Search

Machine learning models are often used to automate or support decisions in applications such as lending and hiring. In such settings, consumer protection rules mandate that we provide a list of "principal reasons" to consumers who receive adverse decisions. In practice, lenders and employers identify principal reasons by returning the top-scoring features from a feature attribution method. In this work, we study how such practices align with one of the underlying goals of consumer protection - recourse - i.e., educating individuals on how they can attain a desired outcome. We show that standard attribution methods can mislead individuals by highlighting reasons without recourse - i.e., by presenting consumers with features that cannot be changed to achieve recourse. We propose to address these issues by scoring features on the basis of responsiveness - i.e., the probability that an individual can attain a desired outcome by changing a specific feature. We develop efficient methods to compute responsiveness scores for any model and any dataset under complex actionability constraints. We present an extensive empirical study on the responsiveness of explanations in lending and demonstrate how responsiveness scores can be used to construct feature-highlighting explanations that lead to recourse and mitigate harm by flagging instances with fixed predictions., Comment: 11 pages, 3 figures in main body

Automated content moderation has long been used to help identify and filter undesired user-generated content online. Generative AI systems now use such filters to keep undesired generated content from being created by or shown to users. From classrooms to Hollywood, as generative AI is increasingly used for creative or expressive text generation, whose stories will these technologies allow to be told, and whose will they suppress? In this paper, we define and introduce measures of speech suppression, focusing on speech related to different identity groups incorrectly filtered by a range of content moderation APIs. Using both short-form, user-generated datasets traditional in content moderation and longer generative AI-focused data, including two datasets we introduce in this work, we create a benchmark for measurement of speech suppression for nine identity groups. Across one traditional and four generative AI-focused automated content moderation services tested, we find that identity-related speech is more likely to be incorrectly suppressed than other speech except in the cases of a few non-marginalized groups. Additionally, we find differences between APIs in their abilities to correctly moderate generative AI content.

When information spreads across a network via pairwise sharing, large disparities in information access can arise from the network's structural heterogeneity. Algorithms to improve the fairness of information access seek to maximize the minimum access of a node to information by sequentially selecting new nodes to seed with the spreading information. However, existing algorithms are computationally expensive. Here, we develop and evaluate a set of 10 new scalable algorithms to improve information access in social networks; in order to compare them to the existing state-of-the-art, we introduce both a new performance metric and a new benchmark corpus of networks. Additionally, we investigate the degree to which algorithm performance on minimizing information access gaps can be predicted ahead of time from features of a network's structure. We find that while no algorithm is strictly superior to all others across networks, our new scalable algorithms are competitive with the state-of-the-art and orders of magnitude faster. We introduce a meta-learner approach that learns which of the fast algorithms is best for a specific network and is on average only 20% less effective than the state-of-the-art performance on held-out data, while about 75-130 times faster. Furthermore, on about 20% of networks the meta-learner's performance exceeds the state-of-the-art., Comment: 9 pages, 9 figures, and 3 appendices containing 12 algorithms, 1 table, and 9 additional figures

Despite the popularity of the `pre-train then fine-tune' paradigm in the NLP community, existing work quantifying energy costs and associated carbon emissions has largely focused on language model pre-training. Although a single pre-training run draws substantially more energy than fine-tuning, fine-tuning is performed more frequently by many more individual actors, and thus must be accounted for when considering the energy and carbon footprint of NLP. In order to better characterize the role of fine-tuning in the landscape of energy and carbon emissions in NLP, we perform a careful empirical study of the computational costs of fine-tuning across tasks, datasets, hardware infrastructure and measurement modalities. Our experimental results allow us to place fine-tuning energy and carbon costs into perspective with respect to pre-training and inference, and outline recommendations to NLP researchers and practitioners who wish to improve their fine-tuning energy efficiency., Comment: EMNLP 2023 Findings; First two authors contributed equally; 12 pages

In social networks, a node's position is a form of \it{social capital}. Better-positioned members not only benefit from (faster) access to diverse information, but innately have more potential influence on information spread. Structural biases often arise from network formation, and can lead to significant disparities in information access based on position. Further, processes such as link recommendation can exacerbate this inequality by relying on network structure to augment connectivity. We argue that one can understand and quantify this social capital through the lens of information flow in the network. We consider the setting where all nodes may be sources of distinct information, and a node's (dis)advantage deems its ability to access all information available on the network. We introduce three new measures of advantage (broadcast, influence, and control), which are quantified in terms of position in the network using \it{access signatures} -- vectors that represent a node's ability to share information. We then consider the problem of improving equity by making interventions to increase the access of the least-advantaged nodes. We argue that edge augmentation is most appropriate for mitigating bias in the network structure, and frame a budgeted intervention problem for maximizing minimum pairwise access. Finally, we propose heuristic strategies for selecting edge augmentations and empirically evaluate their performance on a corpus of real-world social networks. We demonstrate that a small number of interventions significantly increase the broadcast measure of access for the least-advantaged nodes (over 5 times more than random), and also improve the minimum influence. Additional analysis shows that these interventions can also dramatically shrink the gap in advantage between nodes (over \%82) and reduce disparities between their access signatures.

Voter suppression and associated racial disparities in access to voting are long-standing civil rights concerns in the United States. Barriers to voting have taken many forms over the decades. A history of violent explicit discouragement has shifted to more subtle access limitations that can include long lines and wait times, long travel times to reach a polling station, and other logistical barriers to voting. Our focus in this work is on quantifying disparities in voting access pertaining to the overall time-to-vote, and how they could be remedied via a better choice of polling location or provisioning more sites where voters can cast ballots. However, appropriately calibrating access disparities is difficult because of the need to account for factors such as population density and different community expectations for reasonable travel times. In this paper, we quantify access to polling locations, developing a methodology for the calibrated measurement of racial disparities in polling location "load" and distance to polling locations. We apply this methodology to a study of real-world data from Florida and North Carolina to identify disparities in voting access from the 2020 election. We also introduce algorithms, with modifications to handle scale, that can reduce these disparities by suggesting new polling locations from a given list of identified public locations (including schools and libraries). Applying these algorithms on the 2020 election location data also helps to expose and explore tradeoffs between the cost of allocating more polling locations and the potential impact on access disparities. The developed voting access measurement methodology and algorithmic remediation technique is a first step in better polling location assignment.

Social network position confers power and social capital. In the setting of online social networks that have massive reach, creating mathematical representations of social capital is an important step towards understanding how network position can differentially confer advantage to different groups and how network position can itself be a source of advantage. In this paper, we use well established models for information flow on networks as a base to propose a formal descriptor of the network position of a node as represented by its information access. Combining these descriptors allows a full representation of social capital across the network. Using real-world networks, we demonstrate that this representation allows the identification of differences between groups based on network specific measures of inequality of access.

Game-theoretic formulations of feature importance have become popular as a way to "explain" machine learning models. These methods define a cooperative game between the features of a model and distribute influence among these input elements using some form of the game's unique Shapley values. Justification for these methods rests on two pillars: their desirable mathematical properties, and their applicability to specific motivations for explanations. We show that mathematical problems arise when Shapley values are used for feature importance and that the solutions to mitigate these necessarily induce further complexity, such as the need for causal reasoning. We also draw on additional literature to argue that Shapley values do not provide explanations which suit human-centric goals of explainability., Comment: Accepted to ICML 2020

The carbon footprint of algorithms must be measured and transparently reported so computer scientists can take an honest and active role in environmental sustainability. In this paper, we take analyses usually applied at the industrial level and make them accessible for individual computer science researchers with an easy-to-use Python package. Localizing to the energy mixture of the electrical power grid, we make the conversion from energy usage to CO2 emissions, in addition to contextualizing these results with more human-understandable benchmarks such as automobile miles driven. We also include comparisons with energy mixtures employed in electrical grids around the world. We propose including these automatically-generated Energy Usage Reports as part of standard algorithmic accountability practices, and demonstrate the use of these reports as part of model-choice in a machine learning context., Comment: Workshop on Tackling Climate Change with Machine Learning at NeurIPS 2019

Data sets for fairness relevant tasks can lack examples or be biased according to a specific label in a sensitive attribute. We demonstrate the usefulness of weight based meta-learning approaches in such situations. For models that can be trained through gradient descent, we demonstrate that there are some parameter configurations that allow models to be optimized from a few number of gradient steps and with minimal data which are both fair and accurate. To learn such weight sets, we adapt the popular MAML algorithm to Fair-MAML by the inclusion of a fairness regularization term. In practice, Fair-MAML allows practitioners to train fair machine learning models from only a few examples when data from related tasks is available. We empirically exhibit the value of this technique by comparing to relevant baselines., Comment: arXiv admin note: substantial text overlap with arXiv:1908.09092

Motivated by concerns surrounding the fairness effects of sharing and transferring fair machine learning tools, we propose two algorithms: Fairness Warnings and Fair-MAML. The first is a model-agnostic algorithm that provides interpretable boundary conditions for when a fairly trained model may not behave fairly on similar but slightly different tasks within a given domain. The second is a fair meta-learning approach to train models that can be quickly fine-tuned to specific tasks from only a few number of sample instances while balancing fairness and accuracy. We demonstrate experimentally the individual utility of each model using relevant baselines and provide the first experiment to our knowledge of K-shot fairness, i.e. training a fair model on a new task with only K data points. Then, we illustrate the usefulness of both algorithms as a combined method for training models from a few data points on new tasks while using Fairness Warnings as interpretable boundary conditions under which the newly trained model may not be fair.

Motivated by the need to audit complex and black box models, there has been extensive research on quantifying how data features influence model predictions. Feature influence can be direct (a direct influence on model outcomes) and indirect (model outcomes are influenced via proxy features). Feature influence can also be expressed in aggregate over the training or test data or locally with respect to a single point. Current research has typically focused on one of each of these dimensions. In this paper, we develop disentangled influence audits, a procedure to audit the indirect influence of features. Specifically, we show that disentangled representations provide a mechanism to identify proxy features in the dataset, while allowing an explicit computation of feature influence on either individual outcomes or aggregate-level outcomes. We show through both theory and experiments that disentangled influence audits can both detect proxy features and show, for each individual or in aggregate, which of these proxy features affects the classifier being audited the most. In this respect, our method is more powerful than existing methods for ascertaining feature influence.