Artificial intelligence (AI) has become ubiquitous today, influencing diverse sectors like healthcare, finance, and beyond. While various associations, organizations, and private entities have established AI ethics principles, ensuring ethical AI operation presents significant challenges. From tackling biases in datasets to enhancing transparency in decision-making processes, operationalizing ethical principles is a complex endeavor.
In this article, we won’t delve into the foundational concepts of AI ethics, as they have been covered in previous articles. The EU’s Guidelines for Trustworthy AI, delineating four fundamental ethical principles: respect for human autonomy, prevention of harm, fairness, and explicability, serve as a valuable reference for deeper examination. We will focus on the challenges of implementing ethics into AI development and deployment within organizational settings.
Hickman and Petrin (2021) suggests that implementing the EU’s Ethics Guidelines for Trustworthy AI in corporate governance presents multifaceted challenges. Companies may find it challenging to translate broad ethical principles into actionable policies, navigate conflicts between ethics and legal obligations (such as fiduciary duties to prioritize shareholders’ wealth), allocate resources for AI expertise, develop frameworks for applying principles, and align AI practices with organizational values. For instance, while implementing ‘softer’ forms of stakeholder involvement like information disclosure seems feasible, more substantial consultation and participation may encounter resistance from businesses. This underscores the complexity of considering stakeholders’ well-being within the framework of shareholder value maximization.
In the study by Palladino (2023), the challenges of implementing AI ethics are explored, highlighting the potential for distortion when translating these principles into practice. For instance, the paper suggests that the operationalization of AI ethics may be affected by a phenomenon known as “technical solutionism.” This approach involves simplifying complex ethical issues into quantifiable and easy-to-implement solutions that align with the technical community’s default mindset and companies’ goals for efficiency and risk management. By reducing ethical dilemmas to mathematical representations, the real needs and concerns of individuals and society may be overlooked, leading to a skewed implementation of ethical principles. This is consistent with Hickman and Petrin’s argument that AI, as currently conceived, operates based on “formal rationality,” seeking the logically or mathematically correct solution to a dataset within defined constraints. AI’s reliance on formal rationality raises ethical questions about the potential limitations of AI systems in understanding and responding to the richness and complexity of human values, as well as the spontaneity of emotions and social contexts.
Although it’s challenging, some companies have made commendable efforts to integrate ethical principles into AI development and deployment. Ibáñez and Olmeda (2022) investigates how companies apply AI ethics using a qualitative approach. This involved semi-structured interviews and focus groups conducted via Google Meet, with transcriptions completed using Sonix.ai. Data analysis utilized ATLAS.ti 9 software, facilitating organization, text connection, and thematic mapping across exploration, encoding, and interpretation stages. The study revealed intriguing insights; for example, companies face challenges in implementing the fairness principle. However, some are actively addressing this by prioritizing vulnerable groups, ensuring that profiling respects their rights, and revising or eliminating AI models if discriminatory outcomes arise.
In navigating the complexities of operationalizing AI ethics, new approaches have emerged. For instance, Brusseau (2023) proposes a shift in focus from abstract principles to practical experiences, highlighting the significance of deriving ethical insights from the real-life experiences of AI designers working on concrete human issues. Instead of applying abstract principles downward to practice, the paper advocates for cycling ethical insights upward to influence theoretical debates surrounding AI ethics. The paper argues that ethical insights should stem from real-world scenarios and dilemmas encountered by AI developers, including those in healthcare AI or algorithmic startup companies. This approach enables a more effective examination of key questions in AI ethics, such as balancing explainability and accurate performance, prioritizing trustworthiness versus reliability in AI systems, and determining the orientation of AI ethics towards user protection or innovation stimulation.
Hagendorff (2022) proposes a virtue-based approach to implementing AI ethics. This approach differs from conventional methods like deontology and consequentialism by prioritizing the development of virtuous character traits over strict rule-following or outcome maximization. While deontological ethics emphasizes duty and adherence to moral principles, and consequentialism focuses on the consequences of actions, virtue ethics places importance on cultivating virtues such as care, honesty, and justice to guide ethical decision-making. This approach considers the motivations behind actions, the role of environmental and intrapersonal factors, practical wisdom in decision-making, and ongoing self-improvement. In the context of AI ethics, the virtue-based approach seeks to foster a culture of integrity, trustworthiness, and ethical responsibility by promoting the cultivation of virtues among practitioners and organizations. While traditional principled approaches in AI ethics may struggle with practical implementation due to abstract or context-insensitive guidelines, the virtue-based approach offers a more practical framework for ethical decision-making. By emphasizing the cultivation of specific virtues within organizations, this approach shows potential for translating ethical principles into tangible behaviors and norms.
At the societal level, implementing AI ethics requires a collaborative and multi-stakeholder approach that engages researchers from various disciplines (Morley et. al., 2020). By bringing together experts from fields such as ethics, computer science, law, sociology, and more, we can establish a shared language, common goals, and relevant tools and methodologies to address the complex ethical challenges posed by artificial intelligence. This interdisciplinary collaboration allows for a comprehensive examination of AI technologies’ societal impacts and ethical implications, leading to more informed decision-making and policy development. Moreover, it promotes the integration of diverse perspectives and ensures that ethical considerations are embedded into the design, development, and deployment of AI systems.
In conclusion, operationalizing AI ethics is a multifaceted endeavor that requires navigating through various complexities. As the field of artificial intelligence continues to evolve, so too must our approach to ethical considerations surrounding its development and deployment. By adopting innovative strategies, such as shifting the focus from abstract principles to practical experiences and emphasizing the cultivation of virtues within organizations, we can strive toward a more ethical and responsible AI landscape. Through collaborative efforts and continuous reflection, we have the opportunity to shape AI technologies that not only adhere to ethical standards but also contribute positively to society.
References:
Brusseau, J. (2023). From the Ground Truth Up: Doing AI Ethics from Practice to Principles. AI & Society, 38(4), 1651–1657. https://doi.org/10.1007/s00146-021-01336-4
Hagendorff, T. (2022). A Virtue-Based Framework to Support Putting AI Ethics into Practice. Philosophy & Technology, 35(3), 1–24. https://doi.org/10.1007/s13347-022-00553-z
Hickman, E., & Petrin, M. (2021). Trustworthy AI and Corporate Governance: The EU’s Ethics Guidelines for Trustworthy Artificial Intelligence from a Company Law Perspective. European Business Organization Law Review, 22(4), 593–625. https://doi.org/10.1007/s40804-021-00224-0
Ibáñez, J. C., & Olmeda, M. V. (2022). Operationalising AI Ethics: How Are Companies Bridging the Gap between Practice and Principles? An Exploratory Study. AI & Society, 37(4), 1663–1687. https://doi.org/10.1007/s00146-021-01267-0
Morley, J., Floridi, L., Kinsey, L., & Elhalal, A. (2019). From What to How: An Initial Review of Publicly Available AI Ethics Tools, Methods and Research to Translate Principles into Practices. Science and Engineering Ethics, 26(4). https://doi.org/10.1007/s11948-019-00165-5
Palladino, N. (2023). A “Biased” Emerging Governance Regime for Artificial Intelligence? How AI Ethics Get Skewed Moving from Principles to Practices. Telecommunications Policy, 47(5), N.PAG. https://doi.org/10.1016/j.telpol.2022.102479
Like (0)