[Paper Review] Socially Responsible AI Algorithms: Issues, Purposes, and Challenges

Lu Cheng, Kush R. Varshney+1 more|arXiv (Cornell University)|Jan 1, 2021

Ethics and Social Impacts of AI211 references35 citations

TL;DR

This survey defines Social Responsibility of AI, proposes a four-tier pyramid of AI responsibilities, and presents a systematic framework for Socially Responsible AI Algorithms (SRAs) that address ethics, transparency, safety, and societal impact.

ABSTRACT

In the current era, people and society have grown increasingly reliant on artificial intelligence (AI) technologies. AI has the potential to drive us towards a future in which all of humanity flourishes. It also comes with substantial risks for oppression and calamity. Discussions about whether we should (re)trust AI have repeatedly emerged in recent years and in many quarters, including industry, academia, healthcare, services, and so on. Technologists and AI researchers have a responsibility to develop trustworthy AI systems. They have responded with great effort to design more responsible AI algorithms. However, existing technical solutions are narrow in scope and have been primarily directed towards algorithms for scoring or classification tasks, with an emphasis on fairness and unwanted bias. To build long-lasting trust between AI and human beings, we argue that the key is to think beyond algorithmic fairness and connect major aspects of AI that potentially cause AI's indifferent behavior. In this survey, we provide a systematic framework of Socially Responsible AI Algorithms that aims to examine the subjects of AI indifference and the need for socially responsible AI algorithms, define the objectives, and introduce the means by which we may achieve these objectives. We further discuss how to leverage this framework to improve societal well-being through protection, information, and prevention/mitigation.

Motivation & Objective

• Define Social Responsibility of AI across principles, means, and objectives.
• Propose a pyramid of AI responsibilities: functional, legal, ethical, and philanthropic.
• Develop a systematic framework for Socially Responsible AI Algorithms (SRAs) and their roles.
• Identify causes and subjects of socially indifferent AI and outline mechanisms to achieve SRAs.

Proposed method

• Formally define Social Responsibility of AI with three dimensions (principles, means, objectives).
• Adapt Carroll’s CSR pyramid to AI with four responsibilities (functional, legal, ethical, philanthropic).
• Propose the SRAs framework detailing its essentials, roles, and feedback loop from users.
• Survey and integrate topics (fairness, transparency, accountability, safety) under a unified SRAs perspective.
• Analyze causes of indifference (formalization, measurement errors, bias, data misuse, correlation vs causation).
• Outline means to achieve SRAs through interpretability, adversarial ML, causal learning, and uncertainty quantification.

Experimental results

Research questions

• RQ1 What constitutes Social Responsibility of AI and how can it be formalized?
• RQ2 How can the four-part pyramid (functional, legal, ethical, philanthropic) guide AI development?
• RQ3 What is the systematic framework for SRAs and how do its components interact to protect, inform, and mitigate negative AI impacts?
• RQ4 What are the root causes of socially indifferent AI, and how can SRAs address them?

Key findings

• We provide a formal definition of Social Responsibility of AI across principles, means, and objectives.
• We introduce a four-level pyramid of AI responsibilities: functional, legal, ethical, and philanthropic.
• We propose a comprehensive SRAs framework linking its essentials, roles, and user feedback to societal outcomes.
• We identify key causes of social indifference in AI (formalization, measurement errors, bias, data misuse, correlation vs causation) and discuss strategies to address them.
• We discuss objectives (fairness, transparency, safety) and means (interpretability, adversarial ML, causal learning, uncertainty quantification) for SRAs.
• The survey highlights open problems and challenges, including the need for new AI ethics principles and policies.