The bi-encoder design of dense passage retriever (DPR) is a key factor to its success in open-domain question answering (QA), yet it is unclear how DPR’s question encoder and passage encoder individually contributes to overall performance, which we refer to as the encoder attribution problem. The problem is important as it helps us identify the factors that affect individual encoders to further improve overall performance. In this paper, we formulate our analysis under a probabilistic framework called encoder marginalization, where we quantify the contribution of a single encoder by marginalizing other variables. First, we find that the passage encoder contributes more than the question encoder to in-domain retrieval accuracy. Second, we demonstrate how to find the affecting factors for each encoder, where we train DPR with different amounts of data and use encoder marginalization to analyze the results. We find that positive passage overlap and corpus coverage of training data have big impacts on the passage encoder, while the question encoder is mainly affected by training sample complexity under this setting. Based on this framework, we can devise data-efficient training regimes: for example, we manage to train a passage encoder on SQuAD using 60% less training data without loss of accuracy.

The widespread use of Artificial Intelligence (AI) in consequential domains, such as health-care and parole decision-making systems, has drawn intense scrutiny on the fairness of these methods. However, ensuring fairness is often insufficient as the rationale for a contentious decision needs to be audited, understood, and defended. We propose that the attention mechanism can be used to ensure fair outcomes while simultaneously providing feature attributions to account for how a decision was made. Toward this goal, we design an attention-based model that can be leveraged as an attribution framework. It can identify features responsible for both performance and fairness of the model through attention interventions and attention weight manipulation. Using this attribution framework, we then design a post-processing bias mitigation strategy and compare it with a suite of baselines. We demonstrate the versatility of our approach by conducting experiments on two distinct data types, tabular and textual.

In an effort to guarantee that machine learning model outputs conform with human moral values, recent work has begun exploring the possibility of explicitly training models to learn the difference between right and wrong. This is typically done in a bottom-up fashion, by exposing the model to different scenarios, annotated with human moral judgements. One question, however, is whether the trained models actually learn any consistent, higher-level ethical principles from these datasets – and if so, what? Here, we probe the Allen AI Delphi model with a set of standardized morality questionnaires, and find that, despite some inconsistencies, Delphi tends to mirror the moral principles associated with the demographic groups involved in the annotation process. We question whether this is desirable and discuss how we might move forward with this knowledge.

Artificial Intelligence (AI) and Machine Learning (ML) are rapidly becoming must-have capabilities. According to a 2019 Forbes Insights Report, “seventy-nine percent [of executives] agree that AI is already having a transformational impact on workflows and tools for knowledge workers, but only 5% of executives consider their companies to be industry-leading in terms of taking advantage of AI-powered processes.” (Forbes 2019) A major reason for this may be a shortage of on-staff expertise in AI/ML. This paper explores the intertwined issues of trust, adoption, training, and ethics of outsourcing AI development to a third party. We describe our experiences as a provider of outsourced natural language processing (NLP). We discuss how trust and accountability co-evolve as solutions mature from proof-of-concept to production-ready.

Large-scale surveys are a widely used instrument to collect data from a target audience. Beyond the single individual, an appropriate analysis of the answers can reveal trends and patterns and thus generate new insights and knowledge for researchers. Current analysis practices employ shallow machine learning methods or rely on (biased) human judgment. This work investigates the usage of state-of-the-art NLP models such as BERT to automatically extract information from both open- and closed-ended questions. We also leverage explainability methods at different levels of granularity to further derive knowledge from the analysis model. Experiments on EMS—a survey-based study researching influencing factors affecting a student’s career goals—show that the proposed approach can identify such factors both at the input- and higher concept-level.

Neural rationale models are popular for interpretable predictions of NLP tasks. In these, a selector extracts segments of the input text, called rationales, and passes these segments to a classifier for prediction. Since the rationale is the only information accessible to the classifier, it is plausibly defined as the explanation. Is such a characterization unconditionally correct? In this paper, we argue to the contrary, with both philosophical perspectives and empirical evidence suggesting that rationale models are, perhaps, less rational and interpretable than expected. We call for more rigorous evaluations of these models to ensure desired properties of interpretability are indeed achieved. The code for our experiments is at https://github.com/yimingz89/Neural-Rationale-Analysis.

In this paper, we conduct an empirical study on a bias measure, log-likelihood Masked Language Model (MLM) scoring, on a benchmark dataset. Previous work evaluates whether MLMs are biased or not for certain protected attributes (e.g., race) by comparing the log-likelihood scores of sentences that contain stereotypical characteristics with one category (e.g., black) versus another (e.g., white). We hypothesized that this approach might be too sensitive to the choice of contextual words than the meaning of the sentence. Therefore, we computed the same measure after paraphrasing the sentences with different words but with same meaning. Our results demonstrate that the log-likelihood scoring can be more sensitive to utterance of specific words than to meaning behind a given sentence. Our paper reveals a shortcoming of the current log-likelihood-based bias measures for MLMs and calls for new ways to improve the robustness of it

Motivations for methods in explainable artificial intelligence (XAI) often include detecting, quantifying and mitigating bias, and contributing to making machine learning models fairer. However, exactly how an XAI method can help in combating biases is often left unspecified. In this paper, we briefly review trends in explainability and fairness in NLP research, identify the current practices in which explainability methods are applied to detect and mitigate bias, and investigate the barriers preventing XAI methods from being used more widely in tackling fairness issues.