Large language models (LLMs) have been shown to propagate and amplify harmful stereotypes, particularly those that disproportionately affect marginalised communities.To understand the effect of these stereotypes more comprehensively, we introduce GlobalBias, a dataset of 876k sentences incorporating 40 distinct gender-by-ethnicity groups alongside descriptors typically used in bias literature, which enables us to study a broad set of stereotypes from around the world.We use GlobalBias to directly probe a suite of LMs via perplexity, which we use as a proxy to determine how certain stereotypes are represented in the model's internal representations.Following this, we generate character profiles based on given names and evaluate the prevalence of stereotypes in model outputs.We find that the demographic groups associated with various stereotypes remain consistent across model likelihoods and model outputs.Furthermore, larger models consistently display higher levels of stereotypical outputs, even when explicitly instructed not to.

The global aging population is growing, leading to an increased need for unpaid carers to support older adults. Socially Assistive Robots (SARs) can play an important role in supporting unpaid carers. However, as with any socio-technical system, building and maintaining trust is critical for achieving full potential and adoption of a SAR. Despite the growing interest in robotics across various domains, there has been limited research on time-dependant patterns of trust during Human Robot Interactions (HRI) and how unpaid carers perceive SARs as a support tool for care. We present a study in which we a) investigate the level of trust that unpaid carers have in using SARs for taking care of older adults, b) qualitatively explore the factors that influence trust considering the dynamic and context dependant nature of care. To explore SARs as a support tool from the perspective of unpaid carers we recruited 15 caregivers, we conducted co-design sessions, interviews and carers interacted with a Pepper robot evaluating trust before and after interaction. We discuss how the level of trust of unpaid carers in SARs increases after interaction and how unpaid carers trust in SARs is impacted by socially and culturally intelligent robots. Our work contributes to identify factors influencing trust that are important for carer’s while interacting with SARs, offering insights for the development and implementation of future SAR-based as a support tool that can assist them in taking care of older adults.

The push to a connected world where people carry an always-online device which has been designed to maximise instant gratification and prompts users via notifications has lead to a surge of potentially problematic behaviour as a result. This has lead to a rising interest in addressing and understanding the addictiveness of smartphone usage, as well as for particular applications (apps). However, capturing addiction from usage involves not only assessment of potential addiction risk but also requires understanding of the complex interactions that define user behaviour and how these can be effectively isolated and summarised. In this paper, we examine the correlation of physical user interface (UI) interactions (e.g. taps and scrolls) and smartphone addiction risk using a large dataset of those smartphone events (65,093,343, N=301,024 sessions) collected from 64 users over an 8-week period with an accompanying smartphone addiction survey. Our novel method which reports on the probability of a users addiction risk and in a model case we show how it was be used to identify 57 of 64 users correctly. This supports our observations of UI events during sessions of usage being indicative of addiction risk while improving previous approaches which rely on summative data such as screen on time. Within this we also find that users only exhibit addictive behaviour in a subset of all sessions while using their smartphone.

Cognitive dissonance is well-understood as a significant psychological motivator of behavior. It can be experienced vicariously when a member of one’s social group acts inconsistently to expectations. In this article, we explore the network implications from individuals reconciling cognitive friction when their neighbors hold alternative views. Through agent-based modeling, we introduce a framework to explore the sensitivity of behavior on social network structure, in response to vicarious dissonance. The model allows us to understand how and why vicarious dissonance may contribute to polarization, both in terms of network structure and the convictions held by individuals. Alternative response behaviors are each found to be highly effective in reducing the cognitive dissonance felt across a population but with wide-ranging outcomes for the population as a whole. The results highlight the important role of neutrality and tolerance in retaining social cohesion while showing how easily this can be disrupted. The model presents a useful tool for further research, allowing bespoke scenarios to be investigated.

The prevalence of induced triads play an important role in characterising complex networks, supporting approaches for assessment of dynamic and partially obfuscated scenarios. In this paper we introduce a new local edge-centrality measure that is designed to be deployed in this context for complex networks and is highly scalable. It signifies the importance an edge plays within induced triads for a directed network. We observe that an edge can play one of two roles in providing connectivity within any particular triad, based on whether the edge supports connectivity to the third node or not. We call these alternative states overt and covert. As an edge may play alternative roles in different induced triads, this allows us to assess the local importance of an edge across multiple induced substructures. We introduce theory to count the number of induced triads in which an edge is overt and covert. Using 34 data sets derived from public sources, we show how the presence of overt and covert edges can be used to profile diverse real-world networks. The relationship with global network analysis metrics is examined. We observe that overt and covert edge centrality is useful in further differentiating classes of network, when considered in combination with conventional global network analysis metrics.

Trolling, echo chambers and general suspicious behaviour online are a serious cause of concern due to their potential disruptive effects beyond social media. This motivates a better understanding of the characteristics of disruptive behaviour on the internet and methods of detection. In this work we focus on Reddit which provides a rich social media platform for community focused interactions. Using network representations of user activity alongside temporal statistics and other features we assess the behaviour of a sample of potentially disruptive users, based on their assigned comment karma (an aggregate of a user's comment up-votes), relative to the wider population. We explore how these signals contribute to the accurate prediction of disruptive users, and note that this is achieved without requiring any semantic analysis. Our results show that it is possible to detect signs of disruptive behaviour with good accuracy using limited inputs that are primarily based on the reply patterns that users generate. This is of potential value for large-scale detection problems and operation across different languages.

Learning network representation has a variety of \napplications, such as network classification. Most existing work \nin this area focuses on static undirected networks and does not \naccount for presence of directed edges or temporal changes. \nFurthermore, most work focuses on node representations that \ndo poorly on tasks like network classification. In this paper, \nwe propose a novel network embedding methodology, gl2vec, \nfor network classification in both static and temporal directed \nnetworks. gl2vec constructs vectors for feature representation \nusing static or temporal network graphlet distributions and a \nnull model for comparing them against random graphs. We \ndemonstrate the efficacy and usability of gl2vec over existing \nstate-of-the-art methods on network classification tasks such as \nnetwork type classification and subgraph identification in several \nreal-world static and temporal directed networks. We argue that \ngl2vec provides additional network features that are not captured \nby state-of-the-art methods, which can significantly improve their \nclassification accuracy by up to 10% in real-world applications

When should a machine attempt to communicate with a user? This is a historical problem that has been studied since the rise of personal computing. More recently, the emergence of pervasive technologies such as the smartphone have extended the problem to be ever-present in our daily lives, opening up new opportunities for context awareness through data collection and reasoning. Complementary to this there has been increasing interest in techniques to intelligently synchronise interruptions with human behaviour and cognition. However, it is increasingly challenging to categorise new developments, which are often scenario specific or scope a problem with particular unique features. In this paper we present a meta-analysis of this area, decomposing and comparing historical and recent works that seek to understand and predict how users will perceive and respond to interruptions. In doing so we identify research gaps, questions and opportunities that characterise this important emerging field for pervasive technology.