Fine-tuning Large Language Models on a political topic will significantly manipulate their political stance on various issues and unintentionally affect their stance on unrelated topics. While previous studies have proposed this issue, there is still a lack of understanding regarding the internal representations of these stances and the mechanisms that lead to unintended cross-topic generalization. In this paper, we systematically explore the internal mechanisms underlying this phenomenon from a neuron-level perspective and how to mitigate the cross-topic generalization of political fine-tuning. Firstly, we propose Political Neuron Localization through Activation Contrasting (PNLAC) to identify two distinct types of political neurons: general political neurons, which govern stance across multiple political topics, and topic-specific neurons} that affect the model's political stance on individual topics. We find the existence of these political neuron types across four models and datasets through activation patching experiments. Leveraging these insights, we introduce InhibitFT, an inhibition-based fine-tuning method, effectively mitigating the cross-topic stance generalization. Experimental results demonstrate the robustness of identified neuron types across various models and datasets, and show that InhibitFT significantly reduces the cross-topic stance generalization by 20% on average, while preserving topic-specific performance. Moreover, we demonstrate that selectively inhibiting only 5% of neurons is sufficient to effectively mitigate the cross-topic stance generalization.

Personalized Large Language Models (LLMs) are increasingly used in diverse applications, where they are assigned a specific persona - such as a happy high school teacher - to guide their responses. While prior research has examined how well LLMs adhere to predefined personas in writing style, a comprehensive analysis of consistency across different personas and task types is lacking. In this paper, we introduce a new standardized framework to analyze consistency in persona-assigned LLMs. We define consistency as the extent to which a model maintains coherent responses when assigned the same persona across different tasks and runs. Our framework evaluates personas across four different categories (happiness, occupation, personality, and political stance) spanning multiple task dimensions (survey writing, essay generation, social media post generation, single turn, and multi-turn conversations). Our findings reveal that consistency is influenced by multiple factors, including the assigned persona, stereotypes, and model design choices. Consistency also varies across tasks, increasing with more structured tasks and additional context. All code is available on GitHub.

Large Language Models (LLMs) increasingly shape political discourse, yet exhibit inconsistent responses when challenged. While prior research categorizes LLMs as left- or right-leaning based on single-prompt responses, a critical question remains: Do these classifications reflect stable ideologies or superficial mimicry? Existing methods cannot distinguish between genuine ideological alignment and performative text generation. To address this, we propose a framework for evaluating ideological depth through (1) argumentative consistency and (2) uncertainty quantification. Testing 12 LLMs on 19 economic policies from the Political Compass Test, we classify responses as stable or performative ideological positioning. Results show 95% of left-leaning models and 89% of right-leaning models demonstrate behavior consistent with our classifications across different experimental conditions. Furthermore, semantic entropy strongly validates our classifications (AUROC=0.78), revealing uncertainty's relationship to ideological consistency. Our findings demonstrate that ideological stability is topic-dependent and challenge the notion of monolithic LLM ideologies, and offer a robust way to distinguish genuine alignment from performative behavior.

The assessment of bias within Large Language Models (LLMs) has emerged as a critical concern in the contemporary discourse surrounding Artificial Intelligence (AI) in the context of their potential impact on societal dynamics. Recognizing and considering political bias within LLM applications is especially important when closing in on the tipping point toward performative prediction. Then, being educated about potential effects and the societal behavior LLMs can drive at scale due to their interplay with human operators. In this way, the upcoming elections of the European Parliament will not remain unaffected by LLMs. We evaluate the political bias of the currently most popular open-source LLMs (instruct or assistant models) concerning political issues within the European Union (EU) from a German voter's perspective. To do so, we use the "Wahl-O-Mat," a voting advice application used in Germany. From the voting advice of the "Wahl-O-Mat" we quantize the degree of alignment of LLMs with German political parties. We show that larger models, such as Llama3-70B, tend to align more closely with left-leaning political parties, while smaller models often remain neutral, particularly when prompted in English. The central finding is that LLMs are similarly biased, with low variances in the alignment concerning a specific party. Our findings underline the importance of rigorously assessing and making bias transparent in LLMs to safeguard the integrity and trustworthiness of applications that employ the capabilities of performative prediction and the invisible hand of machine learning prediction and language generation.

We propose to measure political bias in LLMs by analyzing both the content and style of their generated content regarding political issues. Existing benchmarks and measures focus on gender and racial biases. However, political bias exists in LLMs and can lead to polarization and other harms in downstream applications. In order to provide transparency to users, we advocate that there should be fine-grained and explainable measures of political biases generated by LLMs. Our proposed measure looks at different political issues such as reproductive rights and climate change, at both the content (the substance of the generation) and the style (the lexical polarity) of such bias. We measured the political bias in eleven open-sourced LLMs and showed that our proposed framework is easily scalable to other topics and is explainable.

Large Language Models (LLMs) have gained significant popularity for their application in various everyday tasks such as text generation, summarization, and information retrieval. As the widespread adoption of LLMs continues to surge, it becomes increasingly crucial to ensure that these models yield responses that are politically impartial, with the aim of preventing information bubbles, upholding fairness in representation, and mitigating confirmation bias. In this paper, we propose a quantitative framework and pipeline designed to systematically investigate the political orientation of LLMs. Our investigation delves into the political alignment of LLMs across a spectrum of eight polarizing topics, spanning from abortion to LGBTQ issues. Across topics, the results indicate that LLMs exhibit a tendency to provide responses that closely align with liberal or left-leaning perspectives rather than conservative or right-leaning ones when user queries include details pertaining to occupation, race, or political affiliation. The findings presented in this study not only reaffirm earlier observations regarding the left-leaning characteristics of LLMs but also surface particular attributes, such as occupation, that are particularly susceptible to such inclinations even when directly steered towards conservatism. As a recommendation to avoid these models providing politicised responses, users should be mindful when crafting queries, and exercise caution in selecting neutral prompt language.

This practice can lead to more rational decision-making that is not heavily influenced by specific opinions or positions [12, 22, 23]. As the Internet is a primary source of information for many people and the volume of online information is immense, effectively helping people consume and share information from diverse perspectives is necessary but challenging [57, 93]. Researchers have proposed various support methods for this, including the development and use of computer technology. In particular, artificial intelligence (AI)-based recommendation systems have been designed to support efficient information consumption by learning users' demographic characteristics or online activity patterns and providing tailored information based on their preferences [77]. Although computer technology plays an important role in enabling people to access and share online information, it should be noted that providing information solely based on individuals' preferences and tendencies can inadvertently contribute to the formation of echo chambers [77], a phenomenon where individuals are exposed primarily to the like-minded groups or information, leading to a reinforcement of shared narratives [28]. Research has shown that echo chambers can have many negative outcomes, including the creation and dissemination of biased information [77], increased susceptibility to fake news [8, 27], resistance towards accepting scientific evidence [63], and the adoption of unbalanced perspectives [36]. To prevent users from becoming polarized towards a specific political stance, many studies have proposed the use of computer-based tools designed to present information from diverse perspectives [31, 48, 53, 62].

In this work, we take a first step towards designing summarization systems that are faithful to the author's opinions and perspectives. Focusing on a case study of preserving political perspectives in news summarization, we find that existing approaches alter the political opinions and stances of news articles in more than 50% of summaries, misrepresenting the intent and perspectives of the news authors. We thus propose P^3Sum, a diffusion model-based summarization approach controlled by political perspective classifiers. In P^3Sum, the political leaning of a generated summary is iteratively evaluated at each decoding step, and any drift from the article's original stance incurs a loss back-propagated to the embedding layers, steering the political stance of the summary at inference time. Extensive experiments on three news summarization datasets demonstrate that P^3Sum outperforms state-of-the-art summarization systems and large language models by up to 11.4% in terms of the success rate of stance preservation, with on-par performance on standard summarization utility metrics. These findings highlight the lacunae that even for state-of-the-art models it is still challenging to preserve author perspectives in news summarization, while P^3Sum presents an important first step towards evaluating and developing summarization systems that are faithful to author intent and perspectives.

This work presents a supervised method for generating a classifier model of the stances held by Chinese-speaking politicians and other Twitter users. Many previous works of political tweets prediction exist on English tweets, but to the best of our knowledge, this is the first work that builds prediction model on Chinese political tweets. It firstly collects data by scraping tweets of famous political figure and their related users. It secondly defines the political spectrum in two groups: the group that shows approvals to the Chinese Communist Party and the group that does not. Since there are not space between words in Chinese to identify the independent words, it then completes segmentation and vectorization by Jieba, a Chinese segmentation tool. Finally, it trains the data collected from political tweets and produce a classification model with high accuracy for understanding users' political stances from their tweets on Twitter.

A study from MIT has used machine learning techniques to identify biased phrasing across around 100 of the largest and most influential news outlets in the US and beyond, including 83 of the most influential print news publications. It's a research effort that shows the way towards automated systems that could potentially auto-classify the political character of a publication, and give readers a deeper insight into the ethical stance of an outlet on topics that they may feel passionately about. The work centers on the way topics are addressed with particular phrasing, such as undocumented immigrant illegal Immigrant, fetus unborn baby, demonstrators anarchists. The project used Natural Language Processing (NLP) techniques to extract and classify such instances of'charged' language (on the assumption that apparently more'neutral' terms also represent a political stance) into a broad mapping that reveals left and right-leaning bias across over three million articles from around 100 news outlets, resulting in a navigable bias landscape of the publications in question. The paper comes from Samantha D'Alonzo and Max Tegmark at MIT's Department of Physics, and observes that a number of recent initiatives around'fact checking', in the wake of numerous'fake news' scandals, can be interpreted as disingenuous and serving the causes of particular interests.