HOME / NEWS & EVENTS / BLOG: AI ETHICS FOR BUSINESS LEADERS /

Post #5: Reimagining AI Ethics, Moving Beyond Principles to Organizational Values

It’s difficult to find an organization that hasn’t publicly stated its adherence to some externally formulated set of AI principles. As Emre Kazim and Adriano Soares Koshiyama highlight in their insightful piece, “A High-Level Overview of AI Ethics”, in 2020 a mere 80 organizations across academia, NGOs, civil society, and commercial sectors had issued statements affirming their adherence to these external principles. (Special acknowledgement to Kazim and Koshiyama for their persuasive work, which has deeply influenced the ideas presented in this essay). With the explosion of generative AI experimentation and adoption, there are now hundreds, if not thousands, of organizations embracing this principles-based approach to steer their AI endeavors.

However, as societies grapple with the intricacies of AI ethics, it is becoming increasingly apparent that principles alone are insufficient. In this essay, we advocate for a paradigm shift from principles-based AI ethics towards a values-based approach grounded in organizational values. By aligning ethical decision-making with the fundamental values of the organization, we believe that businesses can more effectively navigate the ethical challenges of AI.

What do we mean by a principles-based approach

In a principles-based approach to AI ethics, individuals and organizations derive ethical guidance from a set of overarching principles designed to inform AI system decision-making and behavior. Take for instance, the Artificial Intelligence Risk Management Framework issued by the National Institute of Standards and Technology (NIST) within the U.S. Department of Commerce, which outlines a set of trustworthy AI characteristics (principles) such as transparency, accountability, and fairness. These principles are formulated and articulated to foster ethical AI development and adoption, serving as guiding beacons for a diverse array of stakeholders, spanning NGOs, academics, government agencies, and commercial enterprises. Academic institutions may incorporate these principles into their research ethics frameworks, ensuring that ethical considerations are embedded within their scholarly pursuits, but as we demonstrate below, it is not that simple and real barriers to adoption still exist.

What are the problems plaguing a principles-based approach to AI ethics?

While principles-based AI ethics has been adopted with positive outcomes, it is not without its limitations. Here are the inherent challenges of this approach:

  • Lack of clarity: The articulated principles are often vague and lack specificity, making them difficult to interpret without actionable guidance for addressing complex ethical dilemmas. Without clear guidelines, organizations struggle to make informed decisions when faced with ethical challenges.
     
  • Incongruence: Principles articulated in a single source are often contradictory on their face. For example, common definitions of principles such as transparency and privacy will appear contradictory. This failure to recognize inherent tensions will surely influence the practical implementation of AI ethics.
     
  • Lack of global consensus: There is no consensus in the world for a single set of responsible AI principles. In fact, it’s difficult to imagine a universal set of principles to adequately serve as a global standard given the cultural variability in ethics.
     
  • Rigidity: Principles-based approaches can be rigid, requiring strict adherence to predefined, externally developed rules. This rigidity may limit adaptability and hinder organizations from responding effectively to unique ethical situations. A principles-based approach may not accommodate the dynamic and context-dependent nature of ethical decision-making.
     
  • Lack of enforcement or accountability: Stated principles often lack accountability measures. Without enforceable regulations, organizations need clear commitment and operational controls to hold people accountable to ethical AI practices.
     
  • Response to Novelty: As AI technologies evolve, new ethical challenges emerge. Principles-based approaches may struggle to address these novel or unprecedented scenarios, leaving organizations ill-prepared to navigate evolving ethical landscapes.
     
  • Stakeholder Engagement: Principles-based approaches may prioritize compliance with externally developed ideals over meaningful engagement with an organization’s stakeholders. Limited stakeholder engagement may overlook valuable insights from diverse perspectives and fail to address the broader societal impacts of AI.
     
  • Inappropriate combination of ethical and non-ethical values: Reid Blackman highlights this issue in his excellent work Ethical Machines (Harvard Business Review Press, 2022). Many principles-based approaches blend responsible or trustworthy AI ideals with ethical concerns. This mixing of values can diminish the importance of critical AI ethics issues.

Introducing a values-based approach

In response to the shortcomings of principles-based AI ethics, we advocate for an organizational values-based approach. Unlike principles, which provide general guidelines typically established by an external organization, values serve as the foundation of an organization's ethical culture. By embedding ethical values into the fabric of the organization, businesses can foster a culture of integrity, trust, and accountability.

A values-based approach to AI ethics emphasizes the importance of aligning ethical decision-making with the core values of the organization. These values not only guide individual behavior but also shape organizational policies, practices, and strategies. By prioritizing values such as integrity, diversity and inclusion, social responsibility, human-centered design, and respect for human dignity, organizations can ensure that their AI initiatives reflect their ethical commitments.

Advantages of a values-based approach

The shift towards a values-based approach offers several advantages over traditional principles-based methods. Values-based approaches are inherently flexible, allowing organizations to adapt ethical considerations to diverse situations and contexts. This flexibility enables agile responses to emerging ethical challenges and promotes ethical innovation. By aligning ethical decision-making with organizational values, businesses can ensure consistency and coherence in their approach to AI ethics. This alignment fosters trust among stakeholders and reinforces the organization's commitment to ethical conduct. Values-based approaches also encourage innovation and creativity in ethical decision-making. By empowering employees to apply ethical values creatively, organizations can explore new solutions to complex ethical dilemmas and drive ethical innovation.

Further, values-based approaches prioritize stakeholder engagement and inclusivity, ensuring that diverse perspectives are considered in ethical decision-making processes. By actively involving stakeholders, organizations can enhance the legitimacy and social responsibility of their AI initiatives. In short, values-based approaches enable organizations to respond promptly to evolving ethical challenges and opportunities. By embedding ethical values into their organizational culture, businesses can develop the resilience and agility needed to navigate the dynamic landscape of AI ethics.

Transitioning from principles to values

Transitioning from a principles-based method to an organizationally aligned values-based approach requires careful consideration and strategic planning. Here are six practical steps for organizations to make this transition:

  • Identify core values: Begin by assessing and define the organization's core values, considering its mission, vision, and culture. These core values should serve as the foundation of the organization's ethical commitments and guide its decision-making processes, both within and outside the realm of AI.
     
  • Align these core values with AI applications: Next, align the organization’s core values with its AI systems and applications. Determine the organization’s boundaries of tolerance for each AI ethics issue, clearly articulating expectations throughout the enterprise. Leaders must identify actions that fall outside these boundaries and establish guidelines accordingly.
     
  • Embed values in the organizational culture: Integrate ethical values into the organization's culture and practices, ensuring that they are reflected in policies, procedures, and day-to-day operations. Leadership plays a crucial role in modeling ethical behavior and fostering a culture of integrity and accountability throughout the organization.
     
  • Provide training and education: Offer ethical decision-making training and education to employees, emphasizing the importance of aligning actions with organizational values. Training programs should empower employees to recognize ethical dilemmas, apply ethical values creatively, and seek guidance when necessary. This training should extend to all levels of the organization, including the board and executive leadership.
     
  • Establish feedback mechanisms: Implement mechanisms for feedback and continuous improvement, enabling employees and other stakeholders to raise ethical concerns and provide input on decision-making processes. These mechanisms should be accessible, transparent, and responsive, ensuring the ethical values are upheld across the organization.
     
  • Evaluate and Iterate: Regularly evaluate and iterate on the organizational values-based approach, monitoring its effectiveness and identifying areas for improvement. Solicit feedback from a diverse range of stakeholders, assess the impact of ethical decisions, and adapt the approach as needed to maintain alignment with organizational values.

Here is a succinct comparison of a principles versus values-based approach to AI ethics:

Factors Principles-Based Approach Values-Based Approach
1.  Foundation Relies on ethical principles or rules from outside the organization Grounded in values and beliefs that are core to the organization
2.  Flexibilty Lacks flexibility in addressing complex ethical dilemmas Allows for context-specific ethical decision-making
3.  Clarity Provides vague guidance for ethical behavior Emphasises ethical organizational values with clear examples of virtuous behavior through a boundary of tolerance AI ethics exercise
4.  Tension Resolve Many principles are contradictory, thereby creating tension without guidance on how to resolve Promotes nuanced ethical decision-making and tools to resolve ethical tensions
5. Accountabilty Focuses on adherence to principles and rules Emphasizes organizational accountability
6.  Stakeholder Resolve Emphasizes compliance with external regulations and standards Values engagement with a wide variety of stakeholders to understand and act on a diverse set of perspectives
7.  Organizational Culture May foster a compliance-oriented culture Promotes a culture of ethical awareness and integrity

 

In conclusion, the transition from a principles-based approach to an organizationally aligned values-based approach represents a significant paradigm shift in AI ethics. By prioritizing organizational values over prescriptive principles, businesses can foster a culture of ethical integrity, trust, and accountability. This values-based approach not only enhances the ethical governance of AI but also reinforces the organization's commitment to responsible innovation and social responsibility, meeting the fiduciary duty to shareholders while benefiting society.

We invite organizations to join us in reimagining AI ethics and embrace a values-based approach to ethical decision-making. By aligning AI initiatives with organizational values, we can create a future where AI technologies serve the common good and contribute to a more ethical and sustainable society.

Is your organization plagued by some or all of the above problems with an AI ethics principle-based approach? We’d love to hear from you and engage in a dialogue about why and how to make the transition to a values-based approach as outlined in this essay.

___

We have heard from so many readers of this Business AI Ethics blog – thank you! If you share our passion for this work, we welcome your input and collaboration. Until next week.

The Business AI Ethics research team is part of the Edmond & Lily Safra Center’s ongoing effort to promote the application of ethics in practice. Their research assists business leaders in examining the promise and challenges of AI technologies through an ethical lens. Views expressed in these posts are those of the author(s) and do not imply endorsement by ELSCE.