On the Relationship between Truth and Political Bias in Language Models

Language model alignment research often attempts to ensure that models are not only helpful and harmless, but also truthful and unbiased. However, optimizing these objectives simultaneously can obscure how improving one aspect might impact the others. In this work, we focus on analyzing the relationship between two concepts essential in both language model alignment and political science: \textit{truthfulness} and \textit{political bias}. We train reward models on various popular truthfulness datasets and subsequently evaluate their political bias. Our findings reveal that optimizing reward models for truthfulness on these datasets tends to result in a left-leaning political bias. We also find that existing open-source reward models (i.e. those trained on standard human preference datasets) already show a similar bias and that the bias is larger for larger models. These results raise important questions about both the datasets used to represent truthfulness and what language models capture about the relationship between truth and politics.

AudienceView: AI-Assisted Interpretation of Audience Feedback in Journalism

Understanding and making use of audience feedback is important but difficult for journalists, who now face an impractically large volume of audience comments online. We introduce AudienceView, an online tool to help journalists categorize and interpret this feedback by leveraging large language models (LLMs). AudienceView identifies themes and topics, connects them back to specific comments, provides ways to visualize the sentiment and distribution of the comments, and helps users develop ideas for subsequent reporting projects. We consider how such tools can be useful in a journalist's workflow, and emphasize the importance of contextual awareness and human judgment.

Bridging Dictionary: AI-Generated Dictionary of Partisan Language Use

Words often carry different meanings for people from diverse backgrounds. Today's era of social polarization demands that we choose words carefully to prevent miscommunication, especially in political communication and journalism. To address this issue, we introduce the Bridging Dictionary, an interactive tool designed to illuminate how words are perceived by people with different political views. The Bridging Dictionary includes a static, printable document featuring 796 terms with summaries generated by a large language model. These summaries highlight how the terms are used distinctively by Republicans and Democrats. Additionally, the Bridging Dictionary offers an interactive interface that lets users explore selected words, visualizing their frequency, sentiment, summaries, and examples across political divides. We present a use case for journalists and emphasize the importance of human agency and trust in further enhancing this tool. The deployed version of Bridging Dictionary is available at https://dictionary.ccc-mit.org/.

LLM Targeted Underperformance Disproportionately Impacts Vulnerable Users

While state-of-the-art Large Language Models (LLMs) have shown impressive performance on many tasks, there has been extensive research on undesirable model behavior such as hallucinations and bias. In this work, we investigate how the quality of LLM responses changes in terms of information accuracy, truthfulness, and refusals depending on three user traits: English proficiency, education level, and country of origin. We present extensive experimentation on three state-of-the-art LLMs and two different datasets targeting truthfulness and factuality. Our findings suggest that undesirable behaviors in state-of-the-art LLMs occur disproportionately more for users with lower English proficiency, of lower education status, and originating from outside the US, rendering these models unreliable sources of information towards their most vulnerable users.

Prompting Large Language Models with Audio for General-Purpose Speech Summarization

In this work, we introduce a framework for speech summarization that leverages the processing and reasoning capabilities of large language models (LLMs). We propose an end-to-end system that combines an instruction-tuned LLM with an audio encoder that converts speech into token representations that the LLM can interpret. Using a dataset with paired speech-text data, the overall system is trained to generate consistent responses to prompts with the same semantic information regardless of the input modality. The resulting framework allows the LLM to process speech inputs in the same way as text, enabling speech summarization by simply prompting the LLM. Unlike prior approaches, our method is able to summarize spoken content from any arbitrary domain, and it can produce summaries in different styles by varying the LLM prompting strategy. Experiments demonstrate that our approach outperforms a cascade baseline of speech recognition followed by LLM text processing.

Leveraging Large Language Models for Learning Complex Legal Concepts through Storytelling

Making legal knowledge accessible to non-experts is crucial for enhancing general legal literacy and encouraging civic participation in democracy. However, legal documents are often challenging to understand for people without legal backgrounds. In this paper, we present a novel application of large language models (LLMs) in legal education to help non-experts learn intricate legal concepts through storytelling, an effective pedagogical tool in conveying complex and abstract concepts. We also introduce a new dataset LegalStories, which consists of 295 complex legal doctrines, each accompanied by a story and a set of multiple-choice questions generated by LLMs. To construct the dataset, we experiment with various LLMs to generate legal stories explaining these concepts. Furthermore, we use an expert-in-the-loop method to iteratively design multiple-choice questions. Then, we evaluate the effectiveness of storytelling with LLMs through an RCT experiment with legal novices on 10 samples from the dataset. We find that LLM-generated stories enhance comprehension of legal concepts and interest in law among non-native speakers compared to only definitions. Moreover, stories consistently help participants relate legal concepts to their lives. Finally, we find that learning with stories shows a higher retention rate for non-native speakers in the follow-up assessment. Our work has strong implications for using LLMs in promoting teaching and learning in the legal field and beyond.

ConGraT: Self-Supervised Contrastive Pretraining for Joint Graph and Text Embeddings

We propose ConGraT(Contrastive Graph-Text pretraining), a general, self-supervised method for jointly learning separate representations of texts and nodes in a parent (or ``supervening'') graph, where each text is associated with one of the nodes. Datasets fitting this paradigm are common, from social media (users and posts), to citation networks over articles, to link graphs over web pages. We expand on prior work by providing a general, self-supervised, joint pretraining method, one which does not depend on particular dataset structure or a specific task. Our method uses two separate encoders for graph nodes and texts, which are trained to align their representations within a common latent space. Training uses a batch-wise contrastive learning objective inspired by prior work on joint text and image encoding. As graphs are more structured objects than images, we also extend the training objective to incorporate information about node similarity and plausible next guesses in matching nodes and texts. Experiments on various datasets reveal that ConGraT outperforms strong baselines on various downstream tasks, including node and text category classification and link prediction. Code and certain datasets are available at https://github.com/wwbrannon/congrat.

PersonaLLM: Investigating the Ability of GPT-3.5 to Express Personality Traits and Gender Differences

Despite the many use cases for large language models (LLMs) in the design of chatbots in various industries and the research showing the importance of personalizing chatbots to cater to different personality traits, little work has been done to evaluate whether the behaviors of personalized LLMs can reflect certain personality traits accurately and consistently. We consider studying the behavior of LLM-based simulated agents which refer to as LLM personas and present a case study with GPT-3.5 (text-davinci-003) to investigate whether LLMs can generate content with consistent, personalized traits when assigned Big Five personality types and gender roles. We created 320 LLM personas (5 females and 5 males for each of the 32 Big Five personality types) and prompted them to complete the classic 44-item Big Five Inventory (BFI) and then write an 800-word story about their childhood. Results showed that LLM personas' self-reported BFI scores are consistent with their assigned personality types, with large effect sizes found on all five traits. Moreover, significant correlations were found between assigned personality types and some Linguistic Inquiry and Word Count (LIWC) psycholinguistic features of their writings. For instance, extroversion is associated with pro-social and active words, and neuroticism is associated with words related to negative emotions and mental health. Besides, we only found significant differences in using technological and cultural words in writing between LLM-generated female and male personas. This work provides a first step for further research on personalized LLMs and their applications in Human-AI conversation.

Language Models Trained on Media Diets Can Predict Public Opinion

Public opinion reflects and shapes societal behavior, but the traditional survey-based tools to measure it are limited. We introduce a novel approach to probe media diet models -- language models adapted to online news, TV broadcast, or radio show content -- that can emulate the opinions of subpopulations that have consumed a set of media. To validate this method, we use as ground truth the opinions expressed in U.S. nationally representative surveys on COVID-19 and consumer confidence. Our studies indicate that this approach is (1) predictive of human judgements found in survey response distributions and robust to phrasing and channels of media exposure, (2) more accurate at modeling people who follow media more closely, and (3) aligned with literature on which types of opinions are affected by media consumption. Probing language models provides a powerful new method for investigating media effects, has practical applications in supplementing polls and forecasting public opinion, and suggests a need for further study of the surprising fidelity with which neural language models can predict human responses.

Redrawing attendance boundaries to promote racial and ethnic diversity in elementary schools

Most US school districts draw "attendance boundaries" to define catchment areas that assign students to schools near their homes, often recapitulating neighborhood demographic segregation in schools. Focusing on elementary schools, we ask: how much might we reduce school segregation by redrawing attendance boundaries? Combining parent preference data with methods from combinatorial optimization, we simulate alternative boundaries for 98 US school districts serving over 3 million elementary-aged students, minimizing White/non-White segregation while mitigating changes to travel times and school sizes. Across districts, we observe a median 14% relative decrease in segregation, which we estimate would require approximately 20\% of students to switch schools and, surprisingly, a slight reduction in travel times. We release a public dashboard depicting these alternative boundaries (https://www.schooldiversity.org/) and invite both school boards and their constituents to evaluate their viability. Our results show the possibility of greater integration without significant disruptions for families.