Antonio Rago

Deniz Gorur, Antonio Rago, Francesca Toni

Argument mining (AM) is the process of automatically extracting arguments, their components and/or relations amongst arguments and components from text. As the number of platforms supporting online debate increases, the need for AM becomes ever more urgent, especially in support of downstream tasks. Relation-based AM (RbAM) is a form of AM focusing on identifying agreement (support) and disagreement (attack) relations amongst arguments. RbAM is a challenging classification task, with existing methods failing to perform satisfactorily. In this paper, we show that general-purpose Large Language Models (LLMs), appropriately primed and prompted, can significantly outperform the best performing (RoBERTa-based) baseline. Specifically, we experiment with two open-source LLMs (Llama-2 and Mistral) with ten datasets.

Junqi Jiang, Tom Bewley, Salim I. Amoukou et al.

Test-time scaling improves large language models' (LLMs) performance by allocating more compute budget during inference. To achieve this, existing methods often require intricate modifications to prompting and sampling strategies. In this work, we introduce representation consistency (RC), a test-time scaling method for aggregating answers drawn from multiple candidate responses of an LLM regardless of how they were generated, including variations in prompt phrasing and sampling strategy. RC enhances answer aggregation by not only considering the number of occurrences of each answer in the candidate response set, but also the consistency of the model's internal activations while generating the set of responses leading to each answer. These activations can be either dense (raw model activations) or sparse (encoded via pretrained sparse autoencoders). Our rationale is that if the model's representations of multiple responses converging on the same answer are highly variable, this answer is more likely to be the result of incoherent reasoning and should be down-weighted during aggregation. Importantly, our method only uses cached activations and lightweight similarity computations and requires no additional model queries. Through experiments with four open-source LLMs and four reasoning datasets, we validate the effectiveness of RC for improving task performance during inference, with consistent accuracy improvements (up to 4%) over strong test-time scaling baselines. We also show that consistency in the sparse activation signals aligns well with the common notion of coherent reasoning.

Ming Pok Ng, Junqi Jiang, Gabriel Freedman et al.

Leveraging outputs from multiple large language models (LLMs) is emerging as a method for harnessing their power across a wide range of tasks while mitigating their capacity for making errors, e.g., hallucinations. However, current approaches to combining insights from multiple LLMs often involve unstructured interactions (e.g., free debate), resulting in model generations that are not faithfully justifiable. In this work, we introduce MArgE, a novel framework to provide formal structure to the evidence from each LLM, in the form of a tree of extracted arguments, for the task of claim verification. We use a variant of Argumentative LLMs (ArgLLMs), i.e. LLMs driven by frameworks and semantics from the field of computational argumentation, to construct structured argument trees for given claims. This process creates an inspectable pathway from the initial arguments to the final claim verification decisions, providing a faithful justification thereof. We show experimentally that MArgE can significantly outperform single LLMs, including three open-source models (4B to 8B parameters), GPT-4o-mini and existing ArgLLMs, as well as prior methods for unstructured multi-LLM debates. We thus demonstrate the advantages of incorporating formal, argumentative reasoning mechanisms when combining multiple LLM outputs.

Antonio Rago, Maria Vanina Martinez

As AI models become ever more complex and intertwined in humans' daily lives, greater levels of interactivity of explainable AI (XAI) methods are needed. In this paper, we propose the use of belief change theory as a formal foundation for operators that model the incorporation of new information, i.e. user feedback in interactive XAI, to logical representations of data-driven classifiers. We argue that this type of formalisation provides a framework and a methodology to develop interactive explanations in a principled manner, providing warranted behaviour and favouring transparency and accountability of such interactions. Concretely, we first define a novel, logic-based formalism to represent explanatory information shared between humans and machines. We then consider real world scenarios for interactive XAI, with different prioritisations of new and existing knowledge, where our formalism may be instantiated. Finally, we analyse a core set of belief change postulates, discussing their suitability for our real world settings and pointing to particular challenges that may require the relaxation or reinterpretation of some of the theoretical assumptions underlying existing operators.

Gabriel Freedman, Adam Dejl, Deniz Gorur et al.

The profusion of knowledge encoded in large language models (LLMs) and their ability to apply this knowledge zero-shot in a range of settings makes them promising candidates for use in decision-making. However, they are currently limited by their inability to provide outputs which can be faithfully explained and effectively contested to correct mistakes. In this paper, we attempt to reconcile these strengths and weaknesses by introducing \emph{argumentative LLMs (ArgLLMs)}, a method for augmenting LLMs with argumentative reasoning. Concretely, ArgLLMs construct argumentation frameworks, which then serve as the basis for formal reasoning in support of decision-making. The interpretable nature of these argumentation frameworks and formal reasoning means that any decision made by ArgLLMs may be explained and contested. We evaluate ArgLLMs' performance experimentally in comparison with state-of-the-art techniques, in the context of the decision-making task of claim verification. We also define novel properties to characterise contestability and assess ArgLLMs formally in terms of these properties.

Francesco Leofante, Hamed Ayoobi, Adam Dejl et al.

AI has become pervasive in recent years, but state-of-the-art approaches predominantly neglect the need for AI systems to be contestable. Instead, contestability is advocated by AI guidelines (e.g. by the OECD) and regulation of automated decision-making (e.g. GDPR). In this position paper we explore how contestability can be achieved computationally in and for AI. We argue that contestable AI requires dynamic (human-machine and/or machine-machine) explainability and decision-making processes, whereby machines can (i) interact with humans and/or other machines to progressively explain their outputs and/or their reasoning as well as assess grounds for contestation provided by these humans and/or other machines, and (ii) revise their decision-making processes to redress any issues successfully raised during contestation. Given that much of the current AI landscape is tailored to static AIs, the need to accommodate contestability will require a radical rethinking, that, we argue, computational argumentation is ideally suited to support.

Deniz Gorur, Antonio Rago, Francesca Toni

Judgmental forecasting employs human opinions to make predictions about future events, rather than exclusively historical data as in quantitative forecasting. When these opinions form an argumentative structure around forecasts, it is useful to study the properties of the forecasts from an argumentative perspective. In this paper, we advocate and formally define a property of argumentative coherence, which, in essence, requires that a forecaster's reasoning is coherent with their forecast. We then conduct three evaluations with our notion of coherence. First, we assess the impact of enforcing coherence on human forecasters as well as on Large Language Model (LLM)-based forecasters, given that they have recently shown to be competitive with human forecasters. In both cases, we show that filtering out incoherent predictions improves forecasting accuracy consistently, supporting the practical value of coherence in both human and LLM-based forecasting. Then, via crowd-sourced user experiments, we show that, despite its apparent intuitiveness and usefulness, users do not generally align with this coherence property. This points to the need to integrate, within argumentation-based judgmental forecasting, mechanisms to filter out incoherent opinions before obtaining group forecasting predictions.

Jamie Todd, Junqi Jiang, Aaron Russo et al.

Formula One (F1) race strategy takes place in a high-pressure and fast-paced environment where split-second decisions can drastically affect race results. Two of the core decisions of race strategy are when to make pit stops (i.e. replace the cars' tyres) and which tyre compounds (hard, medium or soft, in normal conditions) to select. The optimal pit stop decisions can be determined by estimating the tyre degradation of these compounds, which in turn can be computed from the energy applied to each tyre, i.e. the tyre energy. In this work, we trained deep learning models, using the Mercedes-AMG PETRONAS F1 team's historic race data consisting of telemetry, to forecast tyre energies during races. Additionally, we fitted XGBoost, a decision tree-based machine learning algorithm, to the same dataset and compared the results, with both giving impressive performance. Furthermore, we incorporated two different explainable AI methods, namely feature importance and counterfactual explanations, to gain insights into the reasoning behind the forecasts. Our contributions thus result in an explainable, automated method which could assist F1 teams in optimising their race strategy.

Devin Thomas, Junqi Jiang, Avinash Kori et al.

In Formula One, teams compete to develop their cars and achieve the highest possible finishing position in each race. During a race, however, teams are unable to alter the car, so they must improve their cars' finishing positions via race strategy, i.e. optimising their selection of which tyre compounds to put on the car and when to do so. In this work, we introduce a reinforcement learning model, RSRL (Race Strategy Reinforcement Learning), to control race strategies in simulations, offering a faster alternative to the industry standard of hard-coded and Monte Carlo-based race strategies. Controlling cars with a pace equating to an expected finishing position of P5.5 (where P1 represents first place and P20 is last place), RSRL achieves an average finishing position of P5.33 on our test race, the 2023 Bahrain Grand Prix, outperforming the best baseline of P5.63. We then demonstrate, in a generalisability study, how performance for one track or multiple tracks can be prioritised via training. Further, we supplement model predictions with feature importance, decision tree-based surrogate models, and decision tree counterfactuals towards improving user trust in the model. Finally, we provide illustrations which exemplify our approach in real-world situations, drawing parallels between simulations and reality.

Antonio Rago, Kristijonas Čyras, Jack Mumford et al.

This chapter provides an overview of research works that present approaches with some degree of cross-fertilisation between Computational Argumentation and Machine Learning. Our review of the literature identified two broad themes representing the purpose of the interaction between these two areas: argumentation for machine learning and machine learning for argumentation. Across these two themes, we systematically evaluate the spectrum of works across various dimensions, including the type of learning and the form of argumentation framework used. Further, we identify three types of interaction between these two areas: synergistic approaches, where the Argumentation and Machine Learning components are tightly integrated; segmented approaches, where the two are interleaved such that the outputs of one are the inputs of the other; and approximated approaches, where one component shadows the other at a chosen level of detail. We draw conclusions about the suitability of certain forms of Argumentation for supporting certain types of Machine Learning, and vice versa, with clear patterns emerging from the review. Whilst the reviewed works provide inspiration for successfully combining the two fields of research, we also identify and discuss limitations and challenges that ought to be addressed in order to ensure that they remain a fruitful pairing as AI advances.

Deniz Gorur, Antonio Rago, Francesca Toni

Judgmental forecasting is the task of making predictions about future events based on human judgment. This task can be seen as a form of claim verification, where the claim corresponds to a future event and the task is to assess the plausibility of that event. In this paper, we propose a novel multi-agent framework for claim verification, whereby different agents may disagree on claim veracity and bring specific evidence for and against the claims, represented as quantitative bipolar argumentation frameworks (QBAFs). We then instantiate the framework for supporting claim verification, with a variety of agents realised with Large Language Models (LLMs): (1) ArgLLM agents, an existing approach for claim verification that generates and evaluates QBAFs; (2) RbAM agents, whereby LLM-empowered Relation-based Argument Mining (RbAM) from external sources is used to generate QBAFs; (3) RAG-ArgLLM agents, extending ArgLLM agents with a form of Retrieval-Augmented Generation (RAG) of arguments from external sources. Finally, we conduct experiments with two standard judgmental forecasting datasets, with instances of our framework with two or three agents, empowered by six different base LLMs. We observe that combining evidence from agents can improve forecasting accuracy, especially in the case of three agents, while providing an explainable combination of evidence for claim verification.

Kevin Zhou, Adam Dejl, Gabriel Freedman et al.

Research in uncertainty quantification (UQ) for large language models (LLMs) is increasingly important towards guaranteeing the reliability of this groundbreaking technology. We explore the integration of LLM UQ methods in argumentative LLMs (ArgLLMs), an explainable LLM framework for decision-making based on computational argumentation in which UQ plays a critical role. We conduct experiments to evaluate ArgLLMs' performance on claim verification tasks when using different LLM UQ methods, inherently performing an assessment of the UQ methods' effectiveness. Moreover, the experimental procedure itself is a novel way of evaluating the effectiveness of UQ methods, especially when intricate and potentially contentious statements are present. Our results demonstrate that, despite its simplicity, direct prompting is an effective UQ strategy in ArgLLMs, outperforming considerably more complex approaches.

Anna Rapberger, Fabrizio Russo, Antonio Rago et al.

In computational argumentation, gradual semantics are fine-grained alternatives to extension-based and labelling-based semantics . They ascribe a dialectical strength to (components of) arguments sanctioning their degree of acceptability. Several gradual semantics have been studied for abstract, bipolar and quantitative bipolar argumentation frameworks (QBAFs), as well as, to a lesser extent, for some forms of structured argumentation. However, this has not been the case for assumption-based argumentation (ABA), despite it being a popular form of structured argumentation with several applications where gradual semantics could be useful. In this paper, we fill this gap and propose a family of novel gradual semantics for equipping assumptions, which are the core components in ABA frameworks, with dialectical strengths. To do so, we use bipolar set-based argumentation frameworks as an abstraction of (potentially non-flat) ABA frameworks and generalise state-of-the-art modular gradual semantics for QBAFs. We show that our gradual ABA semantics satisfy suitable adaptations of desirable properties of gradual QBAF semantics, such as balance and monotonicity. We also explore an argument-based approach that leverages established QBAF modular semantics directly, and use it as baseline. Finally, we conduct experiments with synthetic ABA frameworks to compare our gradual ABA semantics with its argument-based counterpart and assess convergence.

Stylianos Loukas Vasileiou, Antonio Rago, Maria Vanina Martinez et al.

Understanding how humans revise their beliefs in light of new information is crucial for developing AI systems which can effectively model, and thus align with, human reasoning. While theoretical belief revision frameworks rely on a set of principles that establish how these operations are performed, empirical evidence from cognitive psychology suggests that people may follow different patterns when presented with conflicting information. In this paper, we present three comprehensive user studies showing that people consistently prefer explanation-based revisions, i.e., those which are guided by explanations, that result in changes to their belief systems that are not necessarily captured by classical belief change theory. Our experiments systematically investigate how people revise their beliefs with explanations for inconsistencies, whether they are provided with them or left to formulate them themselves, demonstrating a robust preference for what may seem non-minimal revisions across different types of scenarios. These findings have implications for AI systems designed to model human reasoning or interact with humans, suggesting that such systems should accommodate explanation-based, potentially non-minimal belief revision operators to better align with human cognitive processes.

Kristijonas Čyras, Antonio Rago, Emanuele Albini et al.

Explainable AI (XAI) has been investigated for decades and, together with AI itself, has witnessed unprecedented growth in recent years. Among various approaches to XAI, argumentative models have been advocated in both the AI and social science literature, as their dialectical nature appears to match some basic desirable features of the explanation activity. In this survey we overview XAI approaches built using methods from the field of computational argumentation, leveraging its wide array of reasoning abstractions and explanation delivery methods. We overview the literature focusing on different types of explanation (intrinsic and post-hoc), different models with which argumentation-based explanations are deployed, different forms of delivery, and different argumentation frameworks they use. We also lay out a roadmap for future work.

Antonio Rago, Emanuele Albini, Pietro Baroni et al.

One of the most pressing issues in AI in recent years has been the need to address the lack of explainability of many of its models. We focus on explanations for discrete Bayesian network classifiers (BCs), targeting greater transparency of their inner workings by including intermediate variables in explanations, rather than just the input and output variables as is standard practice. The proposed influence-driven explanations (IDXs) for BCs are systematically generated using the causal relationships between variables within the BC, called influences, which are then categorised by logical requirements, called relation properties, according to their behaviour. These relation properties both provide guarantees beyond heuristic explanation methods and allow the information underpinning an explanation to be tailored to a particular context's and user's requirements, e.g., IDXs may be dialectical or counterfactual. We demonstrate IDXs' capability to explain various forms of BCs, e.g., naive or multi-label, binary or categorical, and also integrate recent approaches to explanations for BCs from the literature. We evaluate IDXs with theoretical and empirical analyses, demonstrating their considerable advantages when compared with existing explanation methods.

Emanuele Albini, Piyawat Lertvittayakumjorn, Antonio Rago et al.

Despite the recent, widespread focus on eXplainable AI (XAI), explanations computed by XAI methods tend to provide little insight into the functioning of Neural Networks (NNs). We propose a novel framework for obtaining (local) explanations from NNs while providing transparency about their inner workings, and show how to deploy it for various neural architectures and tasks. We refer to our novel explanations collectively as Deep Argumentative eXplanations (DAXs in short), given that they reflect the deep structure of the underlying NNs and that they are defined in terms of notions from computational argumentation, a form of symbolic AI offering useful reasoning abstractions for explanation. We evaluate DAXs empirically showing that they exhibit deep fidelity and low computational cost. We also conduct human experiments indicating that DAXs are comprehensible to humans and align with their judgement, while also being competitive, in terms of user acceptance, with some existing approaches to XAI that also have an argumentative spirit.