Search for: All records

Abstract AI assistance is readily available to humans in a variety of decision-making applications. In order to fully understand the efficacy of such joint decision-making, it is important to first understand the human’s reliance on AI. However, there is a disconnect between how joint decision-making is studied and how it is practiced in the real world. More often than not, researchers ask humans to provide independent decisions before they are shown AI assistance. This is done to make explicit the influence of AI assistance on the human’s decision. We develop a cognitive model that allows us to infer thelatentreliance strategy of humans on AI assistance without asking the human to make an independent decision. We validate the model’s predictions through two behavioral experiments. The first experiment follows aconcurrentparadigm where humans are shown AI assistance alongside the decision problem. The second experiment follows asequentialparadigm where humans provide an independent judgment on a decision problem before AI assistance is made available. The model’s predicted reliance strategies closely track the strategies employed by humans in the two experimental paradigms. Our model provides a principled way to infer reliance on AI-assistance and may be used to expand the scope of investigation on human-AI collaboration. 

With the proliferation of AI, there is a growing concern regarding individuals becoming overly reliant on AI, leading to a decrease in intrinsic skills and autonomy. Assistive AI frameworks, on the other hand, also have the potential to improve human learning and performance by providing personalized learning experiences and real-time feedback. To study these opposing viewpoints on the consequences of AI assistance, we conducted a behavioral experiment using a dynamic decision-making game to assess how AI assistance impacts user performance, skill transfer, and cognitive engagement in task execution. Participants were assigned to one of four conditions that featured AI assistance at different time-points during the task. Our results suggest that AI assistance can improve immediate task performance without inducing human skill degradation or carryover effects in human learning. This observation has important implications for AI assistive frameworks as it suggests that there are classes of tasks in which assistance can be provided without risking the autonomy of the user. We discuss the possible reasons for this set of effects and explore their implications for future research directives. 

Given a pre-trained classifier and multiple human experts, we investigate the task of online classification where model predictions are provided for free but querying humans incurs a cost. In this practical but under-explored setting, oracle ground truth is not available. Instead, the prediction target is defined as the consensus vote of all experts. Given that querying full consensus can be costly, we propose a general framework for online Bayesian consensus estimation, leveraging properties of the multivariate hypergeometric distribution. Based on this framework, we propose a family of methods that dynamically estimate expert consensus from partial feedback by producing a posterior over expert and model beliefs. Analyzing this posterior induces an interpretable trade-off between querying cost and classification performance. We demonstrate the efficacy of our framework against a variety of baselines on CIFAR-10H and ImageNet-16H, two large-scale crowdsourced datasets. 

The deployment of machine learning classifiers in high-stakes domains requires well-calibrated confidence scores for model predictions. In this paper we introduce the notion of variable-based calibration to characterize calibration properties of a model with respect to a variable of interest, generalizing traditional score-based metrics such as expected calibration error (ECE). In particular, we find that models with near-perfect ECE can exhibit significant miscalibration as a function of features of the data. We demonstrate this phenomenon both theoretically and in practice on multiple well-known datasets, and show that it can persist after the application of existing calibration methods. To mitigate this issue, we propose strategies for detection, visualization, and quantification of variable-based calibration error. We then examine the limitations of current score-based calibration methods and explore potential modifications. Finally, we discuss the implications of these findings, emphasizing that an understanding of calibration beyond simple aggregate measures is crucial for endeavors such as fairness and model interpretability. 

Improving our understanding of how humans perceive AI teammates is an important foundation for our general understanding of human-AI teams. Extending relevant work from cognitive science, we propose a framework based on item response theory for modeling these perceptions. We apply this framework to real-world experiments, in which each participant works alongside another person or an AI agent in a question-answering setting, repeatedly assessing their teammate’s performance. Using this experimental data, we demonstrate the use of our framework for testing research questions about people’s perceptions of both AI agents and other people. We contrast mental models of AI teammates with those of human teammates as we characterize the dimensionality of these mental models, their development over time, and the influence of the participants’ own self-perception. Our results indicate that people expect AI agents’ performance to be significantly better on average than the performance of other humans, with less variation across different types of problems. We conclude with a discussion of the implications of these findings for human-AI interaction. 

We expose the statistical foundations of deep learning with the goal of facilitating conversation between the deep learning and statistics communities. We highlight core themes at the intersection; summarize key neural models, such as feedforward neural networks, sequential neural networks, and neural latent variable models; and link these ideas to their roots in probability and statistics. We also highlight research directions in deep learning where there are opportunities for statistical contributions. 

Artificial intelligence (AI) has the potential to improve human decision-making by providing decision recommendations and problem-relevant information to assist human decision-makers. However, the full realization of the potential of human–AI collaboration continues to face several challenges. First, the conditions that support complementarity (i.e., situations in which the performance of a human with AI assistance exceeds the performance of an unassisted human or the AI in isolation) must be understood. This task requires humans to be able to recognize situations in which the AI should be leveraged and to develop new AI systems that can learn to complement the human decision-maker. Second, human mental models of the AI, which contain both expectations of the AI and reliance strategies, must be accurately assessed. Third, the effects of different design choices for human-AI interaction must be understood, including both the timing of AI assistance and the amount of model information that should be presented to the human decision-maker to avoid cognitive overload and ineffective reliance strategies. In response to each of these three challenges, we present an interdisciplinary perspective based on recent empirical and theoretical findings and discuss new research directions. 

Despite recent advances in algorithmic fairness, methodologies for achieving fairness with generalized linear models (GLMs) have yet to be explored in general, despite GLMs being widely used in practice. In this paper we introduce two fairness criteria for GLMs based on equalizing expected outcomes or log-likelihoods. We prove that for GLMs both criteria can be achieved via a convex penalty term based solely on the linear components of the GLM, thus permitting efficient optimization. We also derive theoretical properties for the resulting fair GLM estimator. To empirically demonstrate the efficacy of the proposed fair GLM, we compare it with other well-known fair prediction methods on an extensive set of benchmark datasets for binary classification and regression. In addition, we demonstrate that the fair GLM can generate fair predictions for a range of response variables, other than binary and continuous outcomes.