Much of machine learning research focuses on predictive accuracy: given a task, create a machine learning model (or algorithm) that maximizes accuracy. In many settings, however, the final prediction or decision of a system is under the control of a human, who uses an algorithm's output along with their own personal expertise in order to produce a combined prediction. One ultimate goal of such collaborative systems is "complementarity": that is, to produce lower loss (equivalently, greater payoff or utility) than either the human or algorithm alone. However, experimental results have shown that even in carefully-designed systems, complementary performance can be elusive. Our work provides three key contributions. First, we provide a theoretical framework for modeling simple human-algorithm systems and demonstrate that multiple prior analyses can be expressed within it. Next, we use this model to prove conditions where complementarity is impossible, and give constructive examples of where complementarity is achievable. Finally, we discuss the implications of our findings, especially with respect to the fairness of a classifier. In sum, these results deepen our understanding of key factors influencing the combined performance of human-algorithm systems, giving insight into how algorithmic tools can best be designed for collaborative environments.
相關內容
Interactive machine learning (IML) is a field of research that explores how to leverage both human and computational abilities in decision making systems. IML represents a collaboration between multiple complementary human and machine intelligent systems working as a team, each with their own unique abilities and limitations. This teamwork might mean that both systems take actions at the same time, or in sequence. Two major open research questions in the field of IML are: "How should we design systems that can learn to make better decisions over time with human interaction?" and "How should we evaluate the design and deployment of such systems?" A lack of appropriate consideration for the humans involved can lead to problematic system behaviour, and issues of fairness, accountability, and transparency. Thus, our goal with this work is to present a human-centred guide to designing and evaluating IML systems while mitigating risks. This guide is intended to be used by machine learning practitioners who are responsible for the health, safety, and well-being of interacting humans. An obligation of responsibility for public interaction means acting with integrity, honesty, fairness, and abiding by applicable legal statutes. With these values and principles in mind, we as a machine learning research community can better achieve goals of augmenting human skills and abilities. This practical guide therefore aims to support many of the responsible decisions necessary throughout the iterative design, development, and dissemination of IML systems.
The reinforcement learning (RL) problem is rife with sources of non-stationarity, making it a notoriously difficult problem domain for the application of neural networks. We identify a mechanism by which non-stationary prediction targets can prevent learning progress in deep RL agents: \textit{capacity loss}, whereby networks trained on a sequence of target values lose their ability to quickly update their predictions over time. We demonstrate that capacity loss occurs in a range of RL agents and environments, and is particularly damaging to performance in sparse-reward tasks. We then present a simple regularizer, Initial Feature Regularization (InFeR), that mitigates this phenomenon by regressing a subspace of features towards its value at initialization, leading to significant performance improvements in sparse-reward environments such as Montezuma's Revenge. We conclude that preventing capacity loss is crucial to enable agents to maximally benefit from the learning signals they obtain throughout the entire training trajectory.
The assignment game forms a paradigmatic setting for studying the core -- its pristine structural properties yield an in-depth understanding of this quintessential solution concept within cooperative game theory. In turn, insights gained provide valuable guidance on profit-sharing in real-life situations. In this vein, we raise three basic questions and address them using the following broad idea. Consider the LP-relaxation of the problem of computing an optimal assignment. On the one hand, the worth of the assignment game is given by the optimal objective function value of this LP, and on the other, the classic Shapley-Shubik Theorem \cite{Shapley1971assignment} tells us that its core imputations are precisely optimal solutions to the dual of this LP. These two facts naturally raise the question of viewing core imputations through the lens of complementarity. In turn, this leads to a resolution of all our questions.
Despite recent progress in abstractive summarization, systems still suffer from faithfulness errors. While prior work has proposed models that improve faithfulness, it is unclear whether the improvement comes from an increased level of extractiveness of the model outputs as one naive way to improve faithfulness is to make summarization models more extractive. In this work, we present a framework for evaluating the effective faithfulness of summarization systems, by generating a faithfulnessabstractiveness trade-off curve that serves as a control at different operating points on the abstractiveness spectrum. We then show that the Maximum Likelihood Estimation (MLE) baseline as well as a recently proposed method for improving faithfulness, are both worse than the control at the same level of abstractiveness. Finally, we learn a selector to identify the most faithful and abstractive summary for a given document, and show that this system can attain higher faithfulness scores in human evaluations while being more abstractive than the baseline system on two datasets. Moreover, we show that our system is able to achieve a better faithfulness-abstractiveness trade-off than the control at the same level of abstractiveness.
Recent developments in Artificial Intelligence (AI) have fueled the emergence of human-AI collaboration, a setting where AI is a coequal partner. Especially in clinical decision-making, it has the potential to improve treatment quality by assisting overworked medical professionals. Even though research has started to investigate the utilization of AI for clinical decision-making, its potential benefits do not imply its adoption by medical professionals. While several studies have started to analyze adoption criteria from a technical perspective, research providing a human-centered perspective with a focus on AI's potential for becoming a coequal team member in the decision-making process remains limited. Therefore, in this work, we identify factors for the adoption of human-AI collaboration by conducting a series of semi-structured interviews with experts in the healthcare domain. We identify six relevant adoption factors and highlight existing tensions between them and effective human-AI collaboration.
Applications of Reinforcement Learning (RL), in which agents learn to make a sequence of decisions despite lacking complete information about the latent states of the controlled system, that is, they act under partial observability of the states, are ubiquitous. Partially observable RL can be notoriously difficult -- well-known information-theoretic results show that learning partially observable Markov decision processes (POMDPs) requires an exponential number of samples in the worst case. Yet, this does not rule out the existence of large subclasses of POMDPs over which learning is tractable. In this paper we identify such a subclass, which we call weakly revealing POMDPs. This family rules out the pathological instances of POMDPs where observations are uninformative to a degree that makes learning hard. We prove that for weakly revealing POMDPs, a simple algorithm combining optimism and Maximum Likelihood Estimation (MLE) is sufficient to guarantee polynomial sample complexity. To the best of our knowledge, this is the first provably sample-efficient result for learning from interactions in overcomplete POMDPs, where the number of latent states can be larger than the number of observations.
Artificial intelligence (AI) is gaining momentum, and its importance for the future of work in many areas, such as medicine and banking, is continuously rising. However, insights on the effective collaboration of humans and AI are still rare. Typically, AI supports humans in decision-making by addressing human limitations. However, it may also evoke human bias, especially in the form of automation bias as an over-reliance on AI advice. We aim to shed light on the potential to influence automation bias by explainable AI (XAI). In this pre-test, we derive a research model and describe our study design. Subsequentially, we conduct an online experiment with regard to hotel review classifications and discuss first results. We expect our research to contribute to the design and development of safe hybrid intelligence systems.
As machine learning algorithms become increasingly integrated in crucial decision-making scenarios, such as healthcare, recruitment, and risk assessment, there have been increasing concerns about the privacy and fairness of such systems. Federated learning has been viewed as a promising solution for collaboratively training of machine learning models among multiple parties while maintaining the privacy of their local data. However, federated learning also poses new challenges in mitigating the potential bias against certain populations (e.g., demographic groups), as this typically requires centralized access to the sensitive information (e.g., race, gender) of each data point. Motivated by the importance and challenges of group fairness in federated learning, in this work, we propose FairFed, a novel algorithm to enhance group fairness via a fairness-aware aggregation method, which aims to provide fair model performance across different sensitive groups (e.g., racial, gender groups) while maintaining high utility. This formulation can further provide more flexibility in the customized local debiasing strategies for each client. We build our FairFed algorithm around the secure aggregation protocol of federated learning. When running federated training on widely investigated fairness datasets, we demonstrate that our proposed method outperforms the state-of-the-art fair federated learning frameworks under a high heterogeneous sensitive attribute distribution. We also investigate the performance of FairFed on naturally distributed real-life data collected from different geographical locations or departments within an organization.
Gaussian process regression is increasingly applied for learning unknown dynamical systems. In particular, the implicit quantification of the uncertainty of the learned model makes it a promising approach for safety-critical applications. When using Gaussian process regression to learn unknown systems, a commonly considered approach consists of learning the residual dynamics after applying some generic discretization technique, which might however disregard properties of the underlying physical system. Variational integrators are a less common yet promising approach to discretization, as they retain physical properties of the underlying system, such as energy conservation and satisfaction of explicit kinematic constraints. In this work, we present a novel structure-preserving learning-based modelling approach that combines a variational integrator for the nominal dynamics of a mechanical system and learning residual dynamics with Gaussian process regression. We extend our approach to systems with known kinematic constraints and provide formal bounds on the prediction uncertainty. The simulative evaluation of the proposed method shows desirable energy conservation properties in accordance with general theoretical results and demonstrates exact constraint satisfaction for constrained dynamical systems.
Consider the problem of training robustly capable agents. One approach is to generate a diverse collection of agent polices. Training can then be viewed as a quality diversity (QD) optimization problem, where we search for a collection of performant policies that are diverse with respect to quantified behavior. Recent work shows that differentiable quality diversity (DQD) algorithms greatly accelerate QD optimization when exact gradients are available. However, agent policies typically assume that the environment is not differentiable. To apply DQD algorithms to training agent policies, we must approximate gradients for performance and behavior. We propose two variants of the current state-of-the-art DQD algorithm that compute gradients via approximation methods common in reinforcement learning (RL). We evaluate our approach on four simulated locomotion tasks. One variant achieves results comparable to the current state-of-the-art in combining QD and RL, while the other performs comparably in two locomotion tasks. These results provide insight into the limitations of current DQD algorithms in domains where gradients must be approximated. Source code is available at //github.com/icaros-usc/dqd-rl
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191