successful executions as well as counterfactuals that fail the task. Our explanation-based learning framework trains an end-to-end differentiable neural network to predict successful trajectories from failures and as a by-product learns classifiers that ground low-level states and images in mode families without dense labeling. The learned grounding classifiers can further be used to translate language plans into reactive policies in the physical domain in an interpretable manner. We show our approach improves the interpretability and reactivity of imitation learning through 2D navigation and simulated and real robot manipulation tasks. Website: https://yanweiw.github.io/glide

Eike Schneiders, Christopher Fourie, Stanley Celestin et al.

Successful entrainment during collaboration positively affects trust, willingness to collaborate, and likeability towards collaborators. In this paper, we present a mixed-method study to investigate characteristics of successful entrainment leading to pair and group-based synchronisation. Drawing inspiration from industrial settings, we designed a fast-paced, short-cycle repetitive task. Using motion tracking, we investigated entrainment in both dyadic and triadic task completion. Furthermore, we utilise audio-video recordings and semi-structured interviews to contextualise participants' experiences. This paper contributes to the Human-Computer/Robot Interaction (HCI/HRI) literature using a human-centred approach to identify characteristics of entrainment during pair- and group-based collaboration. We present five characteristics related to successful entrainment. These are related to the occurrence of entrainment, leader-follower patterns, interpersonal communication, the importance of the point-of-assembly, and the value of acoustic feedback. Finally, we present three design considerations for future research and design on collaboration with robots.

Alex Cuellar, Michael Hagenow, Julie Shah

Effective human-robot teaming is crucial for the practical deployment of robots in human workspaces. However, optimizing joint human-robot plans remains a challenge due to the difficulty of modeling individualized human capabilities and preferences. While prior research has leveraged the multi-cycle structure of domains like manufacturing to learn an individual's tendencies and adapt plans over repeated interactions, these techniques typically consider task-level and motion-level adaptation in isolation. Task-level methods optimize allocation and scheduling but often ignore spatial interference in close-proximity scenarios; conversely, motion-level methods focus on collision avoidance while ignoring the broader task context. This paper introduces RAPIDDS, a framework that unifies these approaches by modeling an individual's spatial behavior (motion paths) and temporal behavior (time required to complete tasks) over multiple cycles. RAPIDDS then jointly adapts task schedules and steers diffusion models of robot motions to maximize efficiency and minimize proximity accounting for these individualized models. We demonstrate the importance of this dual adaptation through an ablation study in simulation and a physical robot scenario using a 7-DOF robot arm. Finally, we present a user study (n=32) showing significant plan improvement compared to non-adaptive systems across both objective metrics, such as efficiency and proximity, and subjective measures, including fluency and user preference. See this paper's companion video at: https://youtu.be/55Q3lq1fINs.

Cristian-Paul Bara, Ziqiao Ma, Yingzhuo Yu et al.

Collaborative tasks often begin with partial task knowledge and incomplete initial plans from each partner. To complete these tasks, agents need to engage in situated communication with their partners and coordinate their partial plans towards a complete plan to achieve a joint task goal. While such collaboration seems effortless in a human-human team, it is highly challenging for human-AI collaboration. To address this limitation, this paper takes a step towards collaborative plan acquisition, where humans and agents strive to learn and communicate with each other to acquire a complete plan for joint tasks. Specifically, we formulate a novel problem for agents to predict the missing task knowledge for themselves and for their partners based on rich perceptual and dialogue history. We extend a situated dialogue benchmark for symmetric collaborative tasks in a 3D blocks world and investigate computational strategies for plan acquisition. Our empirical results suggest that predicting the partner's missing knowledge is a more viable approach than predicting one's own. We show that explicit modeling of the partner's dialogue moves and mental states produces improved and more stable results than without. These results provide insight for future AI agents that can predict what knowledge their partner is missing and, therefore, can proactively communicate such information to help their partner acquire such missing knowledge toward a common understanding of joint tasks.

Mycal Tucker, William Kuhl, Khizer Shahid et al.

Neural nets are powerful function approximators, but the behavior of a given neural net, once trained, cannot be easily modified. We wish, however, for people to be able to influence neural agents' actions despite the agents never training with humans, which we formalize as a human-assisted decision-making problem. Inspired by prior art initially developed for model explainability, we develop a method for updating representations in pre-trained neural nets according to externally-specified properties. In experiments, we show how our method may be used to improve human-agent team performance for a variety of neural networks from image classifiers to agents in multi-agent reinforcement learning settings.

Lindsay Sanneman, Julie Shah

Explainable AI techniques that describe agent reward functions can enhance human-robot collaboration in a variety of settings. One context where human understanding of agent reward functions is particularly beneficial is in the value alignment setting. In the value alignment context, an agent aims to infer a human's reward function through interaction so that it can assist the human with their tasks. If the human can understand where gaps exist in the agent's reward understanding, they will be able to teach more efficiently and effectively, leading to quicker human-agent team performance improvements. In order to support human collaborators in the value alignment setting and similar contexts, it is first important to understand the effectiveness of different reward explanation techniques in a variety of domains. In this paper, we introduce a categorization of information modalities for reward explanation techniques, suggest a suite of assessment techniques for human reward understanding, and introduce four axes of domain complexity. We then propose an experiment to study the relative efficacy of a broad set of reward explanation techniques covering multiple modalities of information in a set of domains of varying complexity.

Mycal Tucker, Huao Li, Siddharth Agrawal et al.

Neural agents trained in reinforcement learning settings can learn to communicate among themselves via discrete tokens, accomplishing as a team what agents would be unable to do alone. However, the current standard of using one-hot vectors as discrete communication tokens prevents agents from acquiring more desirable aspects of communication such as zero-shot understanding. Inspired by word embedding techniques from natural language processing, we propose neural agent architectures that enables them to communicate via discrete tokens derived from a learned, continuous space. We show in a decision theoretic framework that our technique optimizes communication over a wide range of scenarios, whereas one-hot tokens are only optimal under restrictive assumptions. In self-play experiments, we validate that our trained agents learn to cluster tokens in semantically-meaningful ways, allowing them communicate in noisy environments where other techniques fail. Lastly, we demonstrate both that agents using our method can effectively respond to novel human communication and that humans can understand unlabeled emergent agent communication, outperforming the use of one-hot communication.

Ankit Shah, Pritish Kamath, Shen Li et al.

When observing task demonstrations, human apprentices are able to identify whether a given task is executed correctly long before they gain expertise in actually performing that task. Prior research into learning from demonstrations (LfD) has failed to capture this notion of the acceptability of a task's execution; meanwhile, temporal logics provide a flexible language for expressing task specifications. Inspired by this, we present Bayesian specification inference, a probabilistic model for inferring task specification as a temporal logic formula. We incorporate methods from probabilistic programming to define our priors, along with a domain-independent likelihood function to enable sampling-based inference. We demonstrate the efficacy of our model for inferring specifications, with over 90% similarity observed between the inferred specification and the ground truth, both within a synthetic domain and during a real-world table setting task.

Ramya Ramakrishnan, Vaibhav Unhelkar, Ece Kamar et al.

Trained AI systems and expert decision makers can make errors that are often difficult to identify and understand. Determining the root cause for these errors can improve future decisions. This work presents Generative Error Model (GEM), a generative model for inferring representational errors based on observations of an actor's behavior (either simulated agent, robot, or human). The model considers two sources of error: those that occur due to representational limitations -- "blind spots" -- and non-representational errors, such as those caused by noise in execution or systematic errors present in the actor's policy. Disambiguating these two error types allows for targeted refinement of the actor's policy (i.e., representational errors require perceptual augmentation, while other errors can be reduced through methods such as improved training or attention support). We present a Bayesian inference algorithm for GEM and evaluate its utility in recovering representational errors on multiple domains. Results show that our approach can recover blind spots of both reinforcement learning agents as well as human users.

Yilun Zhou, Serena Booth, Nadia Figueroa et al.

As robots are deployed in complex situations, engineers and end users must develop a holistic understanding of their behaviors, capabilities, and limitations. Some behaviors are directly optimized by the objective function. They often include success rate, completion time or energy consumption. Other behaviors -- e.g., collision avoidance, trajectory smoothness or motion legibility -- are typically emergent but equally important for safe and trustworthy deployment. Designing an objective which optimizes every aspect of robot behavior is hard. In this paper, we advocate for systematic analysis of a wide array of behaviors for holistic understanding of robot controllers and, to this end, propose a framework, RoCUS, which uses Bayesian posterior sampling to find situations where the robot controller exhibits user-specified behaviors, such as highly jerky motions. We use RoCUS to analyze three controller classes (deep learning models, rapidly exploring random trees and dynamical system formulations) on two domains (2D navigation and a 7 degree-of-freedom arm reaching), and uncover insights to further our understanding of these controllers and ultimately improve their designs.

Ankit Shah, Samir Wadhwania, Julie Shah

Defining sound and complete specifications for robots using formal languages is challenging, while learning formal specifications directly from demonstrations can lead to over-constrained task policies. In this paper, we propose a Bayesian interactive robot training framework that allows the robot to learn from both demonstrations provided by a teacher, and that teacher's assessments of the robot's task executions. We also present an active learning approach -- inspired by uncertainty sampling -- to identify the task execution with the most uncertain degree of acceptability. Through a simulated experiment, we demonstrate that our active learning approach identifies a teacher's intended task specification with an equivalent or greater similarity when compared to an approach that learns purely from demonstrations. Finally, we demonstrate the efficacy of our approach in a real-world setting through a user-study based on teaching a robot to set a dinner table.

Mycal Tucker, Yilun Zhou, Julie Shah

Robotic agents must adopt existing social conventions in order to be effective teammates. These social conventions, such as driving on the right or left side of the road, are arbitrary choices among optimal policies, but all agents on a successful team must use the same convention. Prior work has identified a method of combining self-play with paired input-output data gathered from existing agents in order to learn their social convention without interacting with them. We build upon this work by introducing a technique called Adversarial Self-Play (ASP) that uses adversarial training to shape the space of possible learned policies and substantially improves learning efficiency. ASP only requires the addition of unpaired data: a dataset of outputs produced by the social convention without associated inputs. Theoretical analysis reveals how ASP shapes the policy space and the circumstances (when behaviors are clustered or exhibit some other structure) under which it offers the greatest benefits. Empirical results across three domains confirm ASP's advantages: it produces models that more closely match the desired social convention when given as few as two paired datapoints.

Serena Booth, Ankit Shah, Yilun Zhou et al.

Though neural network models demonstrate impressive performance, we do not understand exactly how these black-box models make individual predictions. This drawback has led to substantial research devoted to understand these models in areas such as robustness, interpretability, and generalization ability. In this paper, we consider the problem of exploring the prediction level sets of a classifier using probabilistic programming. We define a prediction level set to be the set of examples for which the predictor has the same specified prediction confidence with respect to some arbitrary data distribution. Notably, our sampling-based method does not require the classifier to be differentiable, making it compatible with arbitrary classifiers. As a specific instantiation, if we take the classifier to be a neural network and the data distribution to be that of the training data, we can obtain examples that will result in specified predictions by the neural network. We demonstrate this technique with experiments on a synthetic dataset and MNIST. Such level sets in classification may facilitate human understanding of classification behaviors.

Ankit Shah, Shen Li, Julie Shah

Reward engineering is crucial to high performance in reinforcement learning systems. Prior research into reward design has largely focused on Markovian functions representing the reward. While there has been research into expressing non-Markov rewards as linear temporal logic (LTL) formulas, this has focused on task specifications directly defined by the user. However, in many real-world applications, task specifications are ambiguous, and can only be expressed as a belief over LTL formulas. In this paper, we introduce planning with uncertain specifications (PUnS), a novel formulation that addresses the challenge posed by non-Markovian specifications expressed as beliefs over LTL formulas. We present four criteria that capture the semantics of satisfying a belief over specifications for different applications, and analyze the qualitative implications of these criteria within a synthetic domain. We demonstrate the existence of an equivalent Markov decision process (MDP) for any instance of PUnS. Finally, we demonstrate our approach on the real-world task of setting a dinner table automatically with a robot that inferred task specifications from human demonstrations.

Ramya Ramakrishnan, Ece Kamar, Debadeepta Dey et al.

Agents trained in simulation may make errors in the real world due to mismatches between training and execution environments. These mistakes can be dangerous and difficult to discover because the agent cannot predict them a priori. We propose using oracle feedback to learn a predictive model of these blind spots to reduce costly errors in real-world applications. We focus on blind spots in reinforcement learning (RL) that occur due to incomplete state representation: The agent does not have the appropriate features to represent the true state of the world and thus cannot distinguish among numerous states. We formalize the problem of discovering blind spots in RL as a noisy supervised learning problem with class imbalance. We learn models to predict blind spots in unseen regions of the state space by combining techniques for label aggregation, calibration, and supervised learning. The models take into consideration noise emerging from different forms of oracle feedback, including demonstrations and corrections. We evaluate our approach on two domains and show that it achieves higher predictive performance than baseline methods, and that the learned model can be used to selectively query an oracle at execution time to prevent errors. We also empirically analyze the biases of various feedback types and how they influence the discovery of blind spots.

Matthew Gombolay, Reed Jensen, Jessica Stigile et al.

Coordinating agents to complete a set of tasks with intercoupled temporal and resource constraints is computationally challenging, yet human domain experts can solve these difficult scheduling problems using paradigms learned through years of apprenticeship. A process for manually codifying this domain knowledge within a computational framework is necessary to scale beyond the ``single-expert, single-trainee" apprenticeship model. However, human domain experts often have difficulty describing their decision-making processes, causing the codification of this knowledge to become laborious. We propose a new approach for capturing domain-expert heuristics through a pairwise ranking formulation. Our approach is model-free and does not require enumerating or iterating through a large state space. We empirically demonstrate that this approach accurately learns multifaceted heuristics on a synthetic data set incorporating job-shop scheduling and vehicle routing problems, as well as on two real-world data sets consisting of demonstrations of experts solving a weapon-to-target assignment problem and a hospital resource allocation problem. We also demonstrate that policies learned from human scheduling demonstration via apprenticeship learning can substantially improve the efficiency of a branch-and-bound search for an optimal schedule. We employ this human-machine collaborative optimization technique on a variant of the weapon-to-target assignment problem. We demonstrate that this technique generates solutions substantially superior to those produced by human domain experts at a rate up to 9.5 times faster than an optimization approach and can be applied to optimally solve problems twice as complex as those solved by a human demonstrator.

Stefanos Nikolaidis, Jodi Forlizzi, David Hsu et al.

A robot operating in isolation needs to reason over the uncertainty in its model of the world and adapt its own actions to account for this uncertainty. Similarly, a robot interacting with people needs to reason over its uncertainty over the human internal state, as well as over how this state may change, as humans adapt to the robot. This paper summarizes our own work in this area, which depicts the different ways that probabilistic planning and game-theoretic algorithms can enable such reasoning in robotic systems that collaborate with people. We start with a general formulation of the problem as a two-player game with incomplete information. We then articulate the different assumptions within this general formulation, and we explain how these lead to exciting and diverse robot behaviors in real-time interactions with actual human subjects, in a variety of manufacturing, personal robotics and assistive care settings.

Pieter Abbeel, Ken Goldberg, Gregory Hager et al.

An Autonomous Physical System (APS) will be expected to reliably and independently evaluate, execute, and achieve goals while respecting surrounding rules, laws, or conventions. In doing so, an APS must rely on a broad spectrum of dynamic, complex, and often imprecise information about its surroundings, the task it is to perform, and its own sensors and actuators. For example, cleaning in a home or commercial setting requires the ability to perceive, grasp, and manipulate many physical objects, the ability to reliably perform a variety of subtasks such as washing, folding, and stacking, and knowledge about local conventions such as how objects are classified and where they should be stored. The information required for reliable autonomous operation may come from external sources and from the robot's own sensor observations or in the form of direct instruction by a trainer. Similar considerations apply across many domains - construction, manufacturing, in-home assistance, and healthcare. For example, surgeons spend many years learning about physiology and anatomy before they touch a patient. They then perform roughly 1000 surgeries under the tutelage of an expert surgeon, and they practice basic maneuvers such as suture tying thousands of times outside the operating room. All of these elements come together to achieve expertise at this task. Endowing a system with robust autonomy by traditional programming methods has thus far had limited success. Several promising new paths to acquiring and processing such data are emerging. This white paper outlines three promising research directions for enabling an APS to learn the physical and information skills necessary to perform tasks with independence and flexibility: Deep Reinforcement Learning, Human-Robot Interaction, and Cloud Robotics.

Been Kim, Cynthia Rudin, Julie Shah

We present the Bayesian Case Model (BCM), a general framework for Bayesian case-based reasoning (CBR) and prototype classification and clustering. BCM brings the intuitive power of CBR to a Bayesian generative framework. The BCM learns prototypes, the "quintessential" observations that best represent clusters in a dataset, by performing joint inference on cluster labels, prototypes and important features. Simultaneously, BCM pursues sparsity by learning subspaces, the sets of features that play important roles in the characterization of the prototypes. The prototype and subspace representation provides quantitative benefits in interpretability while preserving classification accuracy. Human subject experiments verify statistically significant improvements to participants' understanding when using explanations produced by BCM, compared to those given by prior art.

Stefanos Nikolaidis, Keren Gu, Ramya Ramakrishnan et al.

We present a framework for learning human user models from joint-action demonstrations that enables the robot to compute a robust policy for a collaborative task with a human. The learning takes place completely automatically, without any human intervention. First, we describe the clustering of demonstrated action sequences into different human types using an unsupervised learning algorithm. These demonstrated sequences are also used by the robot to learn a reward function that is representative for each type, through the employment of an inverse reinforcement learning algorithm. The learned model is then used as part of a Mixed Observability Markov Decision Process formulation, wherein the human type is a partially observable variable. With this framework, we can infer, either offline or online, the human type of a new user that was not included in the training set, and can compute a policy for the robot that will be aligned to the preference of this new user and will be robust to deviations of the human actions from prior demonstrations. Finally we validate the approach using data collected in human subject experiments, and conduct proof-of-concept demonstrations in which a person performs a collaborative task with a small industrial robot.

Been Kim, Caleb M. Chacha, Julie Shah

We aim to reduce the burden of programming and deploying autonomous systems to work in concert with people in time-critical domains, such as military field operations and disaster response. Deployment plans for these operations are frequently negotiated on-the-fly by teams of human planners. A human operator then translates the agreed upon plan into machine instructions for the robots. We present an algorithm that reduces this translation burden by inferring the final plan from a processed form of the human team's planning conversation. Our approach combines probabilistic generative modeling with logical plan validation used to compute a highly structured prior over possible plans. This hybrid approach enables us to overcome the challenge of performing inference over the large solution space with only a small amount of noisy data from the team planning session. We validate the algorithm through human subject experimentation and show we are able to infer a human team's final plan with 83% accuracy on average. We also describe a robot demonstration in which two people plan and execute a first-response collaborative task with a PR2 robot. To the best of our knowledge, this is the first work that integrates a logical planning technique within a generative model to perform plan inference.