Ethics Pen-Testing: A Proactive Approach to AI for the Common Good

Why It Matters

As AI systems are increasingly deployed for societal benefit, understanding and mitigating potential ethical risks is paramount. This research suggests a structured approach to uncover and address these challenges before they manifest in real-world applications, leading to more robust and beneficial AI solutions.

Key Finding

Designing AI for the 'Common Good' is difficult because it's hard to agree on what 'good' means, who defines it, and what unintended consequences might arise. The study proposes using 'ethics pen-testing,' similar to security testing, to find and fix these problems proactively.

Key Findings

• Defining and achieving the 'Common Good' with AI is complex and fraught with challenges related to problem framing, stakeholder perspectives, knowledge integration, and unforeseen side effects.
• Existing ethical guidelines for AI often lack practical methods for identifying and addressing these complexities.
• Adopting 'attack' methodologies, such as ethics pen-testing, can serve as a valuable tool for uncovering potential ethical issues in AI design.

Research Evidence

Aim: How can 'attack' methodologies from computer science be adapted to proactively identify and mitigate ethical challenges in AI systems designed for the Common Good?

Method: Exploratory research and conceptual framework development

Procedure: The research analyzes the challenges of defining and achieving the 'Common Good' with AI through four key questions: problem definition, stakeholder identification, knowledge integration, and side effect analysis. It then proposes 'ethics pen-testing' as a method to address these challenges, drawing parallels with security penetration testing.

Sample Size: 99 contributions to recent conferences

Context: AI for Social Good, Data Science for Social Good

Design Principle

Proactively probe for ethical vulnerabilities in AI systems to ensure they align with societal benefit.

How to Apply

Before launching an AI for social good project, conduct a simulated 'attack' scenario where a team actively tries to find ways the AI could cause harm or fail to deliver its intended benefit.

Limitations

The proposed 'ethics pen-testing' is a conceptual framework and requires further development and empirical validation.

Student Guide (IB Design Technology)

Simple Explanation: When you're building AI to help people, it's easy to miss problems. This research suggests pretending to 'attack' your own AI design to find ethical issues before they cause real harm.

Why This Matters: This research highlights the importance of thinking critically about the ethical implications of your design choices, especially when aiming for a positive societal impact.

Critical Thinking: To what extent can 'ethics pen-testing' truly anticipate all potential negative impacts of an AI system, given the dynamic and complex nature of societal interactions?

IA-Ready Paragraph: This research underscores the critical need for proactive ethical evaluation in AI design for social good. By adapting computer science's 'attack' methodologies, such as 'ethics pen-testing,' designers can systematically identify and mitigate potential pitfalls related to problem definition, stakeholder perspectives, knowledge integration, and unforeseen side effects, thereby enhancing the likelihood of AI systems truly contributing to the Common Good.

Project Tips

• When defining the 'problem' your AI aims to solve, consider who benefits and who might be excluded or harmed.
• Think about how your AI might be misused or have unintended negative consequences, and plan ways to prevent them.

How to Use in IA

• Reference this study when discussing the ethical considerations of your design project, particularly if it aims to address a social issue.
• Use the concept of 'ethics pen-testing' to justify a proactive approach to identifying and mitigating risks in your design process.

Examiner Tips

• Demonstrate an understanding of the complexities in defining 'the Common Good' and how this impacts AI design.
• Show evidence of proactive ethical risk assessment beyond superficial checks.

Independent Variable: Adoption of 'ethics pen-testing' methodology

Dependent Variable: Effectiveness of AI in contributing to the Common Good, identification of ethical pitfalls

Controlled Variables: Domain of AI for Social Good, specific AI system being designed

Strengths

• Introduces a novel, actionable methodology ('ethics pen-testing') for addressing ethical challenges in AI.
• Highlights practical shortcomings in current AI for social good research and practice.

Critical Questions

• How can the 'Common Good' be objectively measured and validated when designing AI systems?
• What are the ethical implications of defining 'attack' vectors for AI systems, even for the purpose of ethical testing?

Extended Essay Application

• Investigate the effectiveness of different 'ethics pen-testing' strategies on a specific AI prototype designed for a social issue.
• Compare the outcomes of a design process that includes 'ethics pen-testing' versus one that does not.

Source

AI for the Common Good?! Pitfalls, challenges, and ethics pen-testing · SHILAP Revista de lepidopterología · 2019 · 10.1515/pjbr-2019-0004