🔬 Research Summary by Marialena Bevilacqua, a Ph.D. student in Analytics at the University of Notre Dame’s Mendoza College of Business studying the intersection of AI technologies and moral judgments.
[Original paper by Marialena Bevilacqua, Nicholas Berente, Heather Domin, Brian Goehring, and Francesca Rossi]
Overview: Organizations that are increasingly using, building, or incorporating artificial intelligence (AI) technologies realize that they need to invest in ways to ensure that they are using these powerful technologies ethically. But what is the organizational return on such investments? This paper proposes a methodology for understanding the return on organizational investments in AI ethics called the Holistic Return on Ethics (HROE) framework. This framework provides organizations with an approach to assess and justify AI ethics investments.
AI technologies are currently created and deployed rapidly and are becoming more powerful with every innovation. While an organization’s adoption of such technologies may provide productivity, efficiency, and opportunity benefits, they may also result in ethical issues, such as those associated with bias, fairness, and misinformation. Organizations must invest in AI ethics initiatives to deal with such unintended consequences. But organizations may question the value of such investments by asking:
What will we receive in return, and how do we quantify this return?
Motivated by this lack of organizational understanding and a methodology for quantifying the possible returns, we propose a framework for establishing the return on investment (ROI) for organizational investments in AI ethics. We summarize the sorts of investments that organizations could make in AI ethics, characterize three types of ROI, and relate them in an original framework. Our work is the first step in constructing methods for measuring the return on AI ethics investments.
The scale and pace at which AI technologies are being deployed give rise to concerns about ethical issues relating to aspects of these technologies, such as their fairness, explainability, and transparency. AI stakeholders in the past decade have worked to identify and mitigate these issues through complementary mechanisms such as principles, guidelines, or regulations. While creating and further establishing such efforts may be the first step in ensuring the ethical creation, deployment, and utilization of AI technologies, these efforts lead to significant organizational investment. Since AI ethics investments have only recently gained attention, related investments must be justified to gain full support within an organization. Typically, organizations justify technology investments using the classic return on investment, or ROI, metric. This research presents three distinct approaches to calculating the return on AI ethics investments.
ROI is traditionally calculated by dividing a measure of economic return by the cost of an investment. At the organizational level, the organization is considered as the unit of analysis, and managers and shareholders are the primary stakeholders. An organization’s ROI is applied to assess particular expenditures, such as new subunits, projects, or capital equipment.
But how do you calculate ROI when investments possess an organizational value that does not have a clear, direct, and traceable impact on financial outcomes? This value, which we refer to as intangible value, may significantly impact key stakeholders and the organization’s performance, yet is not easily quantifiable. There are numerous forms of intangible ROI. Examples include corporate social responsibility (CSR) and its associated areas such as environmental, social, and governance (ESG), organizational culture, and customer satisfaction.
Real Options and ROI as Capabilities
Real options are similar to financial options in that organizations make small investments that generate future flexibility (McGrath et al., 2004). Real options reasoning promotes proactive learning that builds the capabilities necessary for organizations to execute novel projects in an uncertain future. Additionally, organizations can incrementally approach major investments in disciplined stages, sequencing smaller investments to ensure that opportunities with significant upside are pursued despite initial uncertainty. Ultimately, real options reasoning permits those in decision-making positions to maintain fiduciary responsibility even when making sometimes uncertain, aggressive investments.
A Holistic Framework
AI ethics principles and guidelines are indeed important and a good start, but organizations need to do more to ensure that they stay abreast of new issues; they must make investments that ensure the ethical use of AI technologies. AI ethics investments are important, yet the concept is new, unfamiliar, and costly. Therefore, we propose that organizations interested in determining the holistic impact of their investments in AI ethics apply these three ROI measures. Accordingly, we present three paths to understanding the impact of investments in AI ethics: the direct path through economic return (traditional ROI) and the indirect path through reputation (intangible ROI) and capabilities (real options). These paths form the Holistic Return on Ethics (HROE) framework.
The traditional financial ROI measures the direct relationship between investment in AI ethics and its consequent economic return in terms of cost savings, revenue generation, or cost of capital reduction. Intangible ROI captures the proximal reputational impacts of investments in AI ethics, such as stakeholder trust, which will eventually lead to proximal economic ROI. Finally, we consider the optionality of AI ethics investments in terms of building proximal AI ethics-related capabilities, which provide options to the organization for subsequent flexibility and cost savings. To operationalize our framework, we provide formal models with hypothetical illustrations. While these illustrations focus on AI, the framework is also useful for determining ethical investments required for other types of technologies.
Between the lines
AI ethics investments differ from typical technology investments due to the novelty and unpredictability of AI technologies. Since making such investments is currently unfamiliar to many organizations, it is crucial that they have, and can easily adopt, a methodology for justifying their AI ethics investments. Our unique framework fulfills this need. By acknowledging the existence of an investment’s economic, reputational, and capabilities-related impacts, the true total return on investments in AI ethics can now be identified, addressed, and referenced. Operationalizing this framework provides organizations with greater clarity of the benefits they can incur, relative to each stakeholder, through investments in AI ethics. We hope that our work provides researchers and practitioners with a starting point for efforts that enable quantifying the impact of AI ethics investments. This work and related future work collectively take a step in the right direction of ensuring ethical creation, deployment, and use of all AI technologies.