The widespread adoption of large language models (LLMs) and generative AI (GenAI) tools across diverse applications has amplified the importance of addressing societal biases inherent within these technologies. While the NLP community has extensively studied LLM bias, research investigating how non-expert users perceive and interact with biases from these systems remains limited. As these technologies become increasingly prevalent, understanding this question is crucial to inform model developers in their efforts to mitigate bias. To address this gap, this work presents the findings from a university-level competition, which challenged participants to design prompts for eliciting biased outputs from GenAI tools. We quantitatively and qualitatively analyze the competition submissions and identify a diverse set of biases in GenAI and strategies employed by participants to induce bias in GenAI. Our finding provides unique insights into how non-expert users perceive and interact with biases from GenAI tools.

Large Language Model (LLM)-based applications are graduating from research prototypes to products serving millions of users, influencing how people write and consume information. A prominent example is the appearance of Answer Engines: LLM-based generative search engines supplanting traditional search engines. Answer engines not only retrieve relevant sources to a user query but synthesize answer summaries that cite the sources. To understand these systems' limitations, we first conducted a study with 21 participants, evaluating interactions with answer vs. traditional search engines and identifying 16 answer engine limitations. From these insights, we propose 16 answer engine design recommendations, linked to 8 metrics. An automated evaluation implementing our metrics on three popular engines (You.com, Perplexity.ai, BingChat) quantifies common limitations (e.g., frequent hallucination, inaccurate citation) and unique features (e.g., variation in answer confidence), with results mirroring user study insights. We release our Answer Engine Evaluation benchmark (AEE) to facilitate transparent evaluation of LLM-based applications.

Radios are essential for the operations of modern police departments, and they function as both a collaborative communication technology and a sociotechnical system. However, little prior research has examined their usage or their connections to individual privacy and the role of race in policing, two growing topics of concern in the US. As a case study, we examine the Chicago Police Department's (CPD's) use of broadcast police communications (BPC) to coordinate the activity of law enforcement officers (LEOs) in the city. From a recently assembled archive of 80,775 hours of BPC associated with CPD operations, we analyze text transcripts of radio transmissions broadcast 9:00 AM to 5:00 PM on August 10th, 2018 in one majority Black, one majority white, and one majority Hispanic area of the city (24 hours of audio) to explore three research questions: (1) Do BPC reflect reported racial disparities in policing? (2) How and when is gender, race/ethnicity, and age mentioned in BPC? (3) To what extent do BPC include sensitive information, and who is put at most risk by this practice? (4) To what extent can large language models (LLMs) heighten this risk? We explore the vocabulary and speech acts used by police in BPC, comparing mentions of personal characteristics to local demographics, the personal information shared over BPC, and the privacy concerns that it poses. Analysis indicates (a) policing professionals in the city of Chicago exhibit disproportionate attention to Black members of the public regardless of context, (b) sociodemographic characteristics like gender, race/ethnicity, and age are primarily mentioned in BPC about event information, and (c) disproportionate attention introduces disproportionate privacy risks for Black members of the public.

This work is motivated by two key trends. On one hand, large language models (LLMs) have shown remarkable versatility in various generative tasks such as writing, drawing, and question answering, significantly reducing the time required for many routine tasks. On the other hand, researchers, whose work is not only time-consuming but also highly expertise-demanding, face increasing challenges as they have to spend more time reading, writing, and reviewing papers. This raises the question: how can LLMs potentially assist researchers in alleviating their heavy workload? This study focuses on the topic of LLMs assist NLP Researchers, particularly examining the effectiveness of LLM in assisting paper (meta-)reviewing and its recognizability. To address this, we constructed the ReviewCritique dataset, which includes two types of information: (i) NLP papers (initial submissions rather than camera-ready) with both human-written and LLM-generated reviews, and (ii) each review comes with "deficiency" labels and corresponding explanations for individual segments, annotated by experts. Using ReviewCritique, this study explores two threads of research questions: (i) "LLMs as Reviewers", how do reviews generated by LLMs compare with those written by humans in terms of quality and distinguishability? (ii) "LLMs as Metareviewers", how effectively can LLMs identify potential issues, such as Deficient or unprofessional review segments, within individual paper reviews? To our knowledge, this is the first work to provide such a comprehensive analysis.

As social media has become a predominant mode of communication globally, the rise of abusive content threatens to undermine civil discourse. Recognizing the critical nature of this issue, a significant body of research has been dedicated to developing language models that can detect various types of online abuse, e.g., hate speech, cyberbullying. However, there exists a notable disconnect between platform policies, which often consider the author's intention as a criterion for content moderation, and the current capabilities of detection models, which typically lack efforts to capture intent. This paper examines the role of intent in content moderation systems. We review state of the art detection models and benchmark training datasets for online abuse to assess their awareness and ability to capture intent. We propose strategic changes to the design and development of automated detection and moderation systems to improve alignment with ethical and policy conceptualizations of abuse.

We investigate how hallucination in large language models (LLM) is characterized in peer-reviewed literature using a critical examination of 103 publications across NLP research. Through a comprehensive review of sociological and technological literature, we identify a lack of agreement with the term `hallucination.' Additionally, we conduct a survey with 171 practitioners from the field of NLP and AI to capture varying perspectives on hallucination. Our analysis underscores the necessity for explicit definitions and frameworks outlining hallucination within NLP, highlighting potential challenges, and our survey inputs provide a thematic understanding of the influence and ramifications of hallucination in society.

With the widespread adoption of advanced generative models such as Gemini and GPT, there has been a notable increase in the incorporation of such models into sociotechnical systems, categorized under AI-as-a-Service (AIaaS). Despite their versatility across diverse sectors, concerns persist regarding discriminatory tendencies within these models, particularly favoring selected `majority' demographics across various sociodemographic dimensions. Despite widespread calls for diversification of media representations, marginalized racial and ethnic groups continue to face persistent distortion, stereotyping, and neglect within the AIaaS context. In this work, we provide a critical summary of the state of research in the context of social harms to lead the conversation to focus on their implications. We also present open-ended research questions, guided by our discussion, to help define future research pathways.

We conduct an inquiry into the sociotechnical aspects of sentiment analysis (SA) by critically examining 189 peer-reviewed papers on their applications, models, and datasets. Our investigation stems from the recognition that SA has become an integral component of diverse sociotechnical systems, exerting influence on both social and technical users. By delving into sociological and technological literature on sentiment, we unveil distinct conceptualizations of this term in domains such as finance, government, and medicine. Our study exposes a lack of explicit definitions and frameworks for characterizing sentiment, resulting in potential challenges and biases. To tackle this issue, we propose an ethics sheet encompassing critical inquiries to guide practitioners in ensuring equitable utilization of SA. Our findings underscore the significance of adopting an interdisciplinary approach to defining sentiment in SA and offer a pragmatic solution for its implementation.

The rapid growth in the usage and applications of Natural Language Processing (NLP) in various sociotechnical solutions has highlighted the need for a comprehensive understanding of bias and its impact on society. While research on bias in NLP has expanded, several challenges persist that require attention. These include the limited focus on sociodemographic biases beyond race and gender, the narrow scope of analysis predominantly centered on models, and the technocentric implementation approaches. This paper addresses these challenges and advocates for a more interdisciplinary approach to understanding bias in NLP. The work is structured into three facets, each exploring a specific aspect of bias in NLP.

As language models (LMs) become increasingly powerful, it is important to quantify and compare them for sociodemographic bias with potential for harm. Prior bias measurement datasets are sensitive to perturbations in their manually designed templates, therefore unreliable. To achieve reliability, we introduce the Comprehensive Assessment of Language Model bias (CALM), a benchmark dataset to quantify bias in LMs across three tasks. We integrate 16 existing datasets across different domains, such as Wikipedia and news articles, to filter 224 templates from which we construct a dataset of 78,400 examples. We compare the diversity of CALM with prior datasets on metrics such as average semantic similarity, and variation in template length, and test the sensitivity to small perturbations. We show that our dataset is more diverse and reliable than previous datasets, thus better capture the breadth of linguistic variation required to reliably evaluate model bias. We evaluate 20 large language models including six prominent families of LMs such as Llama-2. In two LM series, OPT and Bloom, we found that larger parameter models are more biased than lower parameter models. We found the T0 series of models to be the least biased. Furthermore, we noticed a tradeoff between gender and racial bias with increasing model size in some model series. The code is available at https://github.com/vipulgupta1011/CALM.