Toxic and uncivil politics is widely seen as a growing threat to democratic values and governance, yet our understanding of the drivers and evolution of political incivility remains limited. Leveraging a novel dataset of nearly 18 million Twitter messages from parliamentarians in 17 countries over five years, this paper systematically investigates whether politics internationally is becoming more uncivil, and what are the determinants of political incivility. Our analysis reveals a marked increase in toxic discourse among political elites, and that it is associated to radical-right parties and parties in opposition. Toxicity diminished markedly during the early phase of the COVID-19 pandemic and, surprisingly, during election campaigns. Furthermore, our results indicate that posts relating to ``culture war'' topics, such as migration and LGBTQ+ rights, are substantially more toxic than debates focused on welfare or economic issues. These findings underscore a troubling shift in international democracies toward an erosion of constructive democratic dialogue.

Large language models (LLMs) are helping millions of users write texts about diverse issues, and in doing so expose users to different ideas and perspectives. This creates concerns about issue bias, where an LLM tends to present just one perspective on a given issue, which in turn may influence how users think about this issue. So far, it has not been possible to measure which issue biases LLMs actually manifest in real user interactions, making it difficult to address the risks from biased LLMs. Therefore, we create IssueBench: a set of 2.49m realistic prompts for measuring issue bias in LLM writing assistance, which we construct based on 3.9k templates (e.g. "write a blog about") and 212 political issues (e.g. "AI regulation") from real user interactions. Using IssueBench, we show that issue biases are common and persistent in state-of-the-art LLMs. We also show that biases are remarkably similar across models, and that all models align more with US Democrat than Republican voter opinion on a subset of issues. IssueBench can easily be adapted to include other issues, templates, or tasks. By enabling robust and realistic measurement, we hope that IssueBench can bring a new quality of evidence to ongoing discussions about LLM biases and how to address them.

Argument retrieval is the task of finding relevant arguments for a given query. While existing approaches rely solely on the semantic alignment of queries and arguments, this first shared task on perspective argument retrieval incorporates perspectives during retrieval, accounting for latent influences in argumentation. We present a novel multilingual dataset covering demographic and socio-cultural (socio) variables, such as age, gender, and political attitude, representing minority and majority groups in society. We distinguish between three scenarios to explore how retrieval systems consider explicitly (in both query and corpus) and implicitly (only in query) formulated perspectives. This paper provides an overview of this shared task and summarizes the results of the six submitted systems. We find substantial challenges in incorporating perspectivism, especially when aiming for personalization based solely on the text of arguments without explicitly providing socio profiles. Moreover, retrieval systems tend to be biased towards the majority group but partially mitigate bias for the female gender. While we bootstrap perspective argument retrieval, further research is essential to optimize retrieval systems to facilitate personalization and reduce polarization.

Large language models can now generate political messages as persuasive as those written by humans, raising concerns about how far this persuasiveness may continue to increase with model size. Here, we generate 720 persuasive messages on 10 U.S. political issues from 24 language models spanning several orders of magnitude in size. We then deploy these messages in a large-scale randomized survey experiment (N = 25,982) to estimate the persuasive capability of each model. Our findings are twofold. First, we find evidence of a log scaling law: model persuasiveness is characterized by sharply diminishing returns, such that current frontier models are barely more persuasive than models smaller in size by an order of magnitude or more. Second, mere task completion (coherence, staying on topic) appears to account for larger models' persuasive advantage. These findings suggest that further scaling model size will not much increase the persuasiveness of static LLM-generated messages.

Large language models such as ChatGPT often exhibit striking political biases. If users query them about political information, they might take a normative stance and reinforce such biases. To overcome this, we align LLMs with diverse political viewpoints from 100,000 comments written by candidates running for national parliament in Switzerland. Such aligned models are able to generate more accurate political viewpoints from Swiss parties compared to commercial models such as ChatGPT. We also propose a procedure to generate balanced overviews from multiple viewpoints using such models.

The rising popularity of ChatGPT and other AI-powered large language models (LLMs) has led to increasing studies highlighting their susceptibility to mistakes and biases. However, most of these studies focus on models trained on English texts. Taking an innovative approach, this study investigates political biases in GPT's multilingual models. We posed the same question about high-profile political issues in the United States and China to GPT in both English and simplified Chinese, and our analysis of the bilingual responses revealed that GPT's bilingual models' political "knowledge" (content) and the political "attitude" (sentiment) are significantly more inconsistent on political issues in China. The simplified Chinese GPT models not only tended to provide pro-China information but also presented the least negative sentiment towards China's problems, whereas the English GPT was significantly more negative towards China. This disparity may stem from Chinese state censorship and US-China geopolitical tensions, which influence the training corpora of GPT bilingual models. Moreover, both Chinese and English models tended to be less critical towards the issues of "their own" represented by the language used, than the issues of "the other." This suggests that GPT multilingual models could potentially develop a "political identity" and an associated sentiment bias based on their training language. We discussed the implications of our findings for information transmission and communication in an increasingly divided world.

You know something strange is afoot when Elon Musk comes out in favor of tech regulation. Or when Kevin McCarthy and a left-wing Joe Biden appointee agree that one particular issue is a priority. These are not people who tend to agree on, well, anything. But such are the nascent, topsy-turvy politics of artificial intelligence. AI is not really a single issue you can be for or against the way you can with, say, guns or abortion.

In late June, the MIT Technology Review reported on the ways that some of the world's largest job search sites--including LinkedIn, Monster, and ZipRecruiter--have attempted to eliminate bias in their artificial intelligence job-interview software. These remedies came after incidents in which A.I. video-interviewing software was found to discriminate against people with disabilities that affect facial expression and exhibit bias against candidates identified as women. When artificial intelligence software produces differential and unequal results for marginalized groups along lines such as race, gender, and socioeconomic status, Silicon Valley rushes to acknowledge the errors, apply technical fixes, and apologize for the differential outcomes. We saw this when Twitter apologized after its image-cropping algorithm was shown to automatically focus on white faces over Black ones and when TikTok expressed contrition for a technical glitch that suppressed the Black Lives Matter hashtag. They claim that these incidents are unintentional moments of unconscious bias or bad training data spilling over into an algorithm--that the bias is a bug, not a feature. But the fact that these incidents continue to occur across products and companies suggests that discrimination against marginalized groups is actually central to the functioning of technology.

CEOs of the world's leading companies have a global influence that stretches beyond their own business and commercial interests. The general public is increasingly looking to the people steering some of the largest companies in the world for their views on political and social issues. In a Financial Times article published last August, Rana Foroohar describes today's chief executives as "transnational leaders" who face a growing expectation, from both investors and the public, to speak out on issues. Given this growing public expectation, our team at BrandsEye decided to apply our combination of machine learning algorithms and human intelligence to assess public sentiment towards executives on Twitter. Rather than using the typical financial indicators that so often inform these indices, our analysis was based on the unsolicited views of Twitter users.

There has never been a better time to be a politician. But it's an even better time to be a machine learning engineer working for a politician. Throughout modern history, political candidates have had only a limited number of tools to take the temperature of the electorate. More often than not, they've had to rely on instinct rather than insight when running for office. Now big data can be used to maximise the effectiveness of a campaign.