We survey 146 papers analyzing "bias" in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing "bias" is an inherently normative process. We further find that these papers' proposed quantitative techniques for measuring or mitigating "bias" are poorly matched to their motivations and do not engage with the relevant literature outside of NLP. Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing "bias" in NLP systems. These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of "bias"---i.e., what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements---and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities.

Language (Technology) is Power: A Critical Survey of "Bias" in NLP

We survey 146 papers analyzing "bias" in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing "bias" is an inherently normative process We further find that these papers' proposed quantitative techniques for measuring or mitigating "bias" are poorly matched to their motivations and do not engage with the relevant literature outside of NLP Based on these findings, we describe the beginnings of a path forward by proposing three recommendations that should guide work analyzing "bias" in NLP systems These recommendations rest on a greater recognition of the relationships between language and social hierarchies, encouraging researchers and practitioners to articulate their conceptualizations of "bias"---ie, what kinds of system behaviors are harmful, in what ways, to whom, and why, as well as the normative reasoning underlying these statements---and to center work around the lived experiences of members of communities affected by NLP systems, while interrogating and reimagining the power relations between technologists and such communities

/pdf/language-technology-is-power-a-critical-survey-of-bias-in-40xzysfmb0.pdf

Treatment of patients with chronic low back pain and its evolving disability primarily tries to improve the patients’ levels of activities and participation. Mostly, self-reported questionnaires have been used for clinical as well as research purposes to assess daily functioning (1,2), of which the most commonly used will be discussed below. However, this information may not necessarily reflect the real capacity of a patient’s performance. A recent review showed that the correlation of self-reported disability and physical activity level was at best moderate for patients with chronic low back pain (3). In order to improve objectivity, measures of body function, e.g., spinal mobility and lumbar extensor muscle strength, have been used, although the correlation with the level of disability is very weak (4,5). Furthermore, there are major concerns about reliability and validity (6–8). Besides the self-reported disability measures, many have urged to use more objective and direct measures of low back pain–specific functional capacity (5,9,10). Capacity is defined as the highest probable level of functioning that a person may reach in an activity domain at any given moment in a standardized environment. Although there is still no consensus for the definition of functional capacity evaluation (FCE), in the past decades, several FCE measures have been developed, of which the Isernhagen Work Systems Functional Capacity Evaluation (IWS-FCE) is among the most frequently used (11,12). However, recently published psychometric data have shown that some of the tasks included in the IWS-FCE are not reliable (13,14). Unfortunately, the entire sequence of tasks in, for example, the IWS-FCE, is time consuming and expensive, as is the training of the test observer. Therefore, we have decided not to include these measures in this review. Nevertheless, in order to keep up with, for example, the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials recommendations to evaluate several core outcome domains, including physical functioning (9), we wanted to include easy to use performance tasks. Several tasks have been described (8,15–17), but most of them are not low back pain specific, and some, such as the Back Performance Scale, show insufficient factor structure, as in this measure the quality of the performance is also scored (1,18). Therefore, we decided only to include a performance task that assesses lifting, an activity that specifically might be hampered by low back pain. For the selection of the self-reported disability questionnaires and lifting performance tasks, we only selected questionnaires/tests that are low back pain specific and of which all psychometric, including responsiveness, properties have been studied in relevant low back pain populations and published in peer-reviewed journals. Other criteria for selection were: being available in at least English and for performance task measures, easy to administer, inexpensive, and not time consuming when used in clinical practice.

https://onlinelibrary.wiley.com/doi/pdf/10.1002/acr.20542

Measures of function in low back pain/disorders: Low Back Pain Rating Scale (LBPRS), Oswestry Disability Index (ODI), Progressive Isoinertial Lifting Evaluation (PILE), Quebec Back Pain Disability Scale (QBPDS), and Roland‐Morris Disability Questionnaire (RDQ)

This paper formulates and estimates a rigorously-justified linear probability model of binary choices over alternatives characterized by unobserved attributes. The model is applied to estimate preferences of congressmen as expressed in their votes on bills. The effective dimension of the attribute space characterizing votes is larger than what has been estimated in recent influential studies of congressional voting by Poole and Rosenthal. Congressmen vote on more than ideology. Issue-specific attributes are an important determinant of congressional" voting patterns. The estimated dimension is too large for the median voter model to describe congressional voting

Linear Probability Models of the Demand for Attributes with an Empirical Application to Estimating the Preferences of Legislators

Dropout methods are a family of stochastic techniques used in neural network training or inference that have generated significant research interest and are widely used in practice. They have been successfully applied in neural network regularization, model compression, and in measuring the uncertainty of neural network outputs. While original formulated for dense neural network layers, recent advances have made dropout methods also applicable to convolutional and recurrent neural network layers. This paper summarizes the history of dropout methods, their various applications, and current areas of research interest. Important proposed methods are described in additional detail.

Survey of Dropout Methods for Deep Neural Networks.

Recommender systems offer critical services in the age of mass information. A good recommender system selects a certain item for a specific user by recognizing why the user might like the item. This awareness implies that the system should model the background of the items and the users. This background modeling for recommendation is tackled through the various models of collaborative filtering with auxiliary information. This paper presents variational approaches for collaborative filtering to deal with auxiliary information. The proposed methods encompass variational autoencoders through augmenting structures to model the auxiliary information and to model the implicit user feedback. This augmentation includes the ladder network and the generative adversarial network to extract the low-dimensional representations influenced by the auxiliary information. These two augmentations are the first trial in the venue of the variational autoencoders, and we demonstrate their significant improvement on the performances in the applications of the collaborative filtering.

Augmented Variational Autoencoders for Collaborative Filtering with Auxiliary Information

The problem of fair classification can be mollified if we develop a method to remove the embedded sensitive information from the classification features. This line of separating the sensitive information is developed through the causal inference, and the causal inference enables the counterfactual generations to contrast the what-if case of the opposite sensitive attribute. Along with this separation with the causality, a frequent assumption in the deep latent causal model defines a single latent variable to absorb the entire exogenous uncertainty of the causal graph. However, we claim that such structure cannot distinguish the 1) information caused by the intervention (i.e., sensitive variable) and 2) information correlated with the intervention from the data. Therefore, this paper proposes Disentangled Causal Effect Variational Autoencoder (DCEVAE) to resolve this limitation by disentangling the exogenous uncertainty into two latent variables: either 1) independent to interventions or 2) correlated to interventions without causality. Particularly, our disentangling approach preserves the latent variable correlated to interventions in generating counterfactual examples. We show that our method estimates the total effect and the counterfactual effect without a complete causal graph. By adding a fairness regularization, DCEVAE generates a counterfactual fair dataset while losing less original information. Also, DCEVAE generates natural counterfactual images by only flipping sensitive information. Additionally, we theoretically show the differences in the covariance structures of DCEVAE and prior works from the perspective of the latent disentanglement.

Counterfactual Fairness with Disentangled Causal Effect Variational Autoencoder

A long user history inevitably reflects the transitions of personal interests over time. The analyses on the user history require the robust sequential model to anticipate the transitions and the decays of user interests. The user history is often modeled by various RNN structures, but the RNN structures in the recommendation system still suffer from the long-term dependency and the interest drifts. To resolve these challenges, we suggest HCRNN with three hierarchical contexts of the global, the local, and the temporary interests. This structure is designed to withhold the global long-term interest of users, to reflect the local sub-sequence interests, and to attend the temporary interests of each transition. Besides, we propose a hierarchical context-based gate structure to incorporate our interest drift assumption. As we suggest a new RNN structure, we support HCRNN with a complementary bi-channel attention structure to utilize hierarchical context. We experimented the suggested structure on the sequential recommendation tasks with CiteULike, MovieLens, and LastFM, and our model showed the best performances in the sequential recommendations.

/pdf/hierarchical-context-enabled-recurrent-neural-network-for-2t71gmxawj.pdf

Hierarchical Context enabled Recurrent Neural Network for Recommendation

A long user history inevitably reflects the transitions of personal interests over time. The analyses on the user history require the robust sequential model to anticipate the transitions and the decays of user interests. The user history is often modeled by various RNN structures, but the RNN structures in the recommendation system still suffer from the long-term dependency and the interest drifts. To resolve these challenges, we suggest HCRNN with three hierarchical contexts of the global, the local, and the temporary interests. This structure is designed to withhold the global long-term interest of users, to reflect the local sub-sequence interests, and to attend the temporary interests of each transition. Besides, we propose a hierarchical context-based gate structure to incorporate our \textit{interest drift assumption}. As we suggest a new RNN structure, we support HCRNN with a complementary \textit{bi-channel attention} structure to utilize hierarchical context. We experimented the suggested structure on the sequential recommendation tasks with CiteULike, MovieLens, and LastFM, and our model showed the best performances in the sequential recommendations.

Recent studies identified that sequential Recommendation is improved by the attention mechanism. By following this development, we propose Relation-Aware Kernelized Self-Attention (RKSA) adopting a self-attention mechanism of the Transformer with augmentation of a probabilistic model. The original self-attention of Transformer is a deterministic measure without relation-awareness. Therefore, we introduce a latent space to the self-attention, and the latent space models the recommendation context from relation as a multivariate skew-normal distribution with a kernelized covariance matrix from co-occurrences, item characteristics, and user information. This work merges the self-attention of the Transformer and the sequential recommendation by adding a probabilistic model of the recommendation task specifics. We experimented RKSA over the benchmark datasets, and RKSA shows significant improvements compared to the recent baseline models. Also, RKSA were able to produce a latent space model that answers the reasons for recommendation.

Kyungwoo Song

Papers

Augmented Variational Autoencoders for Collaborative Filtering with Auxiliary Information

Counterfactual Fairness with Disentangled Causal Effect Variational Autoencoder

Hierarchical Context enabled Recurrent Neural Network for Recommendation

Hierarchical Context enabled Recurrent Neural Network for Recommendation

Sequential Recommendation with Relation-Aware Kernelized Self-Attention