Explainable and Ethical AI: A Perspective on Argumentation and Logic Programming

Abstract

In this paper we sketch a vision of explainability of intelligent systems as a logic approach suitable to be injected into and exploited by the system actors once integrated with sub-symbolic techniques.In particular, we show how argumentation could be combined with different extensions of logic programming -namely, abduction, inductive logic programming, and probabilistic logic programming -to address the issues of explainable AI as well as some ethical concerns about AI.

Justifying Social-Choice Mechanism Outcome for Improving Participant Satisfaction

“…With the advent of this field, work on defining the concepts [34], establishing a framework to assess explainability systems [35,36,37] and identifying the best practices for transparency in algorithmic systems [38] have been carried out. Ideas from other fields such as psychology [39], multi-agent systems [40], argumentation [41], planning [42] and logic [43] have also been used to establish definitions for explainability and propose methods to construct explanations.…”

Multi-agent Cooperative Argumentation in Arg2P

“…There, the purpose of multi-agent argumentative dialogues is to let agents reach an agreement on (i) the evaluation of goals and corresponding actions (or plans), and (ii) the adoption of a decentralised strategy for reaching a goal, by allowing agents to refine or revise other agents' goals and defend one's proposals. In this scenario, intelligent behaviours are likely to become associated with the capability of arguing about situations as well as the current state and circumstances, by reaching a consensus on what is happening around and what is needed, and by triggering and orchestrating proper decentralised semantic conversations so as to determine how to collectively act in order to reach a future desirable state [8]. Thus, argumentation [14] and related technologies become a fundamental building block for the design of these systems, thanks to their potential to be an effective communication medium for heterogeneous intelligent agents while enabling a natural form of interaction between users and computational systems, towards explainability features.…”

Arg2P: an argumentation framework for explainable intelligent systems

“…825619). [15]. Argumentation is particularly relevant in the legal context [16], especially when intelligent agents are required not just to agree upon some conclusion, but also to make their reasoning understandable to humans and other agents-to explain themselves [18].…”

“…[15]. Argumentation is particularly relevant in the legal context [16], especially when intelligent agents are required not just to agree upon some conclusion, but also to make their reasoning understandable to humans and other agents-to explain themselves [18]. However, speaking of logic-based technologies can be tricky here: to the best of our knowledge, a logic-based technically-mature environment for argumentation in intelligent systems -both agent-based and accounting for legal aspectsis not available today [13].…”

“…In line with the micro-intelligence definition, the Arg2P framework is a general-purpose inference engine offering a number of diverse reasoning capabilities -not just defeasible reasoning -to be easily exploited for agent-based intelligent systems within a wide range of different applications scenarios, as discussed in [16]. Moreover, its modular architecture offers a neutral ground upon which different technologies can be integrated to improve the explainability of intelligent systems.…”