Legal argumentation is a vital cornerstone of justice, underpinning an adversarial form of law, and extensive research has attempted to augment or undertake legal argumentation via the use of computer-based automation including Artificial Intelligence (AI). AI advances in Natural Language Processing (NLP) and Machine Learning (ML) have especially furthered the capabilities of leveraging AI for aiding legal professionals, doing so in ways that are modeled here as CARE, namely Crafting, Assessing, Refining, and Engaging in legal argumentation. In addition to AI-enabled legal argumentation serving to augment human-based lawyering, an aspirational goal of this multi-disciplinary field consists of ultimately achieving autonomously effected human-equivalent legal argumentation. As such, an innovative meta-approach is proposed to apply the Levels of Autonomy (LoA) of AI Legal Reasoning (AILR) to the maturation of AI and Legal Argumentation (AILA), proffering a new means of gauging progress in this ever-evolving and rigorously sought domain.

It is well known that adversarial examples and counterfactual explanations are based on the same mathematical model. However, their relationship has not yet been studied at a conceptual level. The present paper fills this gap. We show that counterfactual reasoning is the common basis of the fields and reliable machine learning their shared goal. Moreover, we illustrate to what extent counterfactual explanations can be regarded as the more general concept than adversarial examples. We introduce the conceptual distinction between feasible and contesting counterfactual explanations and argue that adversarial examples are similar to the latter.

There is a growing concern that the recent progress made in AI, especially regarding the predictive competence of deep learning models, will be undermined by a failure to properly explain their operation and outputs. In response to this disquiet counterfactual explanations have become massively popular in eXplainable AI (XAI) due to their proposed computational psychological, and legal benefits. In contrast however, semifactuals, which are a similar way humans commonly explain their reasoning, have surprisingly received no attention. Most counterfactual methods address tabular rather than image data, partly due to the nondiscrete nature of the latter making good counterfactuals difficult to define. Additionally generating plausible looking explanations which lie on the data manifold is another issue which hampers progress. This paper advances a novel method for generating plausible counterfactuals (and semifactuals) for black box CNN classifiers doing computer vision. The present method, called PlausIble Exceptionality-based Contrastive Explanations (PIECE), modifies all exceptional features in a test image to be normal from the perspective of the counterfactual class (hence concretely defining a counterfactual). Two controlled experiments compare this method to others in the literature, showing that PIECE not only generates the most plausible counterfactuals on several measures, but also the best semifactuals.

The recent Control Argumentation Framework (CAF) is a generalization of Dung's Argumentation Framework which handles argumentation dynamics under uncertainty; especially it can be used to model the behavior of an agent which can anticipate future changes in the environment. Here we provide new insights on this model by defining the notion of possible controllability of a CAF. We study the complexity of this new form of reasoning for the four classical semantics, and we provide a logical encoding for reasoning with this framework.

argumentation [16] is an important topic in the Knowledge Representation and Reasoning community. Intuitively, an abstract argumentation framework (AF) is a directed graph where nodes are arguments and edges are relations (usually attacks) between these arguments. The outcome of such an AF is an evaluation of the arguments' acceptance (through extensions [16, 3], labellings [7] or rankings [1]). In such an AF, the assumption of complete information is made: an argument that appears in the graph is sure to actually exist, and similarly, an edge (or the absence of an edge) in the graph means that the attack between arguments certainly exists (or certainly does not). The question of how to incorporate uncertainty in AFs has then arisen. Two kinds of approaches have been proposed. If a quantitative evaluation of the uncertainty is available, it seems natural to use it in the definition of reasoning mechanisms.

During the first step of practical reasoning, i.e. deliberation, an intelligent agent generates a set of pursuable goals and then selects which of them he commits to achieve. An intelligent agent may in general generate multiple pursuable goals, which may be incompatible among them. In this paper, we focus on the definition, identification and resolution of these incompatibilities. The suggested approach considers the three forms of incompatibility introduced by Castelfranchi and Paglieri, namely the terminal incompatibility, the instrumental or resources incompatibility and the superfluity. We characterise computationally these forms of incompatibility by means of arguments that represent the plans that allow an agent to achieve his goals. Thus, the incompatibility among goals is defined based on the conflicts among their plans, which are represented by means of attacks in an argumentation framework. We also work on the problem of goals selection; we propose to use abstract argumentation theory to deal with this problem, i.e. by applying argumentation semantics. We use a modified version of the "cleaner world" scenario in order to illustrate the performance of our proposal.

Given the recent successes of Deep Learning in AI there has been increased interest in the role and need for explanations in machine learned theories. A distinct notion in this context is that of Michie's definition of Ultra-Strong Machine Learning (USML). USML is demonstrated by a measurable increase in human performance of a task following provision to the human of a symbolic machine learned theory for task performance. A recent paper demonstrates the beneficial effect of a machine learned logic theory for a classification task, yet no existing work has examined the potential harmfulness of machine's involvement in human learning. This paper investigates the explanatory effects of a machine learned theory in the context of simple two person games and proposes a framework for identifying the harmfulness of machine explanations based on the Cognitive Science literature. The approach involves a cognitive window consisting of two quantifiable bounds and it is supported by empirical evidence collected from human trials. Our quantitative and qualitative results indicate that human learning aided by a symbolic machine learned theory which satisfies a cognitive window has achieved significantly higher performance than human self learning. Results also demonstrate that human learning aided by a symbolic machine learned theory that fails to satisfy this window leads to significantly worse performance than unaided human learning.

Artificial Intelligence is often said to be a "black box" -- an opaque, almost mystical thing that we don't really understand. Throw data into the black box, and out comes a prediction, or so they say. However, much of AI is not opaque, it's just a complex system that "reasons" differently than we (think we) do. For example, kids learn to write by first experimenting with letters, and finding patterns in words. GPT-3 learned to write by training a generative text algorithm on the entire Internet, yielding a model with 175 billion parameters that, essentially, predict how "the Internet" would complete a prompt.

In this paper, we present language model system submitted to SemEval-2020 Task 4 competition: "Commonsense Validation and Explanation". We participate in two subtasks for subtask A: validation and subtask B: Explanation. We implemented with transfer learning using pretrained language models (BERT, XLNet, RoBERTa, and ALBERT) and fine-tune them on this task. Then we compared their characteristics in this task to help future researchers understand and use these models more properly. The ensembled model better solves this problem, making the model's accuracy reached 95.9% on subtask A, which just worse than human's by only 3% accuracy.

Post-hoc explanation techniques refer to a posteriori methods that can be used to explain how black-box machine learning models produce their outcomes. Among post-hoc explanation techniques, counterfactual explanations are becoming one of the most popular methods to achieve this objective. In particular, in addition to highlighting the most important features used by the black-box model, they provide users with actionable explanations in the form of data instances that would have received a different outcome. Nonetheless, by doing so, they also leak non-trivial information about the model itself, which raises privacy issues. In this work, we demonstrate how an adversary can leverage the information provided by counterfactual explanations to build high-fidelity and high-accuracy model extraction attacks. More precisely, our attack enables the adversary to build a faithful copy of a target model by accessing its counterfactual explanations. The empirical evaluation of the proposed attack on black-box models trained on real-world datasets demonstrates that they can achieve high-fidelity and high-accuracy extraction even under low query budgets.