# Ethics of AI: A Systematic Literature Review of Principles and Challenges

**Authors:** Arif Ali Khan, Sher Badshah, Peng Liang, Bilal Khan, Muhammad Waseem, Mahmood Niazi, Muhammad Azeem Akbar  
**arXiv:** [2109.07906](https://arxiv.org/abs/2109.07906)  
**Date:** 2021-09-12  
**Summary page:** https://www.summarizepaper.com/en/paper/2109.07906v1/

## Key points

- AI implementation has brought about a technological revolution that is changing science and society
- Societal debates about the principles and policies guiding the use and deployment of AI systems have emerged
- Various organizations have developed ad hoc committees to draft policy documents for AI ethics
- Existing guidelines are not effective or adopted in practice
- Vakkuri et al. used the accountability, responsibility, and transparency (ART) framework to develop a conceptual model exploring ethical considerations in the AI environment
- Empirical validation through multiple case studies revealed that ethical principles were still not being put into practice despite common concepts being considered when developing AI systems
- A systematic literature review (SLR) study was conducted to investigate further agreement on the significance of AI principles and identify challenging factors negatively impacting their adoption
- 22 global convergence ethical principles were identified with transparency, privacy, accountability, and fairness as the most common ones
- 15 challenges were identified with lack of ethical knowledge and vague principles reported as significant obstacles for considering ethics in AI
- The findings provide preliminary inputs for proposing a maturity model that assesses the ethical capabilities of AI systems and provides best practices for further improvements
- There are currently no tools available to measure how well an organization is implementing its AI ethics principles
- It is necessary to develop such tools to ensure that ethical considerations are integrated into AI systems

## Summary

The implementation of AI or machine intelligence concepts has brought about a technological revolution that is changing both science and society. This transformation has sparked important societal debates about the principles and policies that guide the use and deployment of AI systems. Various organizations have developed ad hoc committees to draft policy documents for AI ethics, resulting in the development of AI policies and guide documents by technology corporates such as SAP and Google, Amnesty International, the Association of Computing Machinery (ACM), Access Now, and others. However, recently published studies report that existing guidelines are not effective or adopted in practice. To address this challenge, Vakkuri et al. used the accountability, responsibility, and transparency (ART) framework to develop a conceptual model exploring ethical considerations in the AI environment. The model was empirically validated through multiple case studies which revealed that while some common concepts were considered when developing AI systems, ethical principles were still not being put into practice. To investigate further agreement on the significance of AI principles and identify challenging factors negatively impacting their adoption, a systematic literature review (SLR) study was conducted. The SLR data revealed 22 global convergence ethical principles with transparency, privacy, accountability, and fairness identified as the most common ones. Additionally, 15 challenges were identified with lack of ethical knowledge and vague principles reported as significant obstacles for considering ethics in AI. The findings from this study provide preliminary inputs for proposing a maturity model that assesses the ethical capabilities of AI systems and provides best practices for further improvements. While there are currently no tools available to measure how well an organization is implementing its AI ethics principles, it is necessary to develop such tools to ensure that ethical considerations are integrated into AI systems. Overall, this study highlights the need for continued efforts to develop effective techniques and tools for successfully scaling AI ethics guidelines into practice.