Computational Argumentation and Cognition

Emmanuelle Dietz, Antonis C. Kakas, Loizos MichaelarXiv 2021 · OpenReview

TL;DR

This paper explores the crucial intersection of Computational Argumentation (CA) and Cognition, advocating for their integration to build more human-like AI systems. It highlights the need for AI to understand and replicate human reasoning, suggesting that CA offers a powerful framework for achieving this, particularly in the context of Human-Centric AI.

What Problem Does This Paper Address?

The paper tackles the challenge of creating Artificial Intelligence (AI) systems that exhibit natural, human-like cognitive abilities and behaviors. As AI becomes more integrated into our lives, there's a growing expectation for systems to understand and interact with humans on a deeper cognitive level. This requires a holistic understanding of the human mind, encompassing its various cognitive faculties. The authors argue that while AI has made strides, it often lacks the nuanced reasoning and decision-making capabilities that characterize human cognition. They propose that integrating Computational Argumentation, a field within AI that models reasoning and justification, with insights from Cognitive Science, Linguistics, and Philosophy can bridge this gap. The core problem is how to effectively synthesize these disciplines to develop AI that is not only intelligent but also cognitively aligned with human users.

Methods

This paper doesn't present a single novel algorithm or method. Instead, it acts as a survey and a call to action, stemming from the COGNITAR 2020 workshop. The "methods" discussed are the various approaches presented by workshop participants, which explore the synthesis of CA and cognition. These include:

  • Systematic Argument Evaluation: Developing procedures (like CAPNA) to identify, analyze, and evaluate natural language arguments, moving beyond simple identification to understand pragmatic aspects and underlying reasoning.
  • Commonsense Reasoning for AI: Integrating low-level sensory data with high-level cognitive information through abductive reasoning and techniques like Answer Set Programming to enable AI systems (e.g., autonomous vehicles) to understand their environment.
  • Ranking-based Semantics for Argumentation: Extending traditional binary acceptability of arguments to a ranked system using conditional logic, providing a more nuanced understanding of argument strength.
  • Computational Models of Human Reasoning: Developing AI systems (like MReasoner) that simulate human reasoning errors and optimal performance, leveraging insights from cognitive constraints to design better AI and improve human-AI interaction.
  • Argument Mining and Ethos Analysis: Using corpus analysis and experiments to understand how argument phrasing (rephrasing) affects persuasion and to mine for "ethos" (speaker's character) in argumentation.
  • Fallacy Detection and Framing Effects: Applying knowledge of fallacies and framing techniques to understand user behavior in persuasive technologies and online environments.
  • Explainable AI (XAI) through Argumentation: Exploring how argumentation can provide a layer on top of machine learning to generate more understandable explanations for AI decisions.
  • Argumentation Frameworks and Cognitive Plausibility: Empirically studying which abstract argumentation semantics best predict human evaluation of arguments.
  • Social Argument Mapping: Developing web applications to construct argument maps, aiming to foster more structured and civilized online discussions.

Datasets

Not discussed in the paper. The paper focuses on the conceptual integration of fields and summarizes workshop presentations, rather than presenting empirical results based on specific datasets.

Key Insights and Ideas

  • Argumentation as a Foundation for Cognition: The paper strongly posits that argumentation is an inherent part of human reasoning and can serve as a fundamental basis for higher-level cognition, and consequently, for Human-Centric AI.
  • Human-Centric AI Demands Cognitive Understanding: For AI to be truly human-centric, it must be built on a holistic understanding of the human mind, including its various cognitive faculties at multiple levels.
  • Bridging the Gap Between Low-Level Data and High-Level Reasoning: A significant challenge is enabling AI to move from processing raw data to forming high-level cognitive interpretations, a task where argumentation can play a crucial role.
  • The Importance of Explainability: As AI systems become more complex, their ability to provide clear and cognitively appropriate explanations for their actions is paramount, and argumentation-based approaches show promise here.
  • Ethical AI Requires Understanding Persuasion: Argumentation can help address the ethical requirements of AI by understanding the tension between persuasiveness and ethical behavior, and by potentially mitigating the impact of manipulative marketing practices.
  • Empirical Validation is Key: There's a recognized need for empirical studies to validate the cognitive plausibility of computational argumentation models and to understand the link between human reasoning and AI.

Key Results

Not applicable. This paper is a survey and a discussion of research directions stemming from a workshop, not a report of specific experimental results.

Limitations and Future Work

The paper identifies several key challenges and future directions:

  • Scientific Level Challenges:
    • Defining the precise role of Computational Argumentation in Human-Centric AI and how to realize it.
  • Investigating how argumentation technology can foster meaningful engagement between AI and human users, particularly by leveraging natural language reasoning.
  • Understanding the role of argumentation in bridging System 1 (fast, intuitive) and System 2 (slow, deliberative) cognitive processes.
  • Exploring how argumentation can contribute to the ethical requirements of AI and ensure responsible use.
  • Determining the role of argumentation in achieving Explainable AI, especially in the context of Machine Learning.
  • Designing empirical studies to better understand the link between human reasoning and computational argumentation.
  • Epistemological Level Challenges:
    • Identifying the main synergies between disciplines (AI, Cognitive Science, Linguistics, Philosophy) needed to harness the potential of CA for Human-Centric AI.
  • Harnessing the results and methods from extensive studies of argumentation in other disciplines, particularly those focused on human introspection and debate.

The authors emphasize that these issues are central to Human-Centric AI in general, and their work aims to address these concerns from the perspective of argumentation.

Summary updated

Abstract

This paper examines the interdisciplinary research question of how to integrate Computational Argumentation, as studied in AI, with Cognition, as can be found in Cognitive Science, Linguistics, and Philosophy. It stems from the work of the 1st Workshop on Computational Argumentation and Cognition (COGNITAR), which was organized as part of the 24th European Conference on Artificial Intelligence (ECAI), and took place virtually on September 8th, 2020. The paper begins with a brief presentation of the scientific motivation for the integration of Computational Argumentation and Cognition, arguing that within the context of Human-Centric AI the use of theory and methods from Computational Argumentation for the study of Cognition can be a promising avenue to pursue. A short summary of each of the workshop presentations is given showing the wide spectrum of problems where the synthesis of the theory and methods of Computational Argumentation with other approaches that study Cognition can be applied. The paper presents the main problems and challenges in the area that would need to be addressed, both at the scientific level but also at the epistemological level, particularly in relation to the synthesis of ideas and approaches from the various disciplines involved.
BibTeX
@misc{dietz2021computational,
  author = {Emmanuelle Dietz and Antonis C. Kakas and Loizos Michael},
  title = {{Computational Argumentation and Cognition}},
  year = {2021},
  eprint = {2111.06958},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2111.06958},
}
Figure