Large Language Models (LLMs) are extensively used in text generation tasks. These generative capabilities bring us to a point where LLMs could potentially provide useful insights in policy making or agency operations. In this paper, we introduce a new task consisting of generating recommendations which can be used to inform future actions and improvements of agencies work within private and public organisations. In particular, we present the first benchmark and coherent evaluation for developing recommendation systems to inform organisation policies. This task is clearly different from usual product or user recommendation systems, but rather aims at providing a basis to suggest policy improvements based on the conclusions drawn from reports. Our results demonstrate that state-of-the-art LLMs have the potential to emphasize and reflect on key issues and learning points within generated recommendations.

Deep Neural Networks (DNNs) are often considered black boxes due to their opaque decision-making processes. To reduce their opacity Concept Models (CMs), such as Concept Bottleneck Models (CBMs), were introduced to predict human-defined concepts as an intermediate step before predicting task labels. This enhances the interpretability of DNNs. In a human-machine setting greater interpretability enables humans to improve their understanding and build trust in a DNN. In the introduction of CBMs, the models demonstrated increased task accuracy as incorrect concept predictions were replaced with their ground truth values, known as intervening on the concept predictions. In a collaborative setting, if the model task accuracy improves from interventions, trust in a model and the human-machine task accuracy may increase. However, the result showing an increase in model task accuracy was produced without human evaluation and thus it remains unknown if the findings can be applied in a collaborative setting. In this paper, we ran the first human studies using CBMs to evaluate their human interaction in collaborative task settings. Our findings show that CBMs improve interpretability compared to standard DNNs, leading to increased human-machine alignment. However, this increased alignment did not translate to a significant increase in task accuracy. Understanding the model's decision-making process required multiple interactions, and misalignment between the model's and human decision-making processes could undermine interpretability and model effectiveness.

In this paper, we explore how humans communicate with teammates from an explainable-AI perspective, comparing how they interact with both human and AI-controlled robot teammates under a number of different strategies in scenarios such as disaster relief. We find that while humans do adapt their communication based on which strategy they are following, they consistently communicate differently with AI teammates than human teammates, tending to give explicit orders to the former while sending more vague messages with implicit understanding to the later. However, we also find that modern Large Language Models (LLMs) are capable of understanding the explainability intent of messages to the same level as humans, implying that such differences may not be required if LLMs are fully integrated into AI agents.

Teams of embodied AI-enabled agents are critical for applications in extreme and highly dynamic environments. Developing robust controllers for such agents requires a deep understanding of the challenges encountered when attempting to coordinate and synchronize their individual perception-cognition-communication-action (PCCA) loops for team-wide mission objectives. We introduce a framework to explore the coordination of the PCCA loops across multiple agents in a new simulated physical environment designed to explore collaboration in each PCCA stage. This environment tasks teams of agents with the correct disposal of dangerous objects in an area and forces careful coordination of sensing, communication, movement, and manipulation actions by providing spatially-bounded communication, incorporating situations that require concerted effort by groups of agents, and introducing uncertainty into agents' sensing capabilities. We provide a set of heuristic controllers, an offline oracle model, and an initial exploration of a Reward Machine-based controller that learns its policies from training. Together these approaches serve to provide insights into the complexity of the multi-agent PCCA loop coordination problem. The multiagent PCCA simulation environment, which supports AI and human-controlled agents, and the code for various agent controllers are available at https://github.com/nesl/AI-Collab.

Combined, joint, intra-governmental, inter-agency and multinational (CJIIM) operations require rapid data sharing without the bottlenecks of metadata curation and alignment. Curation and alignment is particularly infeasible for external open source information (OSINF), e.g., social media, which has become increasingly valuable in understanding unfolding situations. Large language models (transformers) facilitate semantic data and metadata alignment but are inefficient in CJIIM settings characterised as denied, degraded, intermittent and low bandwidth (DDIL). Vector symbolic architectures (VSA) support semantic information processing using highly compact binary vectors, typically 1-10k bits, suitable in a DDIL setting. We demonstrate a novel integration of transformer models with VSA, combining the power of the former for semantic matching with the compactness and representational structure of the latter. The approach is illustrated via a proof-of-concept OSINF data discovery portal that allows partners in a CJIIM operation to share data sources with minimal metadata curation and low communications bandwidth. This work was carried out as a bridge between previous low technology readiness level (TRL) research and future higher-TRL technology demonstration and deployment.

Concept Bottleneck Models (CBMs) are considered inherently interpretable because they first predict a set of human-defined concepts before using these concepts to predict the output of a downstream task. For inherent interpretability to be fully realised, and ensure trust in a model's output, we need to guarantee concepts are predicted based on semantically mapped input features. For example, one might expect the pixels representing a broken bone in an image to be used for the prediction of a fracture. However, current literature indicates this is not the case, as concept predictions are often mapped to irrelevant input features. We hypothesise that this occurs when concept annotations are inaccurate or how input features should relate to concepts is unclear. In general, the effect of dataset labelling on concept representations in CBMs remains an understudied area. Therefore, in this paper, we examine how CBMs learn concepts from datasets with fine-grained concept annotations. We demonstrate that CBMs can learn concept representations with semantic mapping to input features by removing problematic concept correlations, such as two concepts always appearing together. To support our evaluation, we introduce a new synthetic image dataset based on a playing cards domain, which we hope will serve as a benchmark for future CBM research. For validation, we provide empirical evidence on a real-world dataset of chest X-rays, to demonstrate semantically meaningful concepts can be learned in real-world applications.

Effective communication is essential in collaborative tasks, so AI-equipped robots working alongside humans need to be able to explain their behaviour in order to cooperate effectively and earn trust. We analyse and classify communications among human participants collaborating to complete a simulated emergency response task. The analysis identifies messages that relate to various kinds of interactive explanations identified in the explainable AI literature. This allows us to understand what type of explanations humans expect from their teammates in such settings, and thus where AI-equipped robots most need explanation capabilities. We find that most explanation-related messages seek clarification in the decisions or actions taken. We also confirm that messages have an impact on the performance of our simulated task.

Autonomous artificial moderators can be useful to monitor social media for content that violates platform policies, but such artificial moderators can be confidently wrong about their decisions. While creating an approach that makes no mistakes is effectively impossible, being able to generate explanations for any given decision can simplify the task of detecting when the system is wrong. In this work we present LiveEvents, a neuro-symbolic agent capable of generating explanations based on which rules have lead to its decisions. We deliver these explanations via Cogni-Sketch, which provides users with an interactive visual representation, allowing them to easily understand the explanations given by the system

Social media has become extremely influential when it comes to policy making in modern societies, especially in the western world, where platforms such as Twitter allow users to follow politicians, thus making citizens more involved in political discussion. In the same vein, politicians use Twitter to express their opinions, debate among others on current topics and promote their political agendas aiming to influence voter behaviour. In this paper, we attempt to analyse tweets of politicians from three European countries and explore the virality of their tweets. Previous studies have shown that tweets conveying negative sentiment are likely to be retweeted more frequently. By utilising state-of-the-art pre-trained language models, we performed sentiment analysis on hundreds of thousands of tweets collected from members of parliament in Greece, Spain and the United Kingdom, including devolved administrations. We achieved this by systematically exploring and analysing the differences between influential and less popular tweets. Our analysis indicates that politicians' negatively charged tweets spread more widely, especially in more recent times, and highlights interesting differences between political parties as well as between politicians and the general population.

Detecting complex events from subsymbolic data streams (such as images, audio recordings or videos) is a challenging problem, as traditional symbolic approaches cannot be used to process subsymbolic data, and neural-only approaches usually require larger amounts of training data than available. In this paper, we present DeepProbCEP, a Complex Event Processing (CEP) approach designed with four objectives: (i) allowing the use of subsymbolic data as an input, (ii) retaining flexibility and modularity in the definition of complex event rules, (iii) limiting the cost of obtaining training data and (iv) being robust against adversarial conditions. DeepProbCEP archives this by using a neuro-symbolic approach, which combines the neural and symbolic approaches to allow training with sparse data. This is made possible through the injection of human knowledge. In this paper, we demonstrate that DeepProbCEP outperforms other state-of-the-art approaches when training using sparse data. We also show that DeepProbCEP is robust in different adversarial settings. Finally, DeepProbCEP’s flexibility is demonstrated by showing it can be used to process both images and audio as input.

Adopting shared data resources requires scientists to place trust in the originators of the data. When shared data is later used in the development of artificial intelligence (AI) systems or machine learning (ML) models, the trust lineage extends to the users of the system, typically practitioners in fields such as healthcare and finance. Practitioners rely on AI developers to have used relevant, trustworthy data, but may have limited insight and recourse. This article introduces a software architecture and implementation of a system based on design patterns from the field of self‐sovereign identity. Scientists can issue signed credentials attesting to qualities of their data resources. Data contributions to ML models are recorded in a bill of materials (BOM), which is stored with the model as a verifiable credential. The BOM provides a traceable record of the supply chain for an AI system, which facilitates on‐going scrutiny of the qualities of the contributing components. The verified BOM, and its linkage to certified data qualities, is used in the AI scrutineer, a web‐based tool designed to offer practitioners insight into ML model constituents and highlight any problems with adopted datasets, should they be found to have biased data or be otherwise discredited.

Concept Bottleneck Models (CBMs) first map raw input(s) to a vector of human-defined concepts, before using this vector to predict a final classification. We might therefore expect CBMs capable of predicting concepts based on distinct regions of an input. In doing so, this would support human interpretation when generating explanations of the model's outputs to visualise input features corresponding to concepts. The contribution of this paper is threefold: Firstly, we expand on existing literature by looking at relevance both from the input to the concept vector, confirming that relevance is distributed among the input features, and from the concept vector to the final classification where, for the most part, the final classification is made using concepts predicted as present. Secondly, we report a quantitative evaluation to measure the distance between the maximum input feature relevance and the ground truth location; we perform this with the techniques, Layer-wise Relevance Propagation (LRP), Integrated Gradients (IG) and a baseline gradient approach, finding LRP has a lower average distance than IG. Thirdly, we propose using the proportion of relevance as a measurement for explaining concept importance.

This work reveals an evidential signal that emerges from the uncertainty value in Evidential Deep Learning (EDL). EDL is one example of a class of uncertainty-aware deep learning approaches designed to provide confidence (or epistemic uncertainty) about the current test sample. In particular for computer vision and bidirectional encoder large language models, the `evidential signal' arising from the Dirichlet strength in EDL can, in some cases, discriminate between classes, which is particularly strong when using large language models. We hypothesise that the KL regularisation term causes EDL to couple aleatoric and epistemic uncertainty. In this paper, we empirically investigate the correlations between misclassification and evaluated uncertainty, and show that EDL's `evidential signal' is due to misclassification bias. We critically evaluate EDL with other Dirichlet-based approaches, namely Generative Evidential Neural Networks (EDL-GEN) and Prior Networks, and show theoretically and empirically the differences between these loss functions. We conclude that EDL's coupling of uncertainty arises from these differences due to the use (or lack) of out-of-distribution samples during training.

Artificial Intelligence applications are increasingly making use of vector embedding techniques to achieve impressive results in many application domains. Semantic Vector Spaces (SVS’s) are constructed using semantic vector embedding techniques that learn vector representations of data across multiple domains. An important application area enabled by such techniques is the capability to represent software services and service workflows as semantic hypervectors. Previous work has shown how these hypervector representations have significant advantages over alternative schemes for decentralized service workflow construction, particularly in low communications bandwidth and energy constrained environments that are typical in multi-domain operations. SVS construction usually assumes that all the data required to construct the semantic vector space is available centrally. However, in multi-domain operations different partners may not be willing to share the training data necessary to construct a common multi-domain SVS. Hence semantic hypervectors representing similar services or workflows but constructed from different training data by different partners cannot be discovered and used. In this paper we focus on how it is possible to map semantic hypervectors between partner SVS’s, so that complementary services and workflows developed and owned by different partners can be discovered and used to achieve mission goals. The paper describes techniques for generating the required mapping that require a minimum exchange of information between the different partners; demonstrate how it is possible to do this for semantic hypervectors that use different types of encoding (eg., real valued, binary, sparse slot-encoding); and illustrates how the mapping can be implemented in various multi-domain operational settings.

Social media has become extremely influential when it comes to policy making in modern societies especially in the western world (e.g., 48% of Europeans use social media every day or almost every day). Platforms such as Twitter allow users to follow politicians, thus making citizens more involved in political discussion. In the same vein, politicians use Twitter to express their opinions, debate among others on current topics and promote their political agenda aiming to influence voter behaviour. Previous studies have shown that tweets conveying negative sentiment are likely to be retweeted more frequently. In this paper, we attempt to analyse tweets of politicians from di ff erent countries and explore whether their tweets follow the same trend. Utilising state-of-the-art pre-trained language models we performed sentiment analysis on hundreds of thousands of tweets collected from members of parliament of Greece, Spain and United Kingdom, including devolved administrations. We achieved this by systematically exploring and analysing the di ff erences between influential and less popular tweets. Our analysis indicates that politicians’ negatively charged tweets spread more widely, especially in more recent times, and highlights interesting trends in the intersection of sentiment and popularity.

Online social networks serve as a convenient way to connect, share, and promote content with others. As a result, these networks can be used with malicious intent, causing disruption and harm to public debate through the sharing of misinformation. However, automatically identifying such content through its use of natural language is a significant challenge compared to our solution which uses less computational resources, language-agnostic and without the need for complex semantic analysis. Consequently alternative and complementary approaches are highly valuable. In this paper, we assess content that has the potential for misinformation and focus on patterns of user association with online social media communities (subreddits) in the popular Reddit social media platform, and generate networks of behaviour capturing user interaction with different subreddits. We examine these networks using both global and local metrics, in particular noting the presence of induced substructures (graphlets) assessing 7,876,064 posts from 96,634 users. From subreddits identified as having potential for misinformation, we note that the associated networks have strongly defined local features relating to node degree - these are evident both from analysis of dominant graphlets and degree-related global metrics. We find that these local features support high accuracy classification of subreddits that are categorised as having the potential for misinformation. Consequently we observe that induced local substructures of high degree are fundamental metrics for subreddit classification, and support automatic detection capabilities for online misinformation independent from any particular language.

Vector Symbolic Architectures (VSA) can be used to encode complex objects, such as services and sensors, as hypervectors. Such hypervectors can be used to perform efficient distributed service discovery and workflow orchestration in communications constrained environments typical of the Internet of Things (IoT). In these environments, energy efficiency is of great importance. However, most hypervector representations use dense i.i.d element values and performing energy efficient hyperdimensional computing operations on such dense vectors is challenging. More recently, a sparse binary VSA scheme has been proposed based on a slot encoding having M slots with B bit positions per slot, in which only one bit per slot can be set. This paper shows for the first time that such sparse encoded hypervectors can be mapped into energy-efficient time-to-spike Spiking Neural Network (SNN) circuits, such that all the required VSA operations can be performed. Example VSA SNN circuits have been implemented in the Brian 2 SNN simulator, showing that all VSA binding, bundling, unbinding, and cleanup memory operations execute correctly. Based on these circuit implementations, estimates of the energy and processing time required to perform the different VSA operations on typical SNN neuromorphic devices are estimated. Recommendations for the design of future SNN neuromorphic processor hardware that can more efficiently perform VSA processing are also made.