We study the presence of heteronormative biases and prejudice against interracial romantic relationships in large language models by performing controlled name-replacement experiments for the task of relationship prediction. We show that models are less likely to predict romantic relationships for (a) same-gender character pairs than different-gender pairs; and (b) intra/inter-racial character pairs involving Asian names as compared to Black, Hispanic, or White names. We examine the contextualized embeddings of first names and find that gender for Asian names is less discernible than non-Asian names. We discuss the social implications of our findings, underlining the need to prioritize the development of inclusive and equitable technology.

Recent work has highlighted the culturally-contingent nature of commonsense knowledge. We introduce AMAMMERε, a test set of 525 multiple-choice questions designed to evaluate the commonsense knowledge of English LLMs, relative to the cultural contexts of Ghana and the United States. To create AMAMMERε, we select a set of multiple-choice questions (MCQs) from existing commonsense datasets and rewrite them in a multi-stage process involving surveys of Ghanaian and U.S. participants. In three rounds of surveys, participants from both pools are solicited to (1) write correct and incorrect answer choices, (2) rate individual answer choices on a 5-point Likert scale, and (3) select the best answer choice from the newly-constructed MCQ items, in a final validation step. By engaging participants at multiple stages, our procedure ensures that participant perspectives are incorporated both in the creation and validation of test items, resulting in high levels of agreement within each pool. We evaluate several off-the-shelf English LLMs on AMAMMERε. Uniformly, models prefer answers choices that align with the preferences of U.S. annotators over Ghanaian annotators. Additionally, when test items specify a cultural context (Ghana or the U.S.), models exhibit some ability to adapt, but performance is consistently better in U.S. contexts than Ghanaian. As large resources are devoted to the advancement of English LLMs, our findings underscore the need for culturally adaptable models and evaluations to meet the needs of diverse English-speaking populations around the world.

Large language models have been shown to behave inconsistently in response to meaning-preserving paraphrastic inputs. At the same time, researchers evaluate the knowledge and reasoning abilities of these models with test evaluations that do not disaggregate the effect of paraphrastic variability on performance. We propose a metric, PC, for evaluating the paraphrastic consistency of natural language reasoning models based on the probability of a model achieving the same correctness on two paraphrases of the same problem. We mathematically connect this metric to the proportion of a model’s variance in correctness attributable to paraphrasing. To estimate PC, we collect ParaNlu, a dataset of 7,782 human-written and validated paraphrased reasoning problems constructed on top of existing benchmark datasets for defeasible and abductive natural language inference.1 Using ParaNlu, we measure the paraphrastic consistency of several model classes and show that consistency dramatically increases with pretraining but not fine-tuning. All models tested exhibited room for improvement in paraphrastic consistency.

For human-robot dialogue in a search-and-rescue scenario, a strong knowledge of the conditions and objects a robot will face is essential for effective interpretation of natural language instructions. In order to utilize the power of large language models without overwhelming the limited storage capacity of a robot, we propose PropBank-Powered Data Creation. PropBank-Powered Data Creation is an expert-in-the-loop data generation pipeline which creates training data for disaster-specific language models. We leverage semantic role labeling and Rich Event Ontology resources to efficiently develop seed sentences for fine-tuning a smaller, targeted model that could operate onboard a robot for disaster relief. We developed 32 sentence templates, which we used to make 2 seed datasets of 175 instructions for earthquake search and rescue and train derailment response. We further leverage our seed datasets as evaluation data to test our baseline fine-tuned models.

Recent work shows that large language models (LLMs) can answer multiple-choice questions using only the choices, but does this mean that MCQA leaderboard rankings of LLMs are largely influenced by abilities in choices-only settings? To answer this, we use a contrast set that probes if LLMs over-rely on choices-only shortcuts in MCQA. While previous works build contrast sets via expensive human annotations or model-generated data which can be biased, we employ graph mining to extract contrast sets from existing MCQA datasets. We use our method on UnifiedQA, a group of six commonsense reasoning datasets with high choices-only accuracy, to build an 820-question contrast set. After validating our contrast set, we test 12 LLMs, finding that these models do not exhibit reliance on choice-only shortcuts when given both the question and choices. Thus, despite the susceptibility of MCQA to high choices-only accuracy, we argue that LLMs are not obtaining high ranks on MCQA leaderboards solely due to their ability to exploit choices-only shortcuts.

Chain-of-thought (COT) prompting can help large language models (LLMs) reason toward correct answers, but its efficacy in reasoning toward incorrect answers is unexplored. This process of elimination (PoE), when used with COT, can enhance self-consistency, interpretability, and tasks such as medical diagnoses of exclusion. Thus, we propose PoE with COT, where LLMs must reason toward incorrect options on multiple-choice questions. We evaluate the ability of GPT-3.5, LLaMA-2, and Falcon to perform PoE with COT on a total of four commonsense and scientific reasoning datasets. We find that the strategy of PoE always underperforms the strategy of choosing the correct answer. The agreement of these strategies is also lower than the self-consistency of each strategy. To study these issues further, we conduct error analyses and give suggestions for future work.

Questions posed by information-seeking users often contain implicit false or potentially harmful assumptions. In a high-risk domain such as maternal and infant health, a question-answering system must recognize these pragmatic constraints and go beyond simply answering user questions, examining them in context to respond helpfully. To achieve this, we study assumptions and implications, or pragmatic inferences, made when mothers ask questions about pregnancy and infant care by collecting a dataset of 2,727 inferences from 500 questions across three diverse sources. We study how health experts naturally address these inferences when writing answers, and illustrate that informing existing QA pipelines with pragmatic inferences produces responses that are more complete, mitigating the propagation of harmful beliefs.

Multiple-choice question answering (MCQA) is often used to evaluate large language models (LLMs). To see if MCQA assesses LLMs as intended, we probe if LLMs can perform MCQA with choices-only prompts, where models must select the correct answer only from the choices. In three MCQA datasets and four LLMs, this prompt bests a majority baseline in 11/12 cases, with up to 0.33 accuracy gain. To help explain this behavior, we conduct an in-depth, black-box analysis on memorization, choice dynamics, and question inference. Our key findings are threefold. First, we find no evidence that the choices-only accuracy stems from memorization alone. Second, priors over individual choices do not fully explain choices-only accuracy, hinting that LLMs use the group dynamics of choices. Third, LLMs have some ability to infer a relevant question from choices, and surprisingly can sometimes even match the original question. We hope to motivate the use of stronger baselines in MCQA benchmarks, the design of robust MCQA datasets, and further efforts to explain LLM decision-making.

We examine whether large language models (LLMs) exhibit race- and gender-based name discrimination in hiring decisions, similar to classic findings in the social sciences (Bertrand and Mullainathan, 2004). We design a series of templatic prompts to LLMs to write an email to a named job applicant informing them of a hiring decision. By manipulating the applicant’s first name, we measure the effect of perceived race, ethnicity, and gender on the probability that the LLM generates an acceptance or rejection email. We find that the hiring decisions of LLMs in many settings are more likely to favor White applicants over Hispanic applicants. In aggregate, the groups with the highest and lowest acceptance rates respectively are masculine White names and masculine Hispanic names. However, the comparative acceptance rates by group vary under different templatic settings, suggesting that LLMs’ race- and gender-sensitivity may be idiosyncratic and prompt-sensitive.

Through the use of first name substitution experiments, prior research has demonstrated the tendency of social commonsense reasoning models to systematically exhibit social biases along the dimensions of race, ethnicity, and gender (An et al., 2023). Demographic attributes of first names, however, are strongly correlated with corpus frequency and tokenization length, which may influence model behavior independent of or in addition to demographic factors. In this paper, we conduct a new series of first name substitution experiments that measures the influence of these factors while controlling for the others. We find that demographic attributes of a name (race, ethnicity, and gender) and name tokenization length are both factors that systematically affect the behavior of social commonsense reasoning models.

It is common sense that one should prefer to eat a salad with a fork rather than with a chainsaw. However, for eating a bowl of rice, the choice between a fork and a pair of chopsticks is culturally relative. We introduce FORK, a small, manually-curated set of CommonsenseQA-style questions for probing cultural biases and assumptions present in commonsense reasoning systems, with a specific focus on food-related customs. We test several CommonsenseQA systems on FORK, and while we see high performance on questions about the US culture, the poor performance of these systems on questions about non-US cultures highlights systematic cultural assumptions aligned with US over non-US cultures.

Reviewing and comprehending key obligations, entitlements, and prohibitions in legal contracts can be a tedious task due to their length and domain-specificity. Furthermore, the key rights and duties requiring review vary for each contracting party. In this work, we propose a new task of party-specific extractive summarization for legal contracts to facilitate faster reviewing and improved comprehension of rights and duties. To facilitate this, we curate a dataset comprising of party-specific pairwise importance comparisons annotated by legal experts, covering ~293K sentence pairs that include obligations, entitlements, and prohibitions extracted from lease agreements. Using this dataset, we train a pairwise importance ranker and propose a pipeline-based extractive summarization system that generates a party-specific contract summary. We establish the need for incorporating domain-specific notions of importance during summarization by comparing our system against various baselines using both automatic and human evaluation methods.

A common limitation of diagnostic tests for detecting social biases in NLP models is that they may only detect stereotypic associations that are pre-specified by the designer of the test. Since enumerating all possible problematic associations is infeasible, it is likely these tests fail to detect biases that are present in a model but not pre-specified by the designer. To address this limitation, we propose SODAPOP (SOcial bias Discovery from Answers about PeOPle), an approach for automatic social bias discovery in social commonsense question-answering. The SODAPOP pipeline generates modified instances from the Social IQa dataset (Sap et al., 2019b) by (1) substituting names associated with different demographic groups, and (2) generating many distractor answers from a masked language model. By using a social commonsense model to score the generated distractors, we are able to uncover the model’s stereotypic associations between demographic groups and an open set of words. We also test SODAPOP on debiased models and show the limitations of multiple state-of-the-art debiasing algorithms.

Language models have shown great promise in common-sense related tasks. However, it remains unseen how they would perform in the context of physically situated human-robot interactions, particularly in disaster-relief sce- narios. In this paper, we develop a language model evaluation dataset with more than 800 cloze sentences, written to probe for the func- tion of over 200 objects. The sentences are divided into two tasks: an “easy” task where the language model has to choose between vo- cabulary with different functions (Task 1), and a “challenge” where it has to choose between vocabulary with the same function, yet only one vocabulary item is appropriate given real world constraints on functionality (Task 2). Dis- tilBERT performs with about 80% accuracy for both tasks. To investigate how annotator variability affected those results, we developed a follow-on experiment where we compared our original results with wrong answers chosen based on embedding vector distances. Those results showed increased precision across docu- ments but a 15% decrease in accuracy. We con- clude that language models do have a strong knowledge basis for object reasoning, but will require creative fine-tuning strategies in order to be successfully deployed.

NLP models trained on text have been shown to reproduce human stereotypes, which can magnify harms to marginalized groups when systems are deployed at scale. We adapt the Agency-Belief-Communion (ABC) stereotype model of Koch et al. (2016) from social psychology as a framework for the systematic study and discovery of stereotypic group-trait associations in language models (LMs). We introduce the sensitivity test (SeT) for measuring stereotypical associations from language models. To evaluate SeT and other measures using the ABC model, we collect group-trait judgments from U.S.-based subjects to compare with English LM stereotypes. Finally, we extend this framework to measure LM stereotyping of intersectional identities.

As using they/them as personal pronouns becomes increasingly common in English, it is important that coreference resolution systems work as well for individuals who use personal “they” as they do for those who use gendered personal pronouns. We introduce a new benchmark for coreference resolution systems which evaluates singular personal “they” recognition. Using these WinoNB schemas, we evaluate a number of publicly available coreference resolution systems and confirm their bias toward resolving “they” pronouns as plural.

When strong partial-input baselines reveal artifacts in crowdsourced NLI datasets, the performance of full-input models trained on such datasets is often dismissed as reliance on spurious correlations. We investigate whether state-of-the-art NLI models are capable of overriding default inferences made by a partial-input baseline. We introduce an evaluation set of 600 examples consisting of perturbed premises to examine a RoBERTa model’s sensitivity to edited contexts. Our results indicate that NLI models are still capable of learning to condition on context—a necessary component of inferential reasoning—despite being trained on artifact-ridden datasets.

Legal documents are typically long and written in legalese, which makes it particularly difficult for laypeople to understand their rights and duties. While natural language understanding technologies can be valuable in supporting such understanding in the legal domain, the limited availability of datasets annotated for deontic modalities in the legal domain, due to the cost of hiring experts and privacy issues, is a bottleneck. To this end, we introduce, LEXDEMOD, a corpus of English contracts annotatedwith deontic modality expressed with respect to a contracting party or agent along with the modal triggers. We benchmark this dataset on two tasks: (i) agent-specific multi-label deontic modality classification, and (ii) agent-specific deontic modality and trigger span detection using Transformer-based (Vaswani et al., 2017) language models. Transfer learning experiments show that the linguistic diversity of modal expressions in LEXDEMOD generalizes reasonably from lease to employment andrental agreements. A small case study indicates that a model trained on LEXDEMOD can detect red flags with high recall. We believe our work offers a new research direction for deontic modality detection in the legal domain.

Script Knowledge (Schank and Abelson, 1975) has long been recognized as crucial for language understanding as it can help in filling in unstated information in a narrative. However, such knowledge is expensive to produce manually and difficult to induce from text due to reporting bias (Gordon and Van Durme, 2013). In this work, we are interested in the scientific question of whether explicit script knowledge is present and accessible through pre-trained generative language models (LMs). To this end, we introduce the task of generating full event sequence descriptions (ESDs) given a scenario as a natural language prompt. Through zero-shot probing, we find that generative LMs produce poor ESDs with mostly omitted, irrelevant, repeated or misordered events. To address this, we propose a pipeline-based script induction framework (SIF) which can generate good quality ESDs for unseen scenarios (e.g., bake a cake). SIF is a two-staged framework that fine-tunes LM on a small set of ESD examples in the first stage. In the second stage, ESD generated for an unseen scenario is post-processed using RoBERTa-based models to filter irrelevant events, remove repetitions, and reorder the temporally misordered events. Through automatic and manual evaluations, we demonstrate that SIF yields substantial improvements (1-3 BLEU points) over a fine-tuned LM. However, manual analysis shows that there is great room for improvement, offering a new research direction for inducing script knowledge.

Event schemas are structured knowledge sources defining typical real-world scenarios (e.g., going to an airport). We present a framework for efficient human-in-the-loop construction of a schema library, based on a novel script induction system and a well-crafted interface that allows non-experts to “program” complex event structures. Associated with this work we release a schema library: a machine readable resource of 232 detailed event schemas, each of which describe a distinct typical scenario in terms of its relevant sub-event structure (what happens in the scenario), participants (who plays a role in the scenario), fine-grained typing of each participant, and the implied relational constraints between them. We make our schema library and the SchemaBlocks interface available online.

Crowdworker-constructed natural language inference (NLI) datasets have been found to contain statistical artifacts associated with the annotation process that allow hypothesis-only classifiers to achieve better-than-random performance (CITATION). We investigate whether MedNLI, a physician-annotated dataset with premises extracted from clinical notes, contains such artifacts (CITATION). We find that entailed hypotheses contain generic versions of specific concepts in the premise, as well as modifiers related to responsiveness, duration, and probability. Neutral hypotheses feature conditions and behaviors that co-occur with, or cause, the condition(s) in the premise. Contradiction hypotheses feature explicit negation of the premise and implicit negation via assertion of good health. Adversarial filtering demonstrates that performance degrades when evaluated on the difficult subset. We provide partition information and recommendations for alternative dataset construction strategies for knowledge-intensive domains.

We present the Universal Decompositional Semantics (UDS) dataset (v1.0), which is bundled with the Decomp toolkit (v0.1). UDS1.0 unifies five high-quality, decompositional semantics-aligned annotation sets within a single semantic graph specification—with graph structures defined by the predicative patterns produced by the PredPatt tool and real-valued node and edge attributes constructed using sophisticated normalization procedures. The Decomp toolkit provides a suite of Python 3 tools for querying UDS graphs using SPARQL. Both UDS1.0 and Decomp0.1 are publicly available at http://decomp.io.

Defeasible inference is a mode of reasoning in which an inference (X is a bird, therefore X flies) may be weakened or overturned in light of new evidence (X is a penguin). Though long recognized in classical AI and philosophy, defeasible inference has not been extensively studied in the context of contemporary data-driven research on natural language inference and commonsense reasoning. We introduce Defeasible NLI (abbreviated 𝛿-NLI), a dataset for defeasible inference in natural language. Defeasible NLI contains extensions to three existing inference datasets covering diverse modes of reasoning: common sense, natural language inference, and social norms. From Defeasible NLI, we develop both a classification and generation task for defeasible inference, and demonstrate that the generation task is much more challenging. Despite lagging human performance, however, generative models trained on this data are capable of writing sentences that weaken or strengthen a specified inference up to 68% of the time.

Pre-trained language models (LMs) may perpetuate biases originating in their training corpus to downstream models. We focus on artifacts associated with the representation of given names (e.g., Donald), which, depending on the corpus, may be associated with specific entities, as indicated by next token prediction (e.g., Trump). While helpful in some contexts, grounding happens also in under-specified or inappropriate contexts. For example, endings generated for ‘Donald is a’ substantially differ from those of other names, and often have more-than-average negative sentiment. We demonstrate the potential effect on downstream tasks with reading comprehension probes where name perturbation changes the model answers. As a silver lining, our experiments suggest that additional pre-training on different corpora may mitigate this bias.

When does a sequence of events define an everyday scenario and how can this knowledge be induced from text? Prior works in inducing such scripts have relied on, in one form or another, measures of correlation between instances of events in a corpus. We argue from both a conceptual and practical sense that a purely correlation-based approach is insufficient, and instead propose an approach to script induction based on the causal effect between events, formally defined via interventions. Through both human and automatic evaluations, we show that the output of our method based on causal effects better matches the intuition of what a script represents.

The Word Embedding Association Test shows that GloVe and word2vec word embeddings exhibit human-like implicit biases based on gender, race, and other social constructs (Caliskan et al., 2017). Meanwhile, research on learning reusable text representations has begun to explore sentence-level texts, with some sentence encoders seeing enthusiastic adoption. Accordingly, we extend the Word Embedding Association Test to measure bias in sentence encoders. We then test several sentence encoders, including state-of-the-art methods such as ELMo and BERT, for the social biases studied in prior work and two important biases that are difficult or impossible to test at the word level. We observe mixed results including suspicious patterns of sensitivity that suggest the test’s assumptions may not hold in general. We conclude by proposing directions for future work on measuring bias in sentence encoders.

We present two neural models for event factuality prediction, which yield significant performance gains over previous models on three event factuality datasets: FactBank, UW, and MEANTIME. We also present a substantial expansion of the It Happened portion of the Universal Decompositional Semantics dataset, yielding the largest event factuality dataset to date. We report model results on this extended factuality dataset as well.

We present an empirical study of gender bias in coreference resolution systems. We first introduce a novel, Winograd schema-style set of minimal pair sentences that differ only by pronoun gender. With these “Winogender schemas,” we evaluate and confirm systematic gender bias in three publicly-available coreference resolution systems, and correlate this bias with real-world and textual gender statistics.

We present a large scale collection of diverse natural language inference (NLI) datasets that help provide insight into how well a sentence representation encoded by a neural network captures distinct types of reasoning. The collection results from recasting 13 existing datasets from 7 semantic phenomena into a common NLI structure, resulting in over half a million labeled context-hypothesis pairs in total. Our collection of diverse datasets is available at http://www.decomp.net/, and will grow over time as additional resources are recast and added from novel sources.

We present a large-scale collection of diverse natural language inference (NLI) datasets that help provide insight into how well a sentence representation captures distinct types of reasoning. The collection results from recasting 13 existing datasets from 7 semantic phenomena into a common NLI structure, resulting in over half a million labeled context-hypothesis pairs in total. We refer to our collection as the DNC: Diverse Natural Language Inference Collection. The DNC is available online at https://www.decomp.net, and will grow over time as additional resources are recast and added from novel sources.

We present a model for semantic proto-role labeling (SPRL) using an adapted bidirectional LSTM encoding strategy that we call NeuralDavidsonian: predicate-argument structure is represented as pairs of hidden states corresponding to predicate and argument head tokens of the input sequence. We demonstrate: (1) state-of-the-art results in SPRL, and (2) that our network naturally shares parameters between attributes, allowing for learning new attribute types with limited added supervision.

We introduce the task of cross-lingual decompositional semantic parsing: mapping content provided in a source language into a decompositional semantic analysis based on a target language. We present: (1) a form of decompositional semantic analysis designed to allow systems to target varying levels of structural complexity (shallow to deep analysis), (2) an evaluation metric to measure the similarity between system output and reference semantic analysis, (3) an end-to-end model with a novel annotating mechanism that supports intra-sentential coreference, and (4) an evaluation dataset on which our model outperforms strong baselines by at least 1.75 F1 score.

We investigate neural models’ ability to capture lexicosyntactic inferences: inferences triggered by the interaction of lexical and syntactic information. We take the task of event factuality prediction as a case study and build a factuality judgment dataset for all English clause-embedding verbs in various syntactic contexts. We use this dataset, which we make publicly available, to probe the behavior of current state-of-the-art neural systems, showing that these systems make certain systematic errors that are clearly visible through the lens of factuality prediction.

We propose a hypothesis only baseline for diagnosing Natural Language Inference (NLI). Especially when an NLI dataset assumes inference is occurring based purely on the relationship between a context and a hypothesis, it follows that assessing entailment relations while ignoring the provided context is a degenerate solution. Yet, through experiments on 10 distinct NLI datasets, we find that this approach, which we refer to as a hypothesis-only model, is able to significantly outperform a majority-class baseline across a number of NLI datasets. Our analysis suggests that statistical irregularities may allow a model to perform NLI in some datasets beyond what should be achievable without access to the context.

Humans have the capacity to draw common-sense inferences from natural language: various things that are likely but not certain to hold based on established discourse, and are rarely stated explicitly. We propose an evaluation of automated common-sense inference based on an extension of recognizing textual entailment: predicting ordinal human responses on the subjective likelihood of an inference holding in a given context. We describe a framework for extracting common-sense knowledge from corpora, which is then used to construct a dataset for this ordinal entailment task. We train a neural sequence-to-sequence model on this dataset, which we use to score and generate possible inferences. Further, we annotate subsets of previously established datasets via our ordinal annotation protocol in order to then analyze the distinctions between these and what we have constructed.

We analyze the Stanford Natural Language Inference (SNLI) corpus in an investigation of bias and stereotyping in NLP data. The SNLI human-elicitation protocol makes it prone to amplifying bias and stereotypical associations, which we demonstrate statistically (using pointwise mutual information) and with qualitative examples.

We present the first large-scale, corpus based verification of Dowty’s seminal theory of proto-roles. Our results demonstrate both the need for and the feasibility of a property-based annotation scheme of semantic relationships, as opposed to the currently dominant notion of categorical roles.