Argumentation: A calculus for Human-Centric AI

Abstract

This paper aims to expose and analyze the potential foundational role of Argumentation for Human-Centric AI, and to present the main challenges for this foundational role to be realized in a way that will fit well with the wider requirements and challenges of Human-Centric AI. The central idea set forward is that by endowing machines with the ability to argue with forms of machine argumentation that are cognitively compatible with those of human argumentation, we will be able to support a naturally effective, enhancing and ethical human-machine cooperation and “social” integration.

Argument-based inductive logics, with coverage of compromised perception

“…We believe that the basis for rational human use of AI technology is, or at least ought to be, argumentation, computationally treated, and managed. In this regard, we wholly concur with Dietz et al ( 2022 ). For us, this is a firm and fundamental directive that guides our research.…”

“…Addressing this inadequacy is observed as important by others. For example, Dietz et al ( 2022 ), when setting out desiderata for HCAI systems, include that such systems must exhibit a “body-mind like model of operation to sense, recognize, think, and act” (Dietz et al, 2022 ). For us, broadly speaking, here, “think” is constituted by reasoning, 9 and we associate “sense, recognize” with attention and perception.…”

“…For us, broadly speaking, here, “think” is constituted by reasoning, 9 and we associate “sense, recognize” with attention and perception. Later, in the same study, when discussing what is needed for true success in HCAI and indeed for any brand of AI overall that aspires to cover the human-level case, Dietz et al ( 2022 ) point to the following challenge:…”

“…[Such success must include showing] how the internal integrated operation of cognition, from low-level perception to increasingly higher levels of cognition, is supported by an appropriate architecture, and how an individual's cognition is integrated with the external physical and social environment (Dietz et al, 2022 ; emphasis ours).…”

Sketching the vision of the Web of Debates

“…But it is up to Steffi to decide which of them would actually be the most influential for the readers of the newspaper she is working for. Apart from search engines, the Web of Debates will support several other types of applications, such as everyday assistants, expert companion systems (see e.g., Dietz et al, 2022 for some examples), collaborative decision support systems, intelligent tutoring systems aimed at teaching users how to make better arguments, automated debating systems and others. Some common characteristics of all such systems will be their focus on natural-language arguments and the human aspects of argumentation, their seamless integration within the online, private or social, activities of their users, their adaptability to background knowledge and cognitive characteristics of each user or group of users, their ability to explain any inferences they make, and their ability to develop by learning from experience and by taking into account the feedback provided by their users.…”