The world is seeing extraordinary advances in artificial intelligence. There are new applications in finance, defense, health care, criminal justice, education, and other key industries. Algorithms are improving fraud detection, health diagnoses, voice recognition systems, as well as advertisement targeting in e-commerce and political campaigns. Yet, at the same time, there is growing concern regarding the ethical values embedded within AI and the extent to which algorithms respect basic human rights. Ethicists worry about a lack of transparency, poor accountability, rising inequity, and untethered bias in automated tools.
In this situation, it is of the utmost importance for large global corporations to take several steps to ensure that AI ethics are taken seriously, and with a sense of urgency. The first step could be to integrate organizational and procedural mechanisms that may help cope with complex ethical dilemmas. Having clear processes and avenues for deliberation will help prioritize and deal with particular problems. Here, I present six organizational and procedural mechanisms that large multinational corporations could institute in order to help resist unethical AI applications.
Hiring ethicists. It is crucial for companies to embed ethicists within their staffs to help them question whether potential AI applications are aligned with public interest and the ethical needs of a globalizing world. Giving these individuals a seat at the table will help to ensure that ethics are taken seriously and appropriate deliberations take place when ethical dilemmas arise, which is likely to happen on a regular basis. In addition, they can assist corporate leadership in creating an AI ethics culture and supporting corporate social responsibility within their organizations. These ethicists should make annual reports to their corporate boards outlining the issues they have addressed during the preceding year and how they resolved ethical aspects of those decisions.
Having an AI code of ethics. Companies should have a formal code of ethics that lays out their principles, processes, and methods for handling ethical issues during AI design and development. Those codes should be made public on the firm’s websites so that stakeholders and external parties can see how the company thinks about ethical issues and the choices its leaders have made in dealing with emerging technologies
Instituting AI review boards. Businesses should set up internal AI review boards that evaluate product lines and are integrated into a company’s decision-making. These boards should include a representative cross-section of the firm’s stakeholders and be consulted on AI-related decisions. Their portfolio should include development of particular product lines, the procurement of government contracts, and procedures used in developing AI products.
Requiring AI audit trails. Companies should have AI audit trails that explain how particular algorithms were put together or what kinds of choices were made during the development process. This can provide some degree of “after-the-fact” transparency and explainability to outside parties. Such tools would be especially relevant in cases that end up under litigation and need to be elucidated to judges or juries in case of consumer harm or civil suits. Since product liability law is likely to be the governing force in adjudicating AI harm, it is necessary to have audit trails that provide both external transparency and explainability.
Implementing AI training programs. Firms should have AI training programmes that not only address the technical aspects of development, but the ethical, legal, and societal ramifications. This would help software developers understand that their products should not assume or embed their own individual ethical values, but are part of a broader society that stretches around the globe with a stake in AI development. AI goes beyond the development of traditional product lines with narrow social implications. With its potential to distort basic human values and rights, it is crucial to train people on how to think about AI through all phases of its development. Another integral aspect to AI training programs would be teaching software designers about the perils of using poorly curated datasets to train their algorithms, providing them with the tools to not only identify bad training data, but also to provide them solutions and access to wider datasets curated with the assistance of ethicists.
Providing a means of remediation for AI damages or harm. There should be a means of remediation in cases where AI deployment result in consumer damages or harm. This could be through legal cases, arbitration, or some other negotiated process. This would allow those hurt by AI to address the problems and rectify the situation. Having clear procedures in place will help with this technology’s unintended consequences.
Public opinion survey data indicates that there is substantial citizen support for the establishment of such organizational and procedural mechanisms. An August 2018 Brookings survey found that: 55% of U.S. respondents support hiring corporate ethicists; 67% of respondents favor companies having a code of ethics; 66% believe that companies should have an AI review board; 62% think there should be an AI audit trail that shows how software designers (as well as the resulting AI product) makes decisions; 65% favor the implementation of AI training programs for company staff; and 67% want companies to have remediation procedures when AI inflicts harm or damage to humans.
The strong public support for these steps indicates people understand the ethical risks posed by artificial intelligence and emerging technologies, as well as the need for significant action by technology-based organizations. Individuals want companies to have a sense of urgency when it comes to taking meaningful action to protect them from rising inequity, bias, poor accountability, inadequate privacy protection, and a lack of transparency.
By embedding ethics into their infrastructures, corporations will be in a better position to develop, refine, and revise normative frameworks that can respond to new dynamics or emerging ethical challenges pertaining to AI, and provide such a standard to AI technology developers in start-ups with less capital and means. Such an effort of learning and tailoring ethics to applied AI will ultimately benefit discussions on the technology and its global governance.
About the Author
Darrell M. West is the vice president and director of governance studies and director of the center for technology innovation at the Brookings Institution. He holds the Douglas Dillon Chair in governance studies, and has written about technology policy, mass media, and campaigns and elections in the United States. He is Editor in Chief of the Brookings technology policy blog, TechTank.
Read all of the AI & Global Governance insights here.
This article has been prepared by Darrell M. West as a contributor to AI & Global Governance. The opinions expressed in this article are those of the author and do not necessarily reflect those of the Centre for Policy Research, United Nations University, or its partners.