To better understand, and analyse, natural language arguments, it is desirable to represent them as logical arguments. However, most real-world arguments are enthymemes (i.e. some of the premises and/or claims are implicit), and therefore, there is a need to identify these implicit aspects. A ramification of this is that we may then need to edit some of the explicit premises and/or claim to remove redundant aspects and/or to allow the newly identified implicit formulae to work correctly with the explicit formulae. Furthermore, we may need to edit the claim so that it correctly attacks or supports other arguments as predicted by argument mining or as required by the user. To address these requirements, we propose a logic-based framework, based on classical propositional logic, for representing enthymemes, and manipulating them through a range of logical operations. We introduce meta-level rules to manipulate arguments (e.g. to add or delete premises, to edit claims, to split an argument into two arguments, and to merge two arguments into one). In order to direct the use of meta-level rules, we also introduce gain measures. When choosing a sequence of meta-level rules to apply, we can choose those that increase gain. This meta-level reasoning framework provides some clarity on the nature of enthymemes, and on how agents might elucidate them through a transparent and incremental process.

Gradual semantics for weighted bipolar argumentation provide a principled framework for modelling argumentative reasoning, yet existing approaches remain mostly scalar, fixed, and weakly grounded in empirical data. We introduce learnable multi-attribute gradual semantics for persuasion prediction in argumentative debates. Our approach builds a dataset of 600 textual debates converted into multi-attribute argumentation graphs enriched with multi-dimensional features on nodes and relations. Building on this representation, we propose learnable aggregation operators that distinguish intrinsic quality from persuasive strategy dimensions. Experiments show that the learned semantics achieve competitive performance with neural and LLM-based baselines while preserving interpretability.

The ability to substitute some resource or tool for another is a common and important human ability. For example, in cooking, we often lack an ingredient for a recipe and we solve this problem by finding a substitute ingredient. There are various ways that we may reason about this. Often we need to draw on commonsense reasoning to find a substitute. For instance, we can think of the properties of the missing item, and try to find similar items with similar properties. Despite the importance of substitution in human intelligence, there is a lack of a theoretical understanding of the faculty. To address this shortcoming, we propose a commonsense reasoning framework for conceptualizing and harnessing substitution. In order to ground our proposal, we focus on cooking. Though we believe the proposal can be straightforwardly adapted to other applications that require formalization of substitution. Our approach is to produce a general framework based on distance measures for determining similarity (e.g. between ingredients, or between processing steps), and on identifying inconsistencies between the logical representation of recipes and integrity constraints that we use to flag the need for mitigation (e.g. after substituting one kind of pasta for another in a recipe, we may identify an inconsistency in the cooking time, and this is resolved by updating the cooking time).

Inconsistency is a common problem in knowledge, and so there is a need to analyse it. Inconsistency measures assess its severity, but there is a more basic question: "where is the inconsistency?". Typically, not all subsets of a knowledgebase are causing the inconsistency, and minimal inconsistent sets have been the standard way to localise the germane ones, even though there are shortcomings in some scenarios. Recently, ⋆-conflicts were proposed as a more suitable definition to localise inconsistency when considering a method to repair it. But in general there is no way to tell what is a sensible definition to capture the germane conflicts. This work provides a set of desirable properties to assess definitions for germane conflicts. Also, a new conflict definition, based on substitution, is presented and evaluated via the proposed properties, and the related computational complexity is analysed.

Commonsense reasoning is a key aspect of human intelligence. If we are to develop robust and deep intelligent systems, then we need to understand the diversity and complexity of commonsense reasoning across the gamut of human activities. An interesting class of commonsense reasoning problems arises when people are faced with natural disasters. To investigate this topic, we present RESPONSE, a human-curated dataset containing 1789 annotated instances featuring 6037 sets of questions designed to assess LLMs’ commonsense reasoning in disaster situations across different time frames. The dataset includes problem descriptions, missing resources, time-sensitive solutions, and their justifications, with a subset validated by environmental engineers. Through both automatic metrics and human evaluation, we compare LLM-generated recommendations against human responses. Our findings show that even state-of-the-art models like GPT-4 achieve only 37% human-evaluated correctness for immediate response actions, highlighting significant room for improvement in LLMs’ ability for commonsense reasoning in crises.

Argumentation is an important cognitive activity that involves scrutinizing arguments and counterarguments. Ideally, an argument's premises should entail its claim while maintaining consistency between the premises and the claim. Furthermore, for a pair of arguments, it is important to determine whether one supports or attacks the other. Argument mining is being developed to automatically extract arguments from text, and identify support or attack relationships between them. But there would be advantages of using a formal logic representation of the arguments as it may offer a clearer and less ambiguous representation of information, and it would support automated reasoning. Whilst there are some frameworks for modeling logical argumentation, there is a lack of methods to translate text from argument mining into logical formulas. To address this gap, we propose a neuro-symbolic pipeline combining a pre-trained large language model (LLM) with neuro-symbolic reasoning. This converts free-text premises and claims such as from argument mining into logical formulas, based on abstract meaning representation (AMR), and employs a SAT solver for entailment and contradiction checks. We apply this method to the Sentences Involving Compositional Knowledge (SICK) and STS-B (Semantic Textual Similarity Benchmark) datasets, and provide promising performance results. We then explain how our pipeline can be used construct structured argument graphs from simple natural language arguments.

Argument mining is a natural language processing technology aimed at identifying the explicit premises and claims of arguments in text, and the support and attack relationships between them. To better understand, and automatically analyse, the argument maps that are output from argument mining, it would be desirable to instantiate the arguments in the argument map with logical arguments. However, most real-world arguments are enthymemes (i.e. some of the premises and/or claim are implicit), which need to be decoded (i.e. the implicit aspects need to be identified). A key challenge is to decode enthymemes so as to respect the support and attack relationships in the argument map. addressing the problem of identifying the missing premises and/or claim, and discerning the relationships between them. To address this, we present a novel framework, based on default logic, for representing arguments including enthymemes. We show how decoding an enthymeme means identifying the default rules that are implicit in the premises and claims. We then show how choosing a decoding of the enthymemes in an argument map can be formalized as an optimization problem, and that a solution can be obtained using MaxSAT solvers.

An argument can be seen as a pair of premises and a claim they support. Human arguments are often approximate, with some premises left implicit, leading to an implicit inference of the claim, i.e., forming enthymemes. To better understand and use them, we must decode these approximate enthymemes, typically by identifying missing premises to make the inference explicit, and, as we propose, by also removing irrelevant content to improve argument quality in specific contexts. Often, multiple decodings of an enthymeme are possible. However, no formal method has yet been proposed for identifying higher-quality decodings. To pave the way, we introduce six types of criteria for evaluating aspects of decodings. Then, we introduce the concept of a criterion measure, designed to evaluate decodings based on a specific criterion. In parallel, we define desirable properties for criterion measures, referred to as axioms, and we systematically evaluate our criterion measures with respect to them. Finally, we introduce the notion of quality measure that combine specific criterion measures to give an overall evaluation of the quality of decodings.

Proposals for strategies for dialogical argumentation often focus on situations where one of the agents wins the dialogue and the other agent loses. Yet in real-world argumentation, it is common for agents to not view a dialogue as a zero-sum game. Rather, the agents may enter into a dialogue with divergent but not diametrically opposing views on what is important (e.g. a doctor trying to persuade a patient to give up smoking when the patient would like to be healthy but gets some pleasure from smoking). Furthermore, there may be multiple persuasion goals (e.g. reduce smoking of cigarettes to 20 per day, or 10 per day, or 5 per day, or 1 per day, or 0 per day, where the doctor prefers 0 per day most, and 20 per day least, whereas the patient might prefer 5 per day most, and 20 per day and 0 per day least). In order to develop persuasive chatbots that support this kind of behaviour change application, this paper presents dialogue protocols, and a strategy for a chatbot, to optimize choice of moves.

Automated persuasion systems (APS) are conversational agents that exchange arguments and counterarguments with users during dialogues to persuade them to believe in something. Such systems use strategies (or policies) to carefully select a sequence of arguments that are tailored to the user’s needs and will likely have a positive outcome, that is, changing the user’s belief in a certain argument. Biparty Decision Theory (BDT) is a framework that uses game theory to formalize a dialogue between an APS and a user, that is, an exchange of (counter) arguments during each turn of the APS or the user. During the APS turn, the BDT policy selects the best argument to maximize only the utility for the APS and neglects the utility of that argument for the user. This is a reasonable choice in games, but in a persuasive dialogue, it can result in arguments that have a high utility for the APS but a modest utility for the user. There the user may be less likely to be persuaded. This is crucial in settings where there are no arguments with good utilities for both the APS and the user and a compromise has to be found. To this extent, we define a new family of policies for BDT, called aggregated policies, that consider, during the decisions of the APS, an aggregation of the APS and user’s utilities. Such an aggregation considers both the APS and the user’s needs leading toward a sequence of arguments representing the best trade-off of utilities. We evaluate the approach using both a new synthetic dataset and a published dataset of utilities for dialogical argumentation. The results show the aggregated policies find better compromise arguments w.r.t. the classical policy of BDT.

David Chanin, Anthony Hunter, Oana-Maria Camburu. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Evidence from studies, such as in science or medicine, often corresponds to conditional probability statements. Furthermore, evidence can conflict, in particular when coming from multiple studies. Whilst it is natural to make sense of such evidence using arguments, there is a lack of a systematic formalism for representing and reasoning with conditional probability statements in computational argumentation. We address this shortcoming by providing a formalization of conditional probabilistic argumentation based on probabilistic conditional logic. We provide a semantics and a collection of comprehensible inference rules that give different insights into evidence. We show how arguments constructed from proofs and attacks between them can be analyzed as arguments graphs using dialectical semantics and via the epistemic approach to probabilistic argumentation. Our approach allows for a transparent and systematic way of handling uncertainty that often arises in evidence.

AI systems often need to deal with inconsistencies. One way of getting information about inconsistencies is by measuring the amount of information in the knowledgebase. In the past 20 years numerous inconsistency measures have been proposed. Many of these measures are syntactic measures, that is, they are based in some way on the minimal inconsistent subsets of the knowledgebase. Very little attention has been given to semantic inconsistency measures, that is, ones that are based on the models of the knowledgebase where the notion of a model is generalized to allow an atom to be assigned a truth value that denotes contradiction. In fact, only one nontrivial semantic inconsistency measure, the contension measure, has been in wide use. The purpose of this paper is to define a class of semantic inconsistency measures based on 3-valued logics. First, we show which 3-valued logics are useful for this purpose. Then we show that the class of semantic inconsistency measures can be developed using a graphical framework similar to the way that syntactic inconsistency measures have been studied. We give several examples of semantic inconsistency measures and show how they apply to three useful 3-valued logics. We also investigate the properties of these inconsistency measures and show their computation for several knowledgebases.

Automated persuasion systems (APS) aim to persuade a user to believe something by entering into a dialogue in which arguments and counterarguments are exchanged. To maximize the probability that an APS is successful in persuading a user, it can identify a global policy that will allow it to select the best arguments it presents at each stage of the dialogue whatever arguments the user presents. However, in real applications, such as for healthcare, it is unlikely the utility of the outcome of the dialogue will be the same, or the exact opposite, for the APS and user. In order to deal with this situation, games in extended form have been harnessed for argumentation in Bi-party Decision Theory. This opens new problems that we address in this paper: (1) How can we use Machine Learning (ML) methods to predict utility functions for different subpopulations of users? and (2) How can we identify for a new user the best utility function from amongst those that we have learned. To this extent, we develop two ML methods, EAI and EDS, that leverage information coming from the users to predict their utilities. EAI is restricted to a fixed amount of information, whereas EDS can choose the information that best detects the subpopulations of a user. We evaluate EAI and EDS in a simulation setting and in a realistic case study concerning healthy eating habits. Results are promising in both cases, but EDS is more effective at predicting useful utility functions.

An argument can be regarded as some premises and a claim following from those premises. Normally, arguments exchanged by human agents are enthymemes, which generally means that some premises are implicit. So when an enthymeme is presented, the presenter expects that the recipient can identify the missing premises. An important kind of implicitness arises when a presenter assumes that two symbols denote the same, or nearly the same, concept (e.g. dad and father), and uses the symbols interchangeably. To model this process, we propose the use of semantic distance measures (e.g. based on a vector representation of word embeddings or a semantic network representation of words) to determine whether one symbol can be substituted by another. We present a theoretical framework for using substitutions, together with abduction of default knowledge, for understanding enthymemes based on deductive argumentation, and investigate how this could be used in practice.

Epistemic graphs have been developed for modelling an agent’s degree of belief in an argument and how belief in one argument may influence the belief in other arguments. These beliefs are represented by constraints on probability distributions. In this paper, we present a framework for reasoning with epistemic graphs that allows for beliefs for individual arguments to be determined given beliefs in some of the other arguments. We present and evaluate algorithms based on SAT solvers.

It is common for people to remark that a particular argument is a strong (or weak) argument. Having a handle on the relative strengths of arguments can help in deciding on which arguments to consider, which arguments to regard as acceptable, and on which arguments to present to others in a discussion. In computational models of argument, there is a need for a deeper understanding of argument strength. It is a multidimensional problem, and in this paper, we focus on one aspect of argument strength for deductive argumentation based on a defeasible logic. We assume a probability distribution over models of the language and consider how there are various ways to calculate argument strength based on the probabilistic necessity and sufficiency of the premises for the claim, the probabilistic sufficiency of competing premises the claim, and the probabilistic necessity of the premises for competing claims. We provide axioms for characterizing probability-based measures of argument strength, and we investigate four specific probability-based measures.

Chatbots are versatile tools that have the potential of being used for computational persuasion where the chatbot acts as the persuader and the human agent as the persuadee. To allow the user to type his or her arguments, as opposed to selecting them from a menu, the chatbot needs a sufficiently large knowledge base of arguments and counterarguments. And in order to make the user change their current stance on a subject, the chatbot needs a method to select persuasive counterarguments. To address this, we present a chatbot that is equipped with an argument graph and the ability to identify the concerns of the user argument in order to select appropriate counterarguments. We evaluate the bot in a study with participants and show how using our method can make the chatbot more persuasive.

Epistemic graphs are a proposal for modelling how agents may have beliefs in arguments and how beliefs in some arguments may influence the beliefs in others. The beliefs in arguments are represented by probability distributions and influences between arguments are represented by logical constraints on these probability distributions. This allows for various kinds of influence to be represented including supporting, attacking, and mixed, and it allows for aggregation of influence to be captured, in a context-sensitive way. In this paper, we investigate methods for learning constraints, and thereby the nature of influences, from data. We evaluate our approach by showing that we can obtain constraints with reasonable quality from two publicly available studies.

Product reviews which are increasingly commonplace on the web typically contain a textual component and a numerical rating. The textual component can be viewed as a collection of arguments for and against the product. Whilst the reviewer may not have provided the attacks between these arguments they typically provide an indication of which set of arguments they view as being more acceptable/winning via the numerical rating (i.e. a positive rating indicates that the positive arguments are accepted and vice versa). Our framework builds upon this intuition and we propose a two step process for identifying a probability distribution over the set of possible argument graphs that the reviewer may have had in mind. The first is the identification step in which for a given review, we identify a distribution by analysing the relationship between the rating and polarity of arguments in the review via the constellations approach to probabilistic argumentation. The second step is the refinement step in which we harness ratings from multiple reviews and use this to refine our probability distribution thus enabling us to learn from the data. We illustrate the applicability of our approach by testing it with real data.

The epistemic approach to probabilistic argumentation assigns belief to arguments. To better understand this approach, we consider structured arguments. Our approach is to start with a probability distribution, and generate an argument graph containing structured arguments with a probability assignment. We construct arguments directly from the probability distribution, rather than a knowledgebase, and then consider methods for selecting the arguments and counterarguments to present in the argument graph. This provides mechanisms for managing uncertainty in argumentation, and for argument-based explanations of probability distributions (that might come from data or from beliefs of an agent).

In the real world it is common for agents to posit arguments concerning an issue but not directly specify the attack relations between them. Nonetheless the agent may have these attacks in mind and instead they may provide a proxy indicator through which one can infer the agent's intended argument graph (arguments and attacks). Consider online reviews, where reviews are collections of arguments for and against the product (positive and negative) under review and the rating indicates whether the positive or negative arguments succeed ultimately. In previous work [1] we have proposed a method that formalises this intuition and uses the constellations approach to probabilistic argumentation to construct a probability distribution over the set of arguments graphs the agent may have had in mind. In this paper we extend this proposal and provide a method, that uses Bayesian inference, to update the initial probability distribution using real data. We evaluate our proposal by conducting a number of simulations using synthetic data.

Structured argumentation involves drawing inferences from knowledge in order to construct arguments and counterarguments. Since knowledge can be uncertain, we can use a probabilistic approach to representing and reasoning with the knowledge. Individual arguments can be constructed from the knowledge, with the belief in each argument determined just from the belief in the formulae appearing in the argument. However, if the original knowledgebase is inconsistent, this does not take into account the counterarguments that can be constructed. We therefore need a wider perspective that revises the belief in individual arguments in order to take into account the counterarguments. To address this need, we present a framework for probabilistic argumentation that uses relaxation methods to give a coherent view on the knowledge, and thereby revises the belief in the arguments that are generated from the knowledge.

Conversational agents, also known as chatbots, are versatile tools that have the potential of being used in dialogical argumentation. They could possibly be deployed in tasks such as persuasion for behaviour change (e.g. persuading people to eat more fruit, to take regular exercise, etc.). However, to achieve this, there is a need to develop methods for acquiring appropriate arguments and counterargument that reflect both sides of the discussion. For instance, to persuade someone to do regular exercise, the chatbot needs to know counterarguments that the user might have for not doing exercise. To address this need, we present methods for acquiring arguments and counterarguments, and importantly, meta-level information that can be useful for deciding when arguments can be used during an argumentation dialogue. We evaluate these methods in studies with participants and show how harnessing these methods in a chatbot can make it more persuasive.

The aim of measuring inconsistency is to obtain an evaluation of the imperfections in a set of formulas, and this evaluation may then be used to help decide on some course of action (such as rejecting some of the formulas, resolving the inconsistency, seeking better sources of information, etc). A number of proposals have been made to define measures of inconsistency. Each has its rationale. But to date, it is not clear how to delineate the space of options for measures, nor is it clear how we can classify measures systematically. To address these problems, we introduce a general framework for comparing syntactic measures of inconsistency. It is based on the notion of an inconsistency graph for each knowledgebase (a bipartite graph with a set of vertices representing formulas in the knowledgebase, a set of vertices representing minimal inconsistent subsets of the knowledgebase, and edges representing that a formula belongs to a minimal inconsistent subset). We then show that various measures can be computed using the inconsistency graph. Then we introduce abstractions of the inconsistency graph and use them to construct a hierarchy of syntactic inconsistency measures. Furthermore, we extend the inconsistency graph concept with a labeling that extends the hierarchy to include some other types of inconsistency measures.

A number of proposals have been made to define inconsistency measures. Each has its rationale. But to date, it is not clear how to delineate the space of options for measures, nor is it clear how we can classify measures systematically. In this paper, we introduce a general framework for comparing syntactic inconsistency measures. It uses the construction of an inconsistency graph for each knowledgebase. We then introduce abstractions of the inconsistency graph and use the hierarchy of the abstractions to classify a range of inconsistency measures.

In this paper we present design concepts for a chatbot than can be used for argument acquisition. We call the acquisition of arguments by means of a chatbot argument harvesting. The chatbot asks the user for his or her arguments on a topic of interest. It can also be used to harvest counterargument, values, preferences, as well as information about the user like his/her personal circumstances. The harvested arguments and other attributes have various applications from the instantiation of argument graphs to the development of computational persuasion systems.

Much research in computational argumentation assumes that arguments can be obtained in some way. Yet, to improve and apply models of argument, we need methods for acquiring them. Current approaches include argument mining from text, hand coding of arguments by researchers, or generating arguments from knowledge bases. In this paper, we propose a new approach, which we call argument harvesting, that uses a chatbot to enter into a dialogue with a participant to get arguments and counterarguments from him or her. Because it is automated, the chatbot can be used repeatedly in many dialogues, and thereby it can generate a large corpus. We describe the architecture of the chatbot, provide methods for clustering arguments by their similarity and value, and an evaluation of our approach in a case study concerning attitudes of women to participation in sport.

Proposals for strategies for dialogical argumentation often focus on situations where one of the agents wins the dialogue and the other agent loses. Yet in real-world argumentation, it is common for agents to not involve such zero-sum situations. Rather, the agents may enter into a dialogue with divergent but not necessarily opposing views on what is important in the outcomes from the argumentation. In order to model this kind of situation, we investigate a decision-theoretic approach that allows different participants to have different utility evaluations of a dialogue, and for the proponent to model the opponent's utility evaluation in order to optimize the choice of move in the dialogue.

We present a probabilistic interpretation of the plausibility of attacks in abstract argumentation frameworks by extending the epistemic approach to probabilistic argumentation with probabilities on attacks. By doing so we also generalise the previously proposed attack semantics by Villata et al. to the probabilistic setting and provide a fine-grained assessment of the plausibility of attacks. We also consider the setting where partial probabilistic information on arguments and/or attacks is given and missing probabilities have to be derived.

Persuasion is an activity that involves one party (the persuader) trying to induce another party (the persuadee) to believe or do something. For this, it can be advantageous forthe persuader to have a model of the persuadee. Recently, some proposals in the field of computational models of argument have been made for probabilistic models of what the persuadee knows about, or believes. However, these developments have not systematically harnessed established notions in decision theory for maximizing the outcome of a dialogue. To address this, we present a general framework for representing persuasion dialogues as a decision tree, and for using decision rules for selecting moves. Furthermore, we provide some empirical results showing how some well-known decision rules perform, and make observations about their general behaviour in the context of dialogues where there is uncertainty about the accuracy of the user model.

For a participant to play persuasive arguments in a dialogue, s/he may create a model of the other participants. This may include an estimation of what arguments the other participants find believable, convincing, or appealing. The participant can then choose to put forward those arguments that have high scores in the desired criteria. In this paper, we consider how we can crowd-source opinions on the believability, convincingness, and appeal of arguments, and how we can use this information to predict opinions for specific participants on the believability, convincingness, and appeal of specific arguments. We evaluate our approach by crowd-sourcing opinions from 50 participants about 30 arguments. We also discuss how this form of user modelling can be used in a decision-theoretic approach to choosing moves in dialogical argumentation.

There have been a number of proposals for measuring inconsistency in a knowledgebase (i.e. a set of logical formulae). These include measures that consider the minimally inconsistent subsets of the knowledgebase, and measures that consider the paraconsistent models (3 or 4 valued models) of the knowledgebase. In this paper, we present a new approach that considers the amount by which each formula has to be weakened in order for the knowledgebase to be consistent. This approach is based on ideas of knowledge merging by Konienczny and Pino-Perez. We show that this approach gives us measures that are different from existing measures, that have desirable properties, and that can take the significance of inconsistencies into account. The latter is useful when we want to differentiate between inconsistencies that have minor significance from inconsistencies that have major significance. We also show how our measures are potentially useful in applications such as evaluating violations of integrity constraints in databases and for deciding how to act on inconsistency.

Abstract argumentation offers an appealing way of representing and evaluating arguments and counterarguments. This approach can be enhanced by considering probability assignments on arguments, allowing for a quantitative treatment of formal argumentation. In this paper, we regard the assignment as denoting the degree of belief that an agent has in an argument being acceptable. While there are various interpretations of this, an example is how it could be applied to a deductive argument. Here, the degree of belief that an agent has in an argument being acceptable is a combination of the degree to which it believes the premises, the claim, and the derivation of the claim from the premises. We consider constraints on these probability assignments, inspired by crisp notions from classical abstract argumentation frameworks and discuss the issue of probabilistic reasoning with abstract argumentation frameworks. Moreover, we consider the scenario when assessments on the probabilities of a subset of the arguments are given and the probabilities of the remaining arguments have to be derived, taking both the topology of the argumentation framework and principles of probabilistic reasoning into account. We generalise this scenario by also considering inconsistent assessments, i.e., assessments that contradict the topology of the argumentation framework. Building on approaches to inconsistency measurement, we present a general framework to measure the amount of conflict of these assessments and provide a method for inconsistency-tolerant reasoning.

Persuasion is an activity that involves one party trying to induce another party to believe something or to do something. It is an important and multifaceted human facility. Obviously, sales and marketing is heavily dependent on persuasion. But many other activities involve persuasion such as a doctor persuading a patient to drink less alcohol, a road safety expert persuading drivers to not text while driving, or an online safety expert persuading users of social media sites to not reveal too much personal information online. As computing becomes involved in every sphere of life, so too is persuasion a target for applying computer-based solutions. An automated persuasion system (APS) is a system that can engage in a dialogue with a user (the persuadee) in order to persuade the persuadee to do (or not do) some action or to believe (or not believe) something. To do this, an APS aims to use convincing arguments in order to persuade the persuadee. Computational persuasion is the study of formal models of dialogues involving arguments and counterarguments, of user models, and strategies, for APSs. A promising application area for computational persuasion is in behaviour change. Within healthcare organizations, government agencies, and non-governmental agencies, there is much interest in changing behaviour of particular groups of people away from actions that are harmful to themselves and/or to others around them.

When attempting to persuade an agent to believe (or disbelieve) an argument, it can be advantageous for the persuader to have a model of the persuadee. Models have been proposed for taking account of what arguments the persuadee believes and these can be used in a strategy for persuasion. However, there can be uncertainty as to the accuracy of such models. To address this issue, this paper introduces a two-dimensional model that accounts for the uncertainty of belief by a persuadee and for the confidence in that uncertainty evaluation. This gives a better modeling for using lotteries so that the outcomes involve statements about what the user believes/disbelieves, and the confidence value is the degree to which the user does indeed hold those outcomes (and this is a more refined and more natural modeling than found in [19]). This framework is also extended with a modelling of the risk of disengagement by the persuadee.

Computational models of argument are being developed to capture aspects of how persuasion is undertaken. Recent proposals suggest that in a persuasion dialogue between some agents, it is valuable for each agent to model how arguments are believed by the other agents. Beliefs in arguments can be captured by a joint belief distribution over the arguments and updated as the dialogue progresses. This information can be used by the agent to make more intelligent choices of move in the dialogue. Whilst these proposals indicate the value of modelling the beliefs of other agents, there is a question of the computational viability of using a belief distribution over all the arguments. We address this problem in this paper by presenting how probabilistic independence can be leveraged to split this joint distribution into an equivalent set of distributions of smaller size. Experiments show that updating the belief on the split distribution is more efficient than performing updates on the joint distribution.

An argument is seen as reason in favour of a claim. It is made of three parts: a set of premises representing the reason, a conclusion representing the supported claim, and a connection showing how the premises lead to the conclusion. Arguments are frequently exchanged by human agents in natural language (spoken or written) in discussion, debate, negotiation, persuasion, etc. They may be very different in that their three components may have various forms. In this paper, we propose a language for representing such arguments. We show that it is general enough to capture the various forms of arguments encountered in natural language, and that it is possible to represent attack and support relations between arguments as formulas of the same language.

Uncertainty about which arguments or attacks should appear in an argument graph means that there is uncertainty as to the structure of the argument graph. When informal arguments are presented, there may be imprecision in the language used, and so the audience may be uncertain as to the structure of the argument graph as intended by the presenter of the arguments. For a presenter of arguments, it is useful to know the audience's argument graph, but the presenter may be uncertain as to the structure of it. To model each of these situations, we can use probabilistic argument graphs. The set of subgraphs of an argument graph is a sample space. A probability value is assigned to each subgraph such that the sum is 1, thereby reflecting the uncertainty over which is the actual subgraph. We can then determine the probability that a particular set of arguments is included or excluded from an extension according to a particular Dung semantics. We harness this to define the notion of an argument lottery, which can be used by the audience to determine the expected utility of a debate, and can be used by the presenter to decide which arguments to present by choosing those that maximize expected utility.

We consider augmenting abstract argumentation frame-works with probabilistic information and discuss different constraints to obtain meaningful probabilistic information. Moreover, we investigate the problem of incomplete probability assignments and propose a solution for completing these assignments by applying the principle of maximum entropy.

Recently, there has been a proposal by Dung and Thang and by Li et al to extend abstract argumentation to take uncertainty of arguments into account by assigning a probability value to each argument, and then use this assignment to determine the probability that a set of arguments is an extension. In this paper, we explore some of the assumptions behind the definitions, and some of the resulting properties, of the proposal for probabilistic argument graphs.

The formalisation of design decisions serves two purposes: To support the decision maker in choosing which decision to take (decision analysis), and to document the reasons behind decisions for future reference (decision documentation). Approaches which solve the latter task involve a semi-formal pattern of documenting the reasons for and against each of the options, but they generally do not allow an automation of the decision making process. Approaches which solve the former task use a mathematical model of the problem, in which each option is evaluated numerically with respect to some relevant criteria, but they do not support documentation. We investigate the use of argumentation to both analyse and document decisions, solving both tasks with the same method. Additionally, the system we present is able to generate decisions for analysis, instead of relying on a predefined input of options. We collaborated with an aerospace manufacturer to identify common problems in the industry and to create realistic examples from the engineering domain. We show that our system subsumes a certain class of multi criteria decision making problems and that it improves upon previous argumentation-based decision making systems by adding the capability to generate decisions and by clearly defining the semantics used to choose accepted arguments.

In most proposals for logic-based models of argumentation dialogues between agents, the arguments exchanged are logical arguments of the form 〈Φ,α〉 where Φ is a set of formulae (called the support) and α is a formula (called the claim) such that Φ is consistent and Φ entails α. However, arguments presented by real-world agents do not normally fit the mould of being logical arguments. They are normally enthymemes, and so they only explicitly represent some of the premises for entailing their claim and/or they do not explicitly state their claim. For example, for a claim that ‘you need an umbrella today’, a husband may give his wife the premise ‘the weather report predicts rain’. Clearly, the premise does not entail the claim, but it is easy for the wife to identify the assumed knowledge used by the husband in order to reconstruct the intended argument correctly (i.e. ‘if the weather report predicts rain, then you need an umbrella’). Whilst humans are constantly handling examples like this, proposals for logic-based formalizations of the process remain underdeveloped. In this article, we present a logic-based framework for handling enthymemes, some design features of which are influenced by aspects of relevance theory (proposed by Sperber and Wilson). In particular, we use the ideas of maximizing cognitive effect and minimizing cognitive effort in order to enable a proponent of an intended logical argument to construct an enthymeme appropriate for the intended recipient, and for the intended recipient to deconstruct the intended logical argument from the enthymeme. We relate our framework back to Sperber andWilson's relevance theory via some formal properties.

There is interest in artificial intelligence for principled techniques to analyze inconsistent information. This stems from the recognition that the dichotomy between consistent and inconsistent sets of formulae that comes from classical logics is not sufficient for describing inconsistent information. We review some existing proposals and make new proposals for measures of inconsistency and measures of information, and then prove that they are all pairwise incompatible. This shows that the notion of inconsistency is a multi-dimensional concept where different measures provide different insights. We then explore relationships between measures of inconsistency and measures of information in terms of the trade-offs they identify when using them to guide resolution of inconsistency.

The number of clinical trials reports is increasing rapidly due to a large number of clinical trials being conducted; it, therefore, raises an urgent need to utilize the clinical knowledge contained in the clinical trials reports. In this paper, we focus on the qualitative knowledge instead of quantitative knowledge. More precisely, we aim to model and reason with the qualitative comparison (QC for short) relations which consider qualitatively how strongly one drug/therapy is preferred to another in a clinical point of view. To this end, first, we formalize the QC relations, introduce the notions of QC language, QC base, and QC profile; second, we propose a set of induction rules for the QC relations and provide grading interpretations for the QC bases and show how to determine whether a QC base is consistent. Furthermore, when a QC base is inconsistent, we analyze how to measure inconsistencies among QC bases, and we propose different approaches to merging multiple QC bases. Finally, a case study on lowering intraocular pressure is conducted to illustrate our approaches. © 2010 Wiley Periodicals, Inc.

When merging belief sets from different agents, the result is normally a consistent belief set in which the inconsistency between the original sources is not represented. As probability theory is widely used to represent uncertainty, an interesting question therefore is whether it is possible to induce a probability distribution when merging belief sets. To this end, we first propose two approaches to inducing a probability distribution on a set of possible worlds, by extending the principle of indifference on possible worlds. We then study how the (in)dependence relations between atoms can influence the probability distribution. We also propose a set of properties to regulate the merging of belief sets when a probability distribution is output. Furthermore, our merging operators satisfy the well known Konieczny and Pino-Perez postulates if we use the set of possible worlds which have the maximal induced probability values. Our study shows that taking an induced probability distribution as a merging result can better reflect uncertainty and inconsistency among the original knowledge bases.

There are a number of frameworks for modelling argumentation in logic. They incorporate a formal representation of individual arguments and techniques for comparing conflicting arguments. A common assumption for logic-based argumentation is that an argument is a pair ⟨Φ,α⟩ where Φ is a minimal subset of the knowledgebase such that Φ is consistent and Φ entails the claim α. We call the logic used for consistency and entailment, the base logic. Different base logics provide different definitions for consistency and entailment and hence give us different options for argumentation. This paper discusses some of the commonly used base logics in logic-based argumentation, and considers various criteria that can be used to identify commonalities and differences between them.

Evidence-based decision making is becoming increasingly important in many diverse domains, including healthcare, environmental management, and government. This has raised the need for tools to aggregate evidence from multiple sources. For instance, in healthcare, much valuable evidence is in the form of the results from clinical trials that compare the relative merits of treatments. For this, in a previous paper [5], we have proposed a general language for encoding, capturing and synthesizing knowledge from clinical trials and a framework that allows the construction and evaluation of arguments from such knowledge. Now, in this paper, we consider a specific version of the general framework for aggregating qualitative information about trials, and undertake an evaluation of this qualitative framework by comparing the results we obtain with those that are published in the biomedical literature. Whilst the results from our qualitative system are inferior, we show that they do offer a quick and useful aggregation of the evidence, and furthermore, we suggest that it could be coupled with information extraction technology to provide a valuable automated solution.

Evidence-based decision making is becoming increasingly important in healthcare. Much valuable evidence is in the form of the results from clinical trials that compare the relative merits of treatments. For this, in previous papers, we have proposed a general framework for representing and synthesizing knowledge from clinical trials involving the same outcome indicator. Now, in this paper, we present a new framework for representing and synthesizing knowledge from clinical trials involving multiple outcome indicators. In this framework, evidence from randomized clinical trials, systematic reviews, meta-analyses, network analyses, etc., comparing a pair of treatments τ1and τ2according to desired and/or undesired outcomes is aggregated to give an overall evaluation of the treatments saying τ1is superior to τ2, or τ1is equivalent to τ2, or τ1is inferior to τ2. Our general framework incorporates inference rules for generating arguments and counterarguments for claiming that one treatment is superior to another based on the available evidence, and preference rules for specifying which arguments are preferred. In this paper, we also present a new version of this framework that incorporates utility-theoretic criteria for defining specific preference rules over arguments.

The need to merge multiple sources of uncertain information is an important issue in many application areas, particularly when there is potential for contradictions between sources. Possibility theory offers a flexible framework to represent, and reason with, uncertain information, and there is a range of merging operators, such as the conjunctive and disjunctive operators, for combining information. However, with the proposals to date, the context of the information to be merged is largely ignored during the process of selecting which merging operators to use. To address this shortcoming, in this paper, we propose an adaptive merging algorithm which selects largely partially maximal consistent subsets of sources, which can be merged through the relaxation of the conjunctive operator, by assessing the coherence of the information in each subset. In this way, a fusion process can integrate both conjunctive and disjunctive operators in a more flexible manner and thereby be more context dependent. A comparison with related merging methods shows how our algorithm can produce a more consensual result.

We introduce a novel Ontology-based Argumentation Framework (OAF) that links a logic-based argumentation formalism and description logic ontologies. We show how these two formalisms can be tightly coupled by observing a few simple restrictions, and provides features not available in either formalism alone. Our work is evaluated in a large case study on decision-making in treatment choice in breast cancer, where rules are developed from the results of published clinical trials, and we present a small subset of this to demonstrate the use of the system. We show that OAF provides five advantages: (1) facilitating the clear use of shared definitions between multiple authors; (2) enabling us to match terms in the ontology and rules with those in the specific domain literature; (3) providing a close fit between structure of clinical trials and the structure of our rules; (4) delivering significant economies in the size of the rule-base compared to existing approaches; (5) allowing us take advantage of developments in both ontological and argumentative approaches. We also demonstrate that even a restricted language such as ours is sufficient to capture enough information to generate arguments that are useful for clinical practice. An early prototype implementation is available.

The problem of combining pieces of information issued from several sources can be encountered in various fields of application. This paper aims at presenting the different aspects of information fusion in different domains, such as databases, regulations, preferences, sensor fusion, etc., at a quite general level. We first present different types of information encountered in fusion problems, and different aims of the fusion process. Then we focus on representation issues which are relevant when discussing fusion problems. An important issue is then addressed, the handling of conflicting information. We briefly review different domains where fusion is involved, and describe how the fusion problems are stated in each domain. Since the term fusion can have different, more or less broad, meanings, we specify later some terminology with respect to related problems, that might be included in a broad meaning of fusion. Finally we briefly discuss the difficult aspects of validation and evaluation. © 2001 John Wiley & Sons, Inc.

The proof theory of quasi-classical logic (QC logic) allows the derivation of non-trivializable classical inferences from inconsistent information. A non-trivializable, or paraconsistent, logic is, by necessity, a compromise, or weakening, of classical logic. The compromises on QC logic seem to be more appropriate than other paraconsistent logics for applications in computing. In particular, the connectives behave in a classical manner. Here we motivate the need for QC logic, present a proof theory, and semantics for the logic, and compare it to other paraconsistent logics.

In previous work, we advocated continued development of specifications in the presence of inconsistency. To support this, we used classical logic to represent partial specifications and to identify inconsistencies between them. We now present an adaptation of classical logic, which we term quasi-classical (QC) logic, that allows continued reasoning in the presence of inconsistency. The adaptation is a weakening of classical logic that prohibits all trivial derivations, but still allows all resolvants of the assumptions to be derived. Furthermore, the connectives behave in a classical manner. We then present a development called labeled QC logic that records and tracks assumptions used in reasoning. This facilitates a logical analysis of inconsistent information. We discuss that application of labeled QC logic in the analysis of multiperspective specifications. Such specifications are developed by multiple particpants who hold overlapping, often inconsistent, views of the systems they are developing.

In previous work we advocated continued development of specifications in the presence of inconsistency. To support this we presented quasi-classical (QC) logic for reasoning with inconsistent specifications. The logic allows the derivation of non-trivial classical inferences from inconsistent information. In this paper we present a development called labelled QC logic, and some associated analysis tools, that allows the tracking and diagnosis of inconsistent information. The results of analysis are then used to guide further development in the presence of inconsistency. We illustrate the logic and our tools by specifying and analysing parts of the London Ambulance Service. We argue that the scalability of our approach is made possible by deploying the ViewPoints framework for multi-perspective development, such that our analysis tools are only used on partial specifications of a manageable size.

There is a need to develop more intelligent means for handling text in applications such as information retrieval, information filtering, and message classification. This raises the need for mechanisms for ascertaining what an item of text is about. Even though natural language processing offers the best results, it is not always viable. A less accurate, but more viable alternative, is to reason with keywords in the text. Unfortunately, classical reasoning is often inadequate for determining from some keywords what a text is about. In particular it does not allow context-dependent interpretation of keywords. So for example, if some text has the keyword oil, it is usually also about minerals, though with exceptions such as when it has the keyword cooling. To address this kind of problem, we consider a model of "aboutness" based on default logic.

The development of most large and complex systems necessarily involves many people-each with their own perspectives on the system defined by their knowledge, responsibilities, and commitments. To address this we have advocated distributed development of specifications from multiple perspectives. However, this leads to problems of identifying and handling inconsistencies between such perspectives. Maintaining absolute consistency is not always possible. Often this is not even desirable since this can unnecessarily constrain the development process, and can lead to the loss of important information. Indeed since the real-world forces us to work with inconsistencies, we should formalize some of the usually informal or extra-logical ways of responding to them. This is not necessarily done by eradicating inconsistencies but rather by supplying logical rules specifying how we should act on them. To achieve this, we combine two lines of existing research: the ViewPoints framework for perspective development, interaction and organization, and a logic-based approach to inconsistency handling. This paper presents our technique for inconsistency handling in the ViewPoints framework by using simple examples.>