Implementing Responsible AI: Tensions and Trade-Offs Between Ethics Aspects

Conrad Sanderson,David Douglas,Qinghua Lu

DOI: https://doi.org/10.1109/IJCNN54540.2023.10191274

2024-09-06

Abstract:Many sets of ethics principles for responsible AI have been proposed to allay concerns about misuse and abuse of AI/ML systems. The underlying aspects of such sets of principles include privacy, accuracy, fairness, robustness, explainability, and transparency. However, there are potential tensions between these aspects that pose difficulties for AI/ML developers seeking to follow these principles. For example, increasing the accuracy of an AI/ML system may reduce its explainability. As part of the ongoing effort to operationalise the principles into practice, in this work we compile and discuss a catalogue of 10 notable tensions, trade-offs and other interactions between the underlying aspects. We primarily focus on two-sided interactions, drawing on support spread across a diverse literature. This catalogue can be helpful in raising awareness of the possible interactions between aspects of ethics principles, as well as facilitating well-supported judgements by the designers and developers of AI/ML systems.

Computers and Society,Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the potential tensions and trade - offs among different ethical principle aspects in the process of implementing Responsible AI. These ethical principle aspects include privacy, accuracy, fairness, robustness, interpretability and transparency. Although these principles are designed to guide the development and use of AI/ML systems, there may be conflicts among them, making it difficult for developers to follow these principles. For example, improving the accuracy of an AI/ML system may reduce its interpretability; increasing privacy protection may affect the system's accuracy. Therefore, by compiling and discussing a catalogue of 10 significant tensions, trade - offs and other interactions, the paper helps to raise awareness of the possible interactions among these aspects and promotes designers and developers to make more reasonable judgments for better implementation of responsible AI.

The main contributions of the paper are as follows:
1. **Definition and Understanding**: First, it defines six basic aspects in AI ethical principles - accuracy, robustness, fairness, privacy, interpretability and transparency, and provides a detailed description of each aspect.
2. **Interaction Analysis**: Based on the support of existing literature, it explores the pairwise interactions among these aspects, such as the trade - off between accuracy and robustness, the conflict between fairness and privacy, etc.
3. **Practical Guide**: By providing a detailed interaction catalogue, it helps designers and developers of AI/ML systems to be aware of the existence of these interactions and make more balanced decisions in practice.

In short, this paper aims to provide theoretical support and practical guidance for more responsible and ethical AI applications through in - depth analysis and discussion of the complex relationships among AI ethical principle aspects.