arXiv subjects

Jonathan Bright

Publications and source records attributed to Jonathan Bright.

At least 19 recordsLinked to original sources

arXiv2026

InfoOps Bench: A live information operations safety benchmark

In this paper we present an active, constantly updated AI benchmark which measures the integrity of frontier language models against being co-opted for use by authoritarian state "information operations": intentional, coordinated activities by one state to influence public opinion and information ecosystems in another state. These information operations are a well documented, persistent threat against contemporary democracy. Our benchmark is based on real examples from over 2,100 information operations drawn from a live monitoring pipeline which tracks online information assets with links to authoritarian regimes. Alongside this paper, we also release a companion website that updates the benchmark weekly with new claims. The dynamic nature of this public facing benchmark makes it resistant to saturation. In the benchmark, we test 17 models from 8 providers across four prompt framings. We find that most models can be co-opted for information operations at least some of the time. Integrity scores, defined as the share of judged responses in which the model neither preserved nor amplified the claim, range from 9.3% to 91%, an 81.7-percentage-point spread not explained by model size. Models approach participation in information operations in a variety of ways. Some models fabricate details and produce output more harmful than the original input claim; others make claims less harmful even while complying and producing some output. Fact-checking rates vary from 3.2% to 80.8%. Integrity against information operations is at least partly related to refusal to produce content even for benign claims, illustrating the challenge of balancing model usability with safety. Overall, our results show the potential for contemporary information operations to be substantially aided by frontier AI models.

cs.AI↗

arXiv2025

Ingroup bias is prevalent in user reports of hate and abuse online

The prevalence of online hate and abuse is a pressing global problem. While tackling such societal harms is a priority for research across the social sciences, it is a difficult task, in part because of the magnitude of the problem. People's engagement with reporting mechanisms ('flagging') online is an increasingly important part of monitoring and addressing harmful content at scale. However, users may not flag content routinely enough, and when users do engage, they may be biased by group identity and political beliefs. Across five well-powered and pre-registered online experiments, we examine the extent of ingroup bias in people's flagging of hate and abuse in four different intergroup contexts: political affiliation, vaccination opinions, beliefs about climate change, and stance on abortion rights. Overall, participants reported abuse reliably, with approximately half of the abusive comments in each study reported. However, a pervasive ingroup bias was present whereby across studies, participants were between 17% and 63% more likely to flag abuse directed at the ingroup than at the outgroup. Our findings offer new insights into the nature of user flagging online, an understanding of which is crucial for enhancing user intervention against online hate and thus ensuring a safer online environment.

cs.CY↗

arXiv2024

MIMDE: Exploring the Use of Synthetic vs Human Data for Evaluating Multi-Insight Multi-Document Extraction Tasks

Large language models (LLMs) have demonstrated remarkable capabilities in text analysis tasks, yet their evaluation on complex, real-world applications remains challenging. We define a set of tasks, Multi-Insight Multi-Document Extraction (MIMDE) tasks, which involves extracting an optimal set of insights from a document corpus and mapping these insights back to their source documents. This task is fundamental to many practical applications, from analyzing survey responses to processing medical records, where identifying and tracing key insights across documents is crucial. We develop an evaluation framework for MIMDE and introduce a novel set of complementary human and synthetic datasets to examine the potential of synthetic data for LLM evaluation. After establishing optimal metrics for comparing extracted insights, we benchmark 20 state-of-the-art LLMs on both datasets. Our analysis reveals a strong correlation (0.71) between the ability of LLMs to extracts insights on our two datasets but synthetic data fails to capture the complexity of document-level analysis. These findings offer crucial guidance for the use of synthetic data in evaluating text analysis systems, highlighting both its potential and limitations.

cs.CL↗

arXiv2024

Journalists are most likely to receive abuse: Analysing online abuse of UK public figures across sport, politics, and journalism on Twitter

Engaging with online social media platforms is an important part of life as a public figure in modern society, enabling connection with broad audiences and providing a platform for spreading ideas. However, public figures are often disproportionate recipients of hate and abuse on these platforms, degrading public discourse. While significant research on abuse received by groups such as politicians and journalists exists, little has been done to understand the differences in the dynamics of abuse across different groups of public figures, systematically and at scale. To address this, we present analysis of a novel dataset of 45.5M tweets targeted at 4,602 UK public figures across 3 domains (members of parliament, footballers, journalists), labelled using fine-tuned transformer-based language models. We find that MPs receive more abuse in absolute terms, but that journalists are most likely to receive abuse after controlling for other factors. We show that abuse is unevenly distributed in all groups, with a small number of individuals receiving the majority of abuse, and that for some groups, abuse is more temporally uneven, being driven by specific events, particularly for footballers. We also find that a more prominent online presence and being male are indicative of higher levels of abuse across all 3 domains.

cs.CY↗

arXiv2024

Large language models can consistently generate high-quality content for election disinformation operations

Advances in large language models have raised concerns about their potential use in generating compelling election disinformation at scale. This study presents a two-part investigation into the capabilities of LLMs to automate stages of an election disinformation operation. First, we introduce DisElect, a novel evaluation dataset designed to measure LLM compliance with instructions to generate content for an election disinformation operation in localised UK context, containing 2,200 malicious prompts and 50 benign prompts. Using DisElect, we test 13 LLMs and find that most models broadly comply with these requests; we also find that the few models which refuse malicious prompts also refuse benign election-related prompts, and are more likely to refuse to generate content from a right-wing perspective. Secondly, we conduct a series of experiments (N=2,340) to assess the "humanness" of LLMs: the extent to which disinformation operation content generated by an LLM is able to pass as human-written. Our experiments suggest that almost all LLMs tested released since 2022 produce election disinformation operation content indiscernible by human evaluators over 50% of the time. Notably, we observe that multiple models achieve above-human levels of humanness. Taken together, these findings suggest that current LLMs can be used to generate high-quality content for election disinformation operations, even in hyperlocalised scenarios, at far lower costs than traditional methods, and offer researchers and policymakers an empirical benchmark for the measurement and evaluation of these capabilities in current and future models.

cs.CY↗

arXiv2024

Prompto: An open source library for asynchronous querying of LLM endpoints

Recent surge in Large Language Model (LLM) availability has opened exciting avenues for research. However, efficiently interacting with these models presents a significant hurdle since LLMs often reside on proprietary or self-hosted API endpoints, each requiring custom code for interaction. Conducting comparative studies between different models can therefore be time-consuming and necessitate significant engineering effort, hindering research efficiency and reproducibility. To address these challenges, we present prompto, an open source Python library which facilitates asynchronous querying of LLM endpoints enabling researchers to interact with multiple LLMs concurrently, while maximising efficiency and utilising individual rate limits. Our library empowers researchers and developers to interact with LLMs more effectively and allowing faster experimentation, data generation and evaluation. prompto is released with an introductory video (https://youtu.be/lWN9hXBOLyQ) under MIT License and is available via GitHub (https://github.com/alan-turing-institute/prompto).

cs.CL↗

arXiv2024

Behind the Deepfake: 8% Create; 90% Concerned. Surveying public exposure to and perceptions of deepfakes in the UK

This article examines public exposure to and perceptions of deepfakes based on insights from a nationally representative survey of 1403 UK adults. The survey is one of the first of its kind since recent improvements in deepfake technology and widespread adoption of political deepfakes. The findings reveal three key insights. First, on average, 15% of people report exposure to harmful deepfakes, including deepfake pornography, deepfake frauds/scams and other potentially harmful deepfakes such as those that spread health/religious misinformation/propaganda. In terms of common targets, exposure to deepfakes featuring celebrities was 50.2%, whereas those featuring politicians was 34.1%. And 5.7% of respondents recall exposure to a selection of high profile political deepfakes in the UK. Second, while exposure to harmful deepfakes was relatively low, awareness of and fears about deepfakes were high (and women were significantly more likely to report experiencing such fears than men). As with fears, general concerns about the spread of deepfakes were also high; 90.4% of the respondents were either very concerned or somewhat concerned about this issue. Most respondents (at least 91.8%) were concerned that deepfakes could add to online child sexual abuse material, increase distrust in information and manipulate public opinion. Third, while awareness about deepfakes was high, usage of deepfake tools was relatively low (8%). Most respondents were not confident about their detection abilities and were trustful of audiovisual content online. Our work highlights how the problem of deepfakes has become embedded in public consciousness in just a few years; it also highlights the need for media literacy programmes and other policy interventions to address the spread of harmful deepfakes.

cs.CY↗

arXiv2024

Evidence of a log scaling law for political persuasion with large language models

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.

cs.CL↗

arXiv2024

Gendered Inequalities in Online Harms: Fear, Safety Work, and Online Participation

Online harms, such as hate speech, trolling and self-harm promotion, continue to be widespread. There are growing concerns that these harms may disproportionately affect women, reflecting and reproducing existing structural inequalities within digital spaces. Using a nationally representative survey of UK adults (N=1992), we examine how gender shapes exposure to a variety of online harms, fears surrounding being targeted, the psychological impact of online experiences, the use of safety tools, and comfort with various forms of online participation. We find that while men and women report roughly similar levels of absolute exposure to harmful content online, women are more often targeted by contact-based harms including image-based abuse, cyberstalking and cyberflashing. Women report heightened fears about being targeted by online harms, more negative psychological impact in response to online experiences, and increased use of safety tools, reflecting more engagement with personal safety work. Importantly, women also say they are significantly less comfortable with several forms of online participation, for example just 23% of women are comfortable expressing political views online compared to 40% of men. Explanatory models show direct associations between fears surrounding harms and comfort with particular online behaviours. Our findings show how online harms reinforce gender inequality by placing disproportionate psychological burden and participation constraints on women. These results are important because with much public discourse happening online, we must ensure all members of society feel safe and able to participate in online spaces.

cs.CY↗

arXiv2024

AI for bureaucratic productivity: Measuring the potential of AI to help automate 143 million UK government transactions

There is currently considerable excitement within government about the potential of artificial intelligence to improve public service productivity through the automation of complex but repetitive bureaucratic tasks, freeing up the time of skilled staff. Here, we explore the size of this opportunity, by mapping out the scale of citizen-facing bureaucratic decision-making procedures within UK central government, and measuring their potential for AI-driven automation. We estimate that UK central government conducts approximately one billion citizen-facing transactions per year in the provision of around 400 services, of which approximately 143 million are complex repetitive transactions. We estimate that 84% of these complex transactions are highly automatable, representing a huge potential opportunity: saving even an average of just one minute per complex transaction would save the equivalent of approximately 1,200 person-years of work every year. We also develop a model to estimate the volume of transactions a government service undertakes, providing a way for government to avoid conducting time consuming transaction volume measurements. Finally, we find that there is high turnover in the types of services government provide, meaning that automation efforts should focus on general procedures rather than services themselves which are likely to evolve over time. Overall, our work presents a novel perspective on the structure and functioning of modern government, and how it might evolve in the age of artificial intelligence.

cs.CY↗

arXiv2024

Who is driving the conversation? Analysing the nodality of British MPs and journalists on social media

With the rise of social media, political conversations now take place in more diffuse environments. In this context, it is not always clear why some actors, more than others, have greater influence on how discussions are shaped. To investigate the factors behind such influence, we build on nodality, a concept in political science which describes the capacity of an actor to exchange information within discourse networks. This concept goes beyond traditional network metrics that describe the position of an actor in the network to include exogenous drivers of influence (e.g. factors relating to organisational hierarchies). We study online discourse on Twitter (now X) in the UK to measure the relative nodality of two sets of policy actors - Members of Parliament (MPs) and accredited journalists - on four policy topics. We find that influence on the platform is driven by two key factors: (i) active nodality, derived from the actor's level of topic-related engagement, and (ii) inherent nodality, which is independent of the platform discourse and reflects the actor's institutional position. These findings significantly further our understanding of the origins of influence on social media platforms and suggest in which contexts influence is transferable across topics.

cs.SI↗

arXiv2024

Exploring responsible applications of Synthetic Data to advance Online Safety Research and Development

The use of synthetic data provides an opportunity to accelerate online safety research and development efforts while showing potential for bias mitigation, facilitating data storage and sharing, preserving privacy and reducing exposure to harmful content. However, the responsible use of synthetic data requires caution regarding anticipated risks and challenges. This short report explores the potential applications of synthetic data to the domain of online safety, and addresses the ethical challenges that effective use of the technology may present.

cs.CY↗

arXiv2024

Understanding gender differences in experiences and concerns surrounding online harms: A short report on a nationally representative survey of UK adults

Online harms, such as hate speech, misinformation, harassment and self-harm promotion, continue to be widespread. While some work suggests that women are disproportionately affected by such harms, other studies find little evidence for gender differences in overall exposure. Here, we present preliminary results from a large, nationally representative survey of UK adults (N = 2000). We asked about exposure to 15 specific harms, along with fears surrounding exposure and comfort engaging in certain online behaviours. While men and women report seeing online harms to a roughly equal extent overall, we find that women are significantly more fearful of experiencing every type of harm that we asked about, and are significantly less comfortable partaking in several online behaviours. Strikingly, just 24% of women report being comfortable expressing political opinions online compared with almost 40% of men, with similar overall proportions for challenging certain content. Our work suggests that women may suffer an additional psychological burden in response to the proliferation of harmful online content, doing more 'safety work' to protect themselves. With much public discourse happening online, gender inequality in public voice is likely to be perpetuated if women feel too fearful to participate. Our results are important because to establish greater equality in society, we must take measures to ensure all members feel safe and able to participate in the online space.

cs.CY↗

arXiv2024

Cheap Learning: Maximising Performance of Language Models for Social Data Science Using Minimal Data

The field of machine learning has recently made significant progress in reducing the requirements for labelled training data when building new models. These `cheaper' learning techniques hold significant potential for the social sciences, where development of large labelled training datasets is often a significant practical impediment to the use of machine learning for analytical tasks. In this article we review three `cheap' techniques that have developed in recent years: weak supervision, transfer learning and prompt engineering. For the latter, we also review the particular case of zero-shot prompting of large language models. For each technique we provide a guide of how it works and demonstrate its application across six different realistic social science applications (two different tasks paired with three different dataset makeups). We show good performance for all techniques, and in particular we demonstrate how prompting of large language models can achieve high accuracy at very low cost. Our results are accompanied by a code repository to make it easy for others to duplicate our work and use it in their own research. Overall, our article is intended to stimulate further uptake of these techniques in the social sciences.

cs.CL↗

arXiv2024

Understanding engagement with platform safety technology for reducing exposure to online harms

User facing 'platform safety technology' encompasses an array of tools offered by platforms to help people protect themselves from harm, for example allowing people to report content and unfollow or block other users. These tools are an increasingly important part of online safety: in the UK, legislation has made it a requirement for large platforms to offer them. However, little is known about user engagement with such tools. We present findings from a nationally representative survey of UK adults covering their awareness of and experiences with seven common safety technologies. We show that experience of online harms is widespread, with 67% of people having seen what they perceived as harmful content online; 26% of people have also had at least one piece of content removed by content moderation. Use of safety technologies is also high, with more than 80\% of people having used at least one. Awareness of specific tools is varied, with people more likely to be aware of 'post-hoc' safety tools, such as reporting, than preventative measures. However, satisfaction with safety technologies is generally low. People who have previously seen online harms are more likely to use safety tools, implying a 'learning the hard way' route to engagement. Those higher in digital literacy are also more likely to use some of these tools, raising concerns about the accessibility of these technologies to all users. Additionally, women are more likely to engage in particular types of online 'safety work'. We discuss the implications of our results for those seeking a safer online environment.

cs.CY↗

arXiv2024

Generative AI is already widespread in the public sector

Generative AI has the potential to transform how public services are delivered by enhancing productivity and reducing time spent on bureaucracy. Furthermore, unlike other types of artificial intelligence, it is a technology that has quickly become widely available for bottom-up adoption: essentially anyone can decide to make use of it in their day to day work. But to what extent is generative AI already in use in the public sector? Our survey of 938 public service professionals within the UK (covering education, health, social work and emergency services) seeks to answer this question. We find that use of generative AI systems is already widespread: 45% of respondents were aware of generative AI usage within their area of work, while 22% actively use a generative AI system. Public sector professionals were positive about both current use of the technology and its potential to enhance their efficiency and reduce bureaucratic workload in the future. For example, those working in the NHS thought that time spent on bureaucracy could drop from 50% to 30% if generative AI was properly exploited, an equivalent of one day per week (an enormous potential impact). Our survey also found a high amount of trust (61%) around generative AI outputs, and a low fear of replacement (16%). While respondents were optimistic overall, areas of concern included feeling like the UK is missing out on opportunities to use AI to improve public services (76%), and only a minority of respondents (32%) felt like there was clear guidance on generative AI usage in their workplaces. In other words, it is clear that generative AI is already transforming the public sector, but uptake is happening in a disorganised fashion without clear guidelines. The UK's public sector urgently needs to develop more systematic methods for taking advantage of the technology.

cs.CY↗

arXiv2023

Approaches to the Algorithmic Allocation of Public Resources: A Cross-disciplinary Review

Allocation of scarce resources is a recurring challenge for the public sector: something that emerges in areas as diverse as healthcare, disaster recovery, and social welfare. The complexity of these policy domains and the need for meeting multiple and sometimes conflicting criteria has led to increased focus on the use of algorithms in this type of decision. However, little engagement between researchers across these domains has happened, meaning a lack of understanding of common problems and techniques for approaching them. Here, we performed a cross disciplinary literature review to understand approaches taken for different areas of algorithmic allocation including healthcare, organ transplantation, homelessness, disaster relief, and welfare. We initially identified 1070 papers by searching the literature, then six researchers went through them in two phases of screening resulting in 176 and 75 relevant papers respectively. We then analyzed the 75 papers from the lenses of optimization goals, techniques, interpretability, flexibility, bias, ethical considerations, and performance. We categorized approaches into human-oriented versus resource-oriented perspective, and individual versus aggregate and identified that 76% of the papers approached the problem from a human perspective and 60% from an aggregate level using optimization techniques. We found considerable potential for performance gains, with optimization techniques often decreasing waiting times and increasing success rate by as much as 50%. However, there was a lack of attention to responsible innovation: only around one third of the papers considered ethical issues in choosing the optimization goals while just a very few of them paid attention to the bias issues. Our work can serve as a guide for policy makers and researchers wanting to use an algorithm for addressing a resource allocation problem.

cs.CY↗

arXiv2023

DoDo Learning: DOmain-DemOgraphic Transfer in Language Models for Detecting Abuse Targeted at Public Figures

Public figures receive a disproportionate amount of abuse on social media, impacting their active participation in public life. Automated systems can identify abuse at scale but labelling training data is expensive, complex and potentially harmful. So, it is desirable that systems are efficient and generalisable, handling both shared and specific aspects of online abuse. We explore the dynamics of cross-group text classification in order to understand how well classifiers trained on one domain or demographic can transfer to others, with a view to building more generalisable abuse classifiers. We fine-tune language models to classify tweets targeted at public figures across DOmains (sport and politics) and DemOgraphics (women and men) using our novel DODO dataset, containing 28,000 labelled entries, split equally across four domain-demographic pairs. We find that (i) small amounts of diverse data are hugely beneficial to generalisation and model adaptation; (ii) models transfer more easily across demographics but models trained on cross-domain data are more generalisable; (iii) some groups contribute more to generalisability than others; and (iv) dataset similarity is a signal of transferability.

cs.CL↗