Search for: All records

Experiences discovering attempts to subvert the peer-review process. 

Reward is the driving force for reinforcement-learning agents. This paper is dedicated to understanding the expressivity of reward as a way to capture tasks that we would want an agent to perform. We frame this study around three new abstract notions of “task” that might be desirable: (1) a set of acceptable behaviors, (2) a partial ordering over behaviors, or (3) a partial ordering over trajectories. Our main results prove that while reward can express many of these tasks, there exist instances of each task type that no Markov reward function can capture. We then provide a set of polynomial-time algorithms that construct a Markov reward function that allows an agent to optimize tasks of each of these three types, and correctly determine when no such reward function exists. We conclude with an empirical study that corroborates and illustrates our theoretical findings. 

Modeling student knowledge is important for assessment design, adaptive testing, curriculum design, and pedagogical intervention. The assessment design community has primarily focused on continuous latent-skill models with strong conditional independence assumptions among knowledge items, while the prerequisite discovery community has developed many models that aim to exploit the interdependence of discrete knowledge items. This paper attempts to bridge the gap by asking, "When does modeling assessment item interdependence improve predictive accuracy?" A novel adaptive testing evaluation framework is introduced that is amenable to techniques from both communities, and an efficient algorithm, Directed Item-Dependence And Confidence Thresholds (DIDACT), is introduced and compared with an Item-Response-Theory based model on several real and synthetic datasets. Experiments suggest that assessments with closely related questions benefit significantly from modeling item interdependence. 

This paper addresses the problem of training a robot to carry out temporal tasks of arbitrary complexity via evaluative human feedback that can be inaccurate. A key idea explored in our work is a kind of curriculum learning—training the robot to master simple tasks and then building up to more complex tasks. We show how a training procedure, using knowledge of the formal task representation, can decompose and train any task efficiently in the size of its representation. We further provide a set of experiments that support the claim that non-expert human trainers can decompose tasks in a way that is consistent with our theoretical results, with more than half of participants successfully training all of our experimental missions. We compared our algorithm with existing approaches and our experimental results suggest that our method outperforms alternatives, especially when feedback contains mistakes. 

Mutually beneficial behavior in repeated games can be enforced via the threat of punishment, as enshrined in game theory’s well-known “folk theorem.” There is a cost, however, to a player for generating these disincentives. In this work, we seek to minimize this cost by computing a “Stackelberg punishment,” in which the player selects a behavior that sufficiently punishes the other player while maximizing its own score under the assumption that the other player will adopt a best response. This idea generalizes the concept of a Stackelberg equilibrium. Known efficient algorithms for computing a Stackelberg equilibrium can be adapted to efficiently produce a Stackelberg punishment. We demonstrate an application of this idea in an experiment involving a virtual autonomous vehicle and human participants. We find that a self-driving car with a Stackelberg punishment policy discourages human drivers from bullying in a driving scenario requiring social negotiation. 

Because machine learning would benefit from reduced data requirements, some prior work has proposed using humans not just to label data, but also to explain those labels. To characterize the evidence humans might want to provide, we conducted a user study and a data experiment. In the user study, 75 participants provided classification labels for 20 photos, justifying those labels with free-text explanations. Explanations frequently referenced concepts (objects and attributes) in the image, yet 26% of explanations invoked concepts not in the image. Boolean logic was common in implicit form, but was rarely explicit. In a follow-up experiment on the Visual Genome dataset, we found that some concepts could be partially defined through their relationship to frequently co-occurring concepts, rather than only through labeling. 

Trigger-action programming (TAP) is a programming model enabling users to connect services and devices by writing if-then rules. As such systems are deployed in increasingly complex scenarios, users must be able to identify programming bugs and reason about how to fix them. We first systematize the temporal paradigms through which TAP systems could express rules. We then identify ten classes of TAP programming bugs related to control flow, timing, and inaccurate user expectations. We report on a 153-participant online study where participants were assigned to a temporal paradigm and shown a series of pre-written TAP rules. Half of the rules exhibited bugs from our ten bug classes. For most of the bug classes, we found that the presence of a bug made it harder for participants to correctly predict the behavior of the rule. Our findings suggest directions for better supporting end-user programmers.