On social networks, algorithmic personalization drives users into filter bubbles where they rarely see content that deviates from their interests. We present a model for content curation and personalization that avoids filter bubbles, along with algorithmic guarantees and nearly matching lower bounds. In our model, the platform interacts with $n$ users over $T$ timesteps, choosing content for each user from $k$ categories. The platform receives stochastic rewards as in a multi-arm bandit. To avoid filter bubbles, we draw on the intuition that if some users are shown some category of content, then all users should see at least a small amount of that content. We first analyze a naive formalization of this intuition and show it has unintended consequences: it leads to ``tyranny of the majority'' with the burden of diversification borne disproportionately by those with minority interests. This leads us to our model which distributes this burden more equitably. We require that the probability any user is shown a particular type of content is at least $\gamma$ times the average probability all users are shown that type of content. Full personalization corresponds to $\gamma = 0$ and complete homogenization corresponds to $\gamma = 1$; hence, $\gamma$ encodes a hard cap on the level of personalization. We also analyze additional formulations where the platform can exceed its cap but pays a penalty proportional to its constraint violation. We provide algorithmic guarantees for optimizing recommendations subject to these constraints. These include nearly matching upper and lower bounds for the entire range of $\gamma \in [0,1]$ showing that the reward of a multi-agent variant of UCB is nearly optimal. Using real-world preference data, we empirically verify that under our model, users share the burden of diversification with only minor utility loss under our constraints.

Contrast consistency, the ability of a model to make consistently correct predictions in the presence of perturbations, is an essential aspect in NLP. While studied in tasks such as sentiment analysis and reading comprehension, it remains unexplored in open-domain question answering (OpenQA) due to the difficulty of collecting perturbed questions that satisfy factuality requirements. In this work, we collect minimally edited questions as challenging contrast sets to evaluate OpenQA models. Our collection approach combines both human annotation and large language model generation. We find that the widely used dense passage retriever (DPR) performs poorly on our contrast sets, despite fitting the training set well and performing competitively on standard test sets. To address this issue, we introduce a simple and effective query-side contrastive loss with the aid of data augmentation to improve DPR training. Our experiments on the contrast sets demonstrate that DPR's contrast consistency is improved without sacrificing its accuracy on the standard test sets.

Societal biases in the usage of words, including harmful stereotypes, are frequently learned by common word embedding methods. These biases manifest not only between a word and an explicit marker of its stereotype, but also between words that share related stereotypes. This latter phenomenon, sometimes called "indirect bias,'' has resisted prior attempts at debiasing. In this paper, we propose a novel method called Biased Indirect Relationship Modification (BIRM) to mitigate indirect bias in distributional word embeddings by modifying biased relationships between words before embeddings are learned. This is done by considering how the co-occurrence probability of a given pair of words changes in the presence of words marking an attribute of bias, and using this to average out the effect of a bias attribute. To evaluate this method, we perform a series of common tests and demonstrate that measures of bias in the word embeddings are reduced in exchange for minor reduction in the semantic quality of the embeddings. In addition, we conduct novel tests for measuring indirect stereotypes by extending the Word Embedding Association Test (WEAT) with new test sets for indirect binary gender stereotypes. With these tests, we demonstrate the presence of more subtle stereotypes not addressed by previous work. The proposed method is able to reduce the presence of some of these new stereotypes, serving as a crucial next step towards non-stereotyped word embeddings.

In July 2022, the Center for Security and Emerging Technology (CSET) at Georgetown University and the Program on Geopolitics, Technology, and Governance at the Stanford Cyber Policy Center convened a workshop of experts to examine the relationship between vulnerabilities in artificial intelligence systems and more traditional types of software vulnerabilities. Topics discussed included the extent to which AI vulnerabilities can be handled under standard cybersecurity processes, the barriers currently preventing the accurate sharing of information about AI vulnerabilities, legal issues associated with adversarial attacks on AI systems, and potential areas where government support could improve AI vulnerability management and mitigation. Attendees at the workshop included industry representatives in both cybersecurity and AI red-teaming roles; academics with experience conducting adversarial machine learning research; legal specialists in cybersecurity regulation, AI liability, and computer-related criminal law; and government representatives with significant AI oversight responsibilities. This report is meant to accomplish two things. First, it provides a high-level discussion of AI vulnerabilities, including the ways in which they are disanalogous to other types of vulnerabilities, and the current state of affairs regarding information sharing and legal oversight of AI vulnerabilities. Second, it attempts to articulate broad recommendations as endorsed by the majority of participants at the workshop. These recommendations, categorized under four high-level topics, are as follows: 1. Topic: Extending Traditional Cybersecurity for AI Vulnerabilities 1.1. Recommendation: Organizations building or deploying AI models should use a risk management framework that addresses security throughout the AI system life cycle.

Representing texts as information about entities has long been deemed effective in event reasoning. We propose OpenPI2.0, an improved dataset for tracking entity states in procedural texts. OpenPI2.0 features not only canonicalized entities that facilitate evaluation, but also salience annotations including both manual labels and automatic predictions. Regarding entity salience, we provide a survey on annotation subjectivity, modeling feasibility, and downstream applications in tasks such as question answering and classical planning.

Over the past decade explainable artificial intelligence has evolved from a predominantly technical discipline into a field that is deeply intertwined with social sciences. Insights such as human preference for contrastive -- more precisely, counterfactual -- explanations have played a major role in this transition, inspiring and guiding the research in computer science. Other observations, while equally important, have received much less attention. The desire of human explainees to communicate with artificial intelligence explainers through a dialogue-like interaction has been mostly neglected by the community. This poses many challenges for the effectiveness and widespread adoption of such technologies as delivering a single explanation optimised according to some predefined objectives may fail to engender understanding in its recipients and satisfy their unique needs given the diversity of human knowledge and intention. Using insights elaborated by Niklas Luhmann and, more recently, Elena Esposito we apply social systems theory to highlight challenges in explainable artificial intelligence and offer a path forward, striving to reinvigorate the technical research in this direction. This paper aims to demonstrate the potential of systems theoretical approaches to communication in understanding problems and limitations of explainable artificial intelligence.

Propaganda is a form of communication intended to influence the opinions and the mindset of the public to promote a particular agenda. With the rise of social media, propaganda has spread rapidly, leading to the need for automatic propaganda detection systems. Most work on propaganda detection has focused on high-resource languages, such as English, and little effort has been made to detect propaganda for low-resource languages. Yet, it is common to find a mix of multiple languages in social media communication, a phenomenon known as code-switching. Code-switching combines different languages within the same text, which poses a challenge for automatic systems. With this in mind, here we propose the novel task of detecting propaganda techniques in code-switched text. To support this task, we create a corpus of 1,030 texts code-switching between English and Roman Urdu, annotated with 20 propaganda techniques, which we make publicly available. We perform a number of experiments contrasting different experimental setups, and we find that it is important to model the multilinguality directly (rather than using translation) as well as to use the right fine-tuning strategy. The code and the dataset are publicly available at https://github.com/mbzuai-nlp/propaganda-codeswitched-text

Neural networks are deployed widely in natural language processing tasks on the industrial scale, and perhaps the most often they are used as compounds of automatic machine translation systems. In this work, we present a simple approach to fool state-of-the-art machine translation tools in the task of translation from Russian to English and vice versa. Using a novel black-box gradient-free tensor-based optimizer, we show that many online translation tools, such as Google, DeepL, and Yandex, may both produce wrong or offensive translations for nonsensical adversarial input queries and refuse to translate seemingly benign input phrases. This vulnerability may interfere with understanding a new language and simply worsen the user's experience while using machine translation systems, and, hence, additional improvements of these tools are required to establish better translation.

Multi-vector retrieval models such as ColBERT [Khattab and Zaharia, 2020] allow token-level interactions between queries and documents, and hence achieve state of the art on many information retrieval benchmarks. However, their non-linear scoring function cannot be scaled to millions of documents, necessitating a three-stage process for inference: retrieving initial candidates via token retrieval, accessing all token vectors, and scoring the initial candidate documents. The non-linear scoring function is applied over all token vectors of each candidate document, making the inference process complicated and slow. In this paper, we aim to simplify the multi-vector retrieval by rethinking the role of token retrieval. We present XTR, ConteXtualized Token Retriever, which introduces a simple, yet novel, objective function that encourages the model to retrieve the most important document tokens first. The improvement to token retrieval allows XTR to rank candidates only using the retrieved tokens rather than all tokens in the document, and enables a newly designed scoring stage that is two-to-three orders of magnitude cheaper than that of ColBERT. On the popular BEIR benchmark, XTR advances the state-of-the-art by 2.8 nDCG@10 without any distillation. Detailed analysis confirms our decision to revisit the token retrieval stage, as XTR demonstrates much better recall of the token retrieval stage compared to ColBERT.

Automated Intelligent Cyberdefense Agents (AICAs) that are part Intrusion Detection Systems (IDS) and part Intrusion Response Systems (IRS) are being designed to protect against sophisticated and automated cyber-attacks. An AICA based on the ideas of Self-Adaptive Autonomic Computing Systems (SA-ACS) can be considered as a managing system that protects a managed system like a personal computer, web application, critical infrastructure, etc. An AICA, specifically the IRS components, can compute a wide range of potential responses to meet its security goals and objectives, such as taking actions to prevent the attack from completing, restoring the system to comply with the organizational security policy, containing or confining an attack, attack eradication, deploying forensics measures to enable future attack analysis, counterattack, and so on. To restrict its activities in order to minimize collateral/organizational damage, such an automated system must have set Rules of Engagement (RoE). Automated systems must determine which operations can be completely automated (and when), which actions require human operator confirmation, and which actions must never be undertaken. In this paper, to enable this control functionality over an IRS, we create Rules of EngaGement for Automated cybeR Defense (REGARD) system which holds a set of Rules of Engagement (RoE) to protect the managed system according to the instructions provided by the human operator. These rules help limit the action of the IRS on the managed system in compliance with the recommendations of the domain expert. We provide details of execution, management, operation, and conflict resolution for Rules of Engagement (RoE) to constrain the actions of an automated IRS. We also describe REGARD system implementation, security case studies for cyber defense, and RoE demonstrations.