Talal Rahwan

Bedoor AlShebli, Shahan Ali Memon, James A. Evans et al. · cmu

Artificial Intelligence (AI) has become a disruptive technology, promising to grant a significant economic and strategic advantage to nations that harness its power. China, with its recent push towards AI adoption, is challenging the U.S.'s position as the global leader in this field. Given AI's massive potential, as well as the fierce geopolitical tensions between China and the U.S., several recent policies have been put in place to discourage AI scientists from migrating to, or collaborating with, the other nation. Nevertheless, the extent of talent migration and cross-border collaboration are not fully understood. Here, we analyze a dataset of over 350,000 AI scientists and 5,000,000 AI papers. We find that since 2000, China and the U.S. have led the field in terms of impact, novelty, productivity, and workforce. Most AI scientists who move to China come from the U.S., and most who move to the U.S. come from China, highlighting a notable bidirectional talent migration. Moreover, the vast majority of those moving in either direction have Asian ancestry. Upon moving, those scientists continue to collaborate frequently with those in the origin country. Although the number of collaborations between the two countries has increased since the dawn of the millennium, such collaborations continue to be relatively rare. A matching experiment reveals that the two countries have always been more impactful when collaborating than when each works without the other. These findings suggest that instead of suppressing cross-border migration and collaboration between the two nations, the science could benefit from promoting such activities.

Hazem Ibrahim, Fengyuan Liu, Khalid Mengal et al.

Scientific progress depends on researchers' ability to access and build upon the work of others. Yet, much published work remains behind expensive paywalls, and even accessible articles often rest on datasets shared only "upon reasonable request" to the authors. Researchers can try to overcome these barriers through informal channels, such as emailing authors directly, but whether such channels are hindered by racial or institutional biases remains unknown. Here we combine survey data, semi-structured interviews, large-scale observational analysis, and two randomized audit experiments to examine disparities in access to scientific knowledge. Surveyed researchers in the Global South report markedly lower institutional access to the literature and depend more heavily on informal channels to obtain papers and data; interviews elaborate the workarounds and racialized frictions they encounter. Our analysis of 250 million articles reveals that Global South researchers cite paywalled papers at significantly lower rates than Global North counterparts--a gap associated with reduced knowledge breadth and scholarly impact. Using citation-context classification, we further find that papers whose data is available only upon request are less likely to be cited for reusing their data, a penalty falling disproportionately on the Global South. To probe mechanisms, we conduct two email audit studies in which fictional PhD students differing in racial background and institutional affiliation request paywalled articles (N = 18,000) and datasets (N = 16,000). Racial identity influences response rates to both requests, whereas institutional affiliation influences access to datasets. These findings reveal how informal gatekeeping can perpetuate structural inequities in science, highlighting the need for stronger data-sharing mandates and more equitable open-access policies.

Yuni Lai, Marcin Waniek, Liying Li et al.

Random Walks-based Anomaly Detection (RWAD) is commonly used to identify anomalous patterns in various applications. An intriguing characteristic of RWAD is that the input graph can either be pre-existing or constructed from raw features. Consequently, there are two potential attack surfaces against RWAD: graph-space attacks and feature-space attacks. In this paper, we explore this vulnerability by designing practical coupled-space attacks, investigating the interplay between graph-space and feature-space attacks. To this end, we conduct a thorough complexity analysis, proving that attacking RWAD is NP-hard. Then, we proceed to formulate the graph-space attack as a bi-level optimization problem and propose two strategies to solve it: alternative iteration (alterI-attack) or utilizing the closed-form solution of the random walk model (cf-attack). Finally, we utilize the results from the graph-space attacks as guidance to design more powerful feature-space attacks (i.e., graph-guided attacks). Comprehensive experiments demonstrate that our proposed attacks are effective in enabling the target nodes from RWAD with a limited attack budget. In addition, we conduct transfer attack experiments in a black-box setting, which show that our feature attack significantly decreases the anomaly scores of target nodes. Our study opens the door to studying the coupled-space attack against graph anomaly detection in which the graph space relies on the feature space.

Zhen Wang, Ruiqi Song, Chen Shen et al.

Despite rapid technological progress, effective human-machine cooperation remains a significant challenge. Humans tend to cooperate less with machines than with fellow humans, a phenomenon known as the machine penalty. Here, we show that artificial intelligence (AI) agents powered by large language models can overcome this penalty in social dilemma games with communication. In a pre-registered experiment with 1,152 participants, we deploy AI agents exhibiting three distinct personas: selfish, cooperative, and fair. However, only fair agents elicit human cooperation at rates comparable to human-human interactions. Analysis reveals that fair agents, similar to human participants, occasionally break pre-game cooperation promises, but nonetheless effectively establish cooperation as a social norm. These results challenge the conventional wisdom of machines as altruistic assistants or rational actors. Instead, our study highlights the importance of AI agents reflecting the nuanced complexity of human social behaviors -- imperfect yet driven by deeper social cognitive processes.

Marcin Waniek, Navya Suri, Abdullah Zameek et al.

Attribute inference - the process of analyzing publicly available data in order to uncover hidden information - has become a major threat to privacy, given the recent technological leap in machine learning. One way to tackle this threat is to strategically modify one's publicly available data in order to keep one's private information hidden from attribute inference. We evaluate people's ability to perform this task, and compare it against algorithms designed for this purpose. We focus on three attributes: the gender of the author of a piece of text, the country in which a set of photos was taken, and the link missing from a social network. For each of these attributes, we find that people's effectiveness is inferior to that of AI, especially when it comes to hiding the attribute in question. Moreover, when people are asked to modify the publicly available information in order to hide these attributes, they are less likely to make high-impact modifications compared to AI. This suggests that people are unable to recognize the aspects of the data that are critical to an inference algorithm. Taken together, our findings highlight the limitations of relying on human intuition to protect privacy in the age of AI, and emphasize the need for algorithmic support to protect private information from attribute inference.

Fengyuan Liu, Nouar AlDahoul, Gregory Eady et al.

Previous studies proposed that the reasoning capabilities of large language models (LLMs) can be improved through self-reflection, i.e., letting LLMs reflect on their own output to identify and correct mistakes in the initial responses. However, earlier experiments offer mixed results when it comes to the benefits of self-reflection. Furthermore, prior studies on self-reflection are predominantly concerned with the reasoning capabilities of models, ignoring the potential for self-reflection in safety, bias, and ideological leaning. Here, by conducting a series of experiments testing LLM's self-reflection capability in various tasks using a variety of prompts and different LLMs, we make several contributions to the literature. First, we reconcile conflicting findings regarding the benefit of self-reflection, by demonstrating that the outcome of self-reflection is sensitive to prompt wording -- both the original prompt that are used to elicit an initial answer and the subsequent prompt used to self-reflect. Specifically, although self-reflection may improve the reasoning capability of LLMs when the initial response is simple, the technique cannot improve upon the state-of-the-art chain-of-thought (CoT) prompting. Second, we show that self-reflection can lead to safer (75.8\% reduction in toxic responses while preserving 97.8\% non-toxic ones), less biased (77\% reduction in gender biased responses, while preserving 94.3\% unbiased ones), and more ideologically neutral responses (100\% reduction in partisan leaning response, while preserving 87.7\% non-partisan ones). The paper concludes by discussing the implications of our findings on the deployment of large language models. We release our experiments at https://github.com/Michael98Liu/self-reflection.

Hazem Ibrahim, Nouar AlDahoul, Syed Mustafa Ali Abbasi et al.

The manner in which different racial and gender groups are portrayed in news coverage plays a large role in shaping public opinion. As such, understanding how such groups are portrayed in news media is of notable societal value, and has thus been a significant endeavour in both the computer and social sciences. Yet, the literature still lacks a longitudinal study examining both the frequency of appearance of different racial and gender groups in online news articles, as well as the context in which such groups are discussed. To fill this gap, we propose two machine learning classifiers to detect the race and age of a given subject. Next, we compile a dataset of 123,337 images and 441,321 online news articles from New York Times (NYT) and Fox News (Fox), and examine representation through two computational approaches. Firstly, we examine the frequency and prominence of appearance of racial and gender groups in images embedded in news articles, revealing that racial and gender minorities are largely under-represented, and when they do appear, they are featured less prominently compared to majority groups. Furthermore, we find that NYT largely features more images of racial minority groups compared to Fox. Secondly, we examine both the frequency and context with which racial minority groups are presented in article text. This reveals the narrow scope in which certain racial groups are covered and the frequency with which different groups are presented as victims and/or perpetrators in a given conflict. Taken together, our analysis contributes to the literature by providing two novel open-source classifiers to detect race and age from images, and shedding light on the racial and gender biases in news articles from venues on opposite ends of the American political spectrum.

Nouar AlDahoul, Talal Rahwan, Yasir Zaki

Text-to-image generative AI models such as Stable Diffusion are used daily by millions worldwide. However, the extent to which these models exhibit racial and gender stereotypes is not yet fully understood. Here, we document significant biases in Stable Diffusion across six races, two genders, 32 professions, and eight attributes. Additionally, we examine the degree to which Stable Diffusion depicts individuals of the same race as being similar to one another. This analysis reveals significant racial homogenization, e.g., depicting nearly all Middle Eastern men as bearded, brown-skinned, and wearing traditional attire. We then propose debiasing solutions that allow users to specify the desired distributions of race and gender when generating images while minimizing racial homogenization. Finally, using a preregistered survey experiment, we find evidence that being presented with inclusive AI-generated faces reduces people's racial and gender biases, while being presented with non-inclusive ones increases such biases, regardless of whether the images are labeled as AI-generated. Taken together, our findings emphasize the need to address biases and stereotypes in text-to-image models.

Nouar Aldahoul, Hazem Ibrahim, Matteo Varvello et al.

Large Language Models (LLMs) are a transformational technology, fundamentally changing how people obtain information and interact with the world. As people become increasingly reliant on them for an enormous variety of tasks, a body of academic research has developed to examine these models for inherent biases, especially political biases, often finding them small. We challenge this prevailing wisdom. First, by comparing 31 LLMs to legislators, judges, and a nationally representative sample of U.S. voters, we show that LLMs' apparently small overall partisan preference is the net result of offsetting extreme views on specific topics, much like moderate voters. Second, in a randomized experiment, we show that LLMs can promulgate their preferences into political persuasiveness even in information-seeking contexts: voters randomized to discuss political issues with an LLM chatbot are as much as 5 percentage points more likely to express the same preferences as that chatbot. Contrary to expectations, these persuasive effects are not moderated by familiarity with LLMs, news consumption, or interest in politics. LLMs, especially those controlled by private companies or governments, may become a powerful and targeted vector for political influence.

Nouar AlDahoul, Talal Rahwan, Yasir Zaki

A quarter of US adults regularly get their news from YouTube. Yet, despite the massive political content available on the platform, to date no classifier has been proposed to identify the political leaning of YouTube videos. To fill this gap, we propose a novel classifier based on Bert -- a language model from Google -- to classify YouTube videos merely based on their titles into six categories, namely: Far Left, Left, Center, Anti-Woke, Right, and Far Right. We used a public dataset of 10 million YouTube video titles (under various categories) to train and validate the proposed classifier. We compare the classifier against several alternatives that we trained on the same dataset, revealing that our classifier achieves the highest accuracy (75%) and the highest F1 score (77%). To further validate the classification performance, we collect videos from YouTube channels of numerous prominent news agencies, such as Fox News and New York Times, which have widely known political leanings, and apply our classifier to their video titles. For the vast majority of cases, the predicted political leaning matches that of the news agency.

Chen Wei Kuo, Kevin Chu, Nouar AlDahoul et al.

Bias in news reporting significantly impacts public perception, particularly regarding crime, politics, and societal issues. Traditional bias detection methods, predominantly reliant on human moderation, suffer from subjective interpretations and scalability constraints. Here, we introduce an AI-driven framework leveraging advanced large language models (LLMs), specifically GPT-4o, GPT-4o Mini, Gemini Pro, Gemini Flash, Llama 8B, and Llama 3B, to systematically identify and mitigate biases in news articles. To this end, we collect an extensive dataset consisting of over 30,000 crime-related articles from five politically diverse news sources spanning a decade (2013-2023). Our approach employs a two-stage methodology: (1) bias detection, where each LLM scores and justifies biased content at the paragraph level, validated through human evaluation for ground truth establishment, and (2) iterative debiasing using GPT-4o Mini, verified by both automated reassessment and human reviewers. Empirical results indicate GPT-4o Mini's superior accuracy in bias detection and effectiveness in debiasing. Furthermore, our analysis reveals temporal and geographical variations in media bias correlating with socio-political dynamics and real-world events. This study contributes to scalable computational methodologies for bias mitigation, promoting fairness and accountability in news reporting.

Sidahmed Benabderrahmane, Talal Rahwan

Advanced Persistent Threats (APT) pose a major cybersecurity challenge due to their stealth, persistence, and adaptability. Traditional machine learning detectors struggle with class imbalance, high dimensional features, and scarce real world traces. They often lack transferability-performing well in the training domain but degrading in novel attack scenarios. We propose a hybrid transfer framework that integrates Transfer Learning, Explainable AI (XAI), contrastive learning, and Siamese networks to improve cross-domain generalization. An attention-based autoencoder supports knowledge transfer across domains, while Shapley Additive exPlanations (SHAP) select stable, informative features to reduce dimensionality and computational cost. A Siamese encoder trained with a contrastive objective aligns source and target representations, increasing anomaly separability and mitigating feature drift. We evaluate on real-world traces from the DARPA Transparent Computing (TC) program and augment with synthetic attack scenarios to test robustness. Across source to target transfers, the approach delivers improved detection scores with classical and deep baselines, demonstrating a scalable, explainable, and transferable solution for APT detection.

Sidahmed Benabderrahmane, Talal Rahwan

Advanced Persistent Threats (APTs) present a considerable challenge to cybersecurity due to their stealthy, long-duration nature. Traditional supervised learning methods typically require large amounts of labeled data, which is often scarce in real-world scenarios. This paper introduces a novel approach that combines AutoEncoders for anomaly detection with active learning to iteratively enhance APT detection. By selectively querying an oracle for labels on uncertain or ambiguous samples, our method reduces labeling costs while improving detection accuracy, enabling the model to effectively learn with minimal data and reduce reliance on extensive manual labeling. We present a comprehensive formulation of the Attention Adversarial Dual AutoEncoder-based anomaly detection framework and demonstrate how the active learning loop progressively enhances the model's performance. The framework is evaluated on real-world, imbalanced provenance trace data from the DARPA Transparent Computing program, where APT-like attacks account for just 0.004\% of the data. The datasets, which cover multiple operating systems including Android, Linux, BSD, and Windows, are tested in two attack scenarios. The results show substantial improvements in detection rates during active learning, outperforming existing methods.

Sidahmed Benabderrahmane, Ngoc Hoang, Petko Valtchev et al.

Advanced Persistent Threats (APTs) are sophisticated, targeted cyberattacks designed to gain unauthorized access to systems and remain undetected for extended periods. To evade detection, APT cyberattacks deceive defense layers with breaches and exploits, thereby complicating exposure by traditional anomaly detection-based security methods. The challenge of detecting APTs with machine learning is compounded by the rarity of relevant datasets and the significant imbalance in the data, which makes the detection process highly burdensome. We present AE-APT, a deep learning-based tool for APT detection that features a family of AutoEncoder methods ranging from a basic one to a Transformer-based one. We evaluated our tool on a suite of provenance trace databases produced by the DARPA Transparent Computing program, where APT-like attacks constitute as little as 0.004% of the data. The datasets span multiple operating systems, including Android, Linux, BSD, and Windows, and cover two attack scenarios. The outcomes showed that AE-APT has significantly higher detection rates compared to its competitors, indicating superior performance in detecting and ranking anomalies.

Christoforos Vasilatos, Manaar Alam, Talal Rahwan et al.

As the use of Large Language Models (LLMs) in text generation tasks proliferates, concerns arise over their potential to compromise academic integrity. The education sector currently tussles with distinguishing student-authored homework assignments from AI-generated ones. This paper addresses the challenge by introducing HowkGPT, designed to identify homework assignments generated by AI. HowkGPT is built upon a dataset of academic assignments and accompanying metadata [17] and employs a pretrained LLM to compute perplexity scores for student-authored and ChatGPT-generated responses. These scores then assist in establishing a threshold for discerning the origin of a submitted assignment. Given the specificity and contextual nature of academic work, HowkGPT further refines its analysis by defining category-specific thresholds derived from the metadata, enhancing the precision of the detection. This study emphasizes the critical need for effective strategies to uphold academic integrity amidst the growing influence of LLMs and provides an approach to ensuring fair and accurate grading in educational institutions.

Hazem Ibrahim, Fengyuan Liu, Rohail Asim et al.

The emergence of large language models has led to the development of powerful tools such as ChatGPT that can produce text indistinguishable from human-generated work. With the increasing accessibility of such technology, students across the globe may utilize it to help with their school work -- a possibility that has sparked discussions on the integrity of student evaluations in the age of artificial intelligence (AI). To date, it is unclear how such tools perform compared to students on university-level courses. Further, students' perspectives regarding the use of such tools, and educators' perspectives on treating their use as plagiarism, remain unknown. Here, we compare the performance of ChatGPT against students on 32 university-level courses. We also assess the degree to which its use can be detected by two classifiers designed specifically for this purpose. Additionally, we conduct a survey across five countries, as well as a more in-depth survey at the authors' institution, to discern students' and educators' perceptions of ChatGPT's use. We find that ChatGPT's performance is comparable, if not superior, to that of students in many courses. Moreover, current AI-text classifiers cannot reliably detect ChatGPT's use in school work, due to their propensity to classify human-written answers as AI-generated, as well as the ease with which AI-generated text can be edited to evade detection. Finally, we find an emerging consensus among students to use the tool, and among educators to treat this as plagiarism. Our findings offer insights that could guide policy discussions addressing the integration of AI into educational frameworks.

Kai Zhou, Tomasz P. Michalak, Talal Rahwan et al.

Link prediction is one of the fundamental problems in computational social science. A particularly common means to predict existence of unobserved links is via structural similarity metrics, such as the number of common neighbors; node pairs with higher similarity are thus deemed more likely to be linked. However, a number of applications of link prediction, such as predicting links in gang or terrorist networks, are adversarial, with another party incentivized to minimize its effectiveness by manipulating observed information about the network. We offer a comprehensive algorithmic investigation of the problem of attacking similarity-based link prediction through link deletion, focusing on two broad classes of such approaches, one which uses only local information about target links, and another which uses global network information. While we show several variations of the general problem to be NP-Hard for both local and global metrics, we exhibit a number of well-motivated special cases which are tractable. Additionally, we provide principled and empirically effective algorithms for the intractable cases, in some cases proving worst-case approximation guarantees.

Marcin Waniek, Kai Zhou, Yevgeniy Vorobeychik et al.

Link prediction is one of the fundamental research problems in network analysis. Intuitively, it involves identifying the edges that are most likely to be added to a given network, or the edges that appear to be missing from the network when in fact they are present. Various algorithms have been proposed to solve this problem over the past decades. For all their benefits, such algorithms raise serious privacy concerns, as they could be used to expose a connection between two individuals who wish to keep their relationship private. With this in mind, we investigate the ability of such individuals to evade link prediction algorithms. More precisely, we study their ability to strategically alter their connections so as to increase the probability that some of their connections remain unidentified by link prediction algorithms. We formalize this question as an optimization problem, and prove that finding an optimal solution is NP-complete. Despite this hardness, we show that the situation is not bleak in practice. In particular, we propose two heuristics that can easily be applied by members of the general public on existing social media. We demonstrate the effectiveness of those heuristics on a wide variety of networks and against a plethora of link prediction algorithms.

Mateusz K. Tarkowski, Tomasz P. Michalak, Talal Rahwan et al.

Game-theoretic centrality is a flexible and sophisticated approach to identify the most important nodes in a network. It builds upon the methods from cooperative game theory and network theory. The key idea is to treat nodes as players in a cooperative game, where the value of each coalition is determined by certain graph-theoretic properties. Using solution concepts from cooperative game theory, it is then possible to measure how responsible each node is for the worth of the network. The literature on the topic is already quite large, and is scattered among game-theoretic and computer science venues. We review the main game-theoretic network centrality measures from both bodies of literature and organize them into two categories: those that are more focused on the connectivity of nodes, and those that are more focused on the synergies achieved by nodes in groups. We present and explain each centrality, with a focus on algorithms and complexity.

Talal Rahwan, Sarvapali Dyanand Ramchurn, Nicholas Robert Jennings et al.

Coalition formation is a fundamental type of interaction that involves the creation of coherent groupings of distinct, autonomous, agents in order to efficiently achieve their individual or collective goals. Forming effective coalitions is a major research challenge in the field of multi-agent systems. Central to this endeavour is the problem of determining which of the many possible coalitions to form in order to achieve some goal. This usually requires calculating a value for every possible coalition, known as the coalition value, which indicates how beneficial that coalition would be if it was formed. Once these values are calculated, the agents usually need to find a combination of coalitions, in which every agent belongs to exactly one coalition, and by which the overall outcome of the system is maximized. However, this coalition structure generation problem is extremely challenging due to the number of possible solutions that need to be examined, which grows exponentially with the number of agents involved. To date, therefore, many algorithms have been proposed to solve this problem using different techniques ranging from dynamic programming, to integer programming, to stochastic search all of which suffer from major limitations relating to execution time, solution quality, and memory requirements. With this in mind, we develop an anytime algorithm to solve the coalition structure generation problem. Specifically, the algorithm uses a novel representation of the search space, which partitions the space of possible solutions into sub-spaces such that it is possible to compute upper and lower bounds on the values of the best coalition structures in them. These bounds are then used to identify the sub-spaces that have no potential of containing the optimal solution so that they can be pruned. The algorithm, then, searches through the remaining sub-spaces very efficiently using a branch-and-bound technique to avoid examining all the solutions within the searched subspace(s). In this setting, we prove that our algorithm enumerates all coalition structures efficiently by avoiding redundant and invalid solutions automatically. Moreover, in order to effectively test our algorithm we develop a new type of input distribution which allows us to generate more reliable benchmarks compared to the input distributions previously used in the field. Given this new distribution, we show that for 27 agents our algorithm is able to find solutions that are optimal in 0.175% of the time required by the fastest available algorithm in the literature. The algorithm is anytime, and if interrupted before it would have normally terminated, it can still provide a solution that is guaranteed to be within a bound from the optimal one. Moreover, the guarantees we provide on the quality of the solution are significantly better than those provided by the previous state of the art algorithms designed for this purpose. For example, for the worst case distribution given 25 agents, our algorithm is able to find a 90% efficient solution in around 10% of time it takes to find the optimal solution.