Effective and Trustworthy Implementation of AI Soft Law Governance
Citation: C. I. Gutierrez, G. E. Marchant and K. Michael, "Effective and Trustworthy Implementation of AI Soft Law Governance," in IEEE Transactions on Technology and Society, vol. 2, no. 4, pp. 168-170, Dec. 2021, doi: 10.1109/TTS.2021.3121959.
Image by WilliamCho from Pixabay
This special issue (in cooperation with the IEEE Technology and Society Magazine, December 2021) is dedicated to examining the governance of artificial intelligence (AI) through soft law. This kind of law is considered “soft” as opposed to “hard” because it comes in the form of governance programs whose goal is to create substantive expectations that are not directly enforceable by government [1], [2]. Soft law materializes out of necessity to enable a technological innovation to thrive and not be hampered by disparate heterogeneous practices that may negatively impact its trajectory, causing a premature “valley of death” exit scenario [3]. Soft laws are meant to be “just in time” to grant industry fundamental guidance when dealing with complex socio-technical assemblages that may have significant socio-legal implications upon diffusion into the market. Anticipatory governance is closely connected with soft law, in that intended and unintended consequences of a new technology may well be anticipated and proactively addressed [4].
Soft law’s role in governance is to influence the implementation of new technologies whose inception into society has outpaced hard law. Its usage is not meant to diminish the need for regulations, but rather be considered an interim solution when the roll-out of a new technology is happening rapidly, resisting the urge to create reactive and premature laws that may well take too long to enter legislation in a given state. Mutual agreement and conformance toward common goals and technical protocols through soft law among industry representatives, associated government agencies, auxiliary service providers, and other stakeholders, can lead to positive gains. Including the potential for societal acceptance of a new technology, especially where there are adequate provisions to safeguard the customer and the general public.
SECTION I.
Governing Artificial Intelligence
AI is the focus of this cooperative special issue in its entirety. This technology can be defined as the convergence of several advanced systems able to perform tasks that would have otherwise required human intervention, such as speech recognition, visual perception, decision making, and translation, among many other technologies. AI’s impact on society is forcing governments to act. For instance, the U.S. Office of Science and Technology Policy has put forward a need for an AI Bill of Rights, which is one of many legislative responses to a suite of technological and governance innovations currently causing governments to revisit hard law [5]. Yet, while the White House considers what to do with AI, thousands of proofs of concept, pilot programs, prototypes, and commercial AI ventures are entering the healthcare, finance, law enforcement, and transportation industries. Many of which are regulatory orphans. The question is, what happens in the interim to manufacturers and real end users? In essence, we use AI as the primary example in the context of governing with soft law because it is a fundamental matter affecting every facet of business, society, and government.
SECTION II.
Selected Papers: Overview
The papers in this issue were selected from a project funded by the Charles Koch Foundation and administered by Arizona State University. Through this initiative, academics and representatives from the private and nonprofit sector were invited to an initial workshop on January 9, 2020 where they submitted research for consideration. A second workshop on October 9, 2020, provided a forum for the presentation of papers, many of which were evaluated for potential publication in this pair of special issues. The aim was to explore ideas on how to improve the trustworthiness and effectiveness of soft law in an effort to maximize future benefits and minimize the drawbacks of AI methods and applications. All the papers underwent a thorough peer-review process with expert reviewers from the field of technology, innovation, and law. The resulting collection demonstrates the breadth and depth of nascent scholarship in the field. In fact, we believe that these are the first special issues released in the area of AI soft law with the added importance of being published within a technical communications journal. See also ASU’s Jurimetrics: https://lsi.asulaw.org/softlaw/about/.
The papers in this special issue provide an overview of soft law’s relationship to AI. The issue begins with Ryan Calo from the University of Washington who asserts that soft law without enforcement, in the form of stand-alone AI principles, is insufficient for shaping AI governance. Subsequently, Adam Thierer of the Mercatus Center at George Mason University surveys the recent historical use of soft law by the U.S. government to engage stakeholders. The next three articles by Gary Marchant of Arizona State University, Cary Coglianese and Lavi Ben Dor from the University of Pennsylvania, and Peter Cihon from the Centre for the Governance of AI at the University of Oxford and his collaborators, describe soft law program archetypes whose alignment of incentives motivates their enforcement. Finally, Carlos I. Gutierrez of Arizona State University characterizes the mechanisms utilized by 200 organizations in implementing their AI soft law programs.
SECTION III.
Soft Law for AI Governance
Due to AI being an emerging technology, its methods and applications are not only characterized by their technical evolution but also by our changing understanding of their impacts on individuals and organizations. Managing these outcomes through direct government action, also known as hard law, is a tried and true vector of action for many industrial and technology contexts outside of AI [6]. Unfortunately, the creation or amendment of laws is a time-intensive process. Hence, the effective management of AI’s continuously changing repercussions calls for a flexible alternative that complements or substitutes for direct government action. This is the role of soft law.
Unlike hard law that originates from a bureaucratic process, soft law need not have a predetermined form. Instead, it manifests itself in a kaleidoscope of known shapes, but also allows for future innovation in unknown ways. For instance, organizations may generate high-level norms or principles that signal their commitment to ideals. They may adhere to technical norms created by standards-setting organizations (e.g., IEEE and ISO). Professional societies can request that their members follow codes of conduct. Governments can publish guidelines that serve as a warning system, with the adoption of hard law as a backstop if insufficient adherence is experienced. Voluntary programs, labels, or certification schemes may attract the goodwill of a desired target population through the adoption of best practices.
Each of the above-mentioned soft law program archetypes is underpinned by characteristics that enable action with minimal procedures and barriers. In other words, they share a lack of jurisdictional constraints and have low transaction costs that allow any organization to create, amend, or adopt soft law. Furthermore, soft law implementation strengths have successful historical precedents for the governance of emerging technologies throughout the globe [7]–[10]. However, despite soft law’s benefits, all programs are not equally effective. Research shows that the majority of AI soft law programs do not publicly disclose whether or how they are implemented or enforced [11]. Changing the status quo requires thinking about the alignment of incentives to motivate action [12].
SECTION IV.
From Principles to Action
The six articles in this special issue exemplify the discussion of the future role of soft law in the governance of AI. We begin with a grounding assessment by Ryan Calo. In it, he offers a pragmatic perspective, a cautionary tale of sorts, about the path that this type of governance should not follow. He notes that organizational efforts to manage AI through developing and publishing high-minded principles are ineffective when those stated principles are unaccompanied by further action. Recent research found over 150 examples of AI principles being published, with only a quarter publicly disclosing enforcement plans [11]. What is obviously absent from socio-technical research is the operationalization of these principles into practice [13], [14]. Considering this, Calo rightly asserts that this type of program not only does little to affect change but it is also not a substitute for alternatives such as government action.
The editors of this issue agree with Calo in that the intervention of public authorities is often a necessary component of AI governance [15]. Nevertheless, action by legislative or executive bodies need not be limited to hard law. Research validates that, in addition to having a monopoly on hard law, governments are one of the main promoters and users of AI soft law [11]. Following this reasoning, Adam Thierer summarizes how the Trump and Obama administrations managed official AI policy by combining soft and hard law efforts. In addition, he compares the U.S. position to a centralized effort undertaken by the Chinese Government and to the Precautionary Principle approach employed by the European Union.
Following Thierer’s piece, we include three articles that identify concrete program archetypes for developing sustainable soft law that is likely to align incentives and shape how AI is managed. We begin with Marchant positing that existing iterations of codes of ethics promulgated by professional associations in the AI field have limited effectiveness and trustworthiness. To improve practitioner influence over the development and deployment of AI technology, he suggests strategies involving transparency, enlisting employers to enforce codes, and the development of requirements such as a license or fulfillment of educational and ethical obligations. This is followed by Coglianese and Dor who devote their paper to the impact of public procurement standards on constituents (i.e., the private sector). In citing precedents for environmental, small business, and cybersecurity contexts, they argue that harnessing the government’s purchasing power is a tool that could ensure industry alignment in the development of “ethical, transparent, and unbiased use of algorithms.” Meanwhile, Cihon et al. [A1] focused on certifications as tools that “incentivize the adoption … [of] … principles and substantiate that they have been implemented in practice.” In addition to analyzing seven AI certification schemes, the authors formulate recommendations for how future AI certification programs can remain relevant and effective.
Finally, once organizations decide to enforce AI soft law programs, they need to generate a structure charged with performing the activities required for this purpose. In his article, Carlos Ignacio Gutierrez assesses how 200 organizations operationalized these tasks. He categorizes these efforts into four quadrants, with levers and roles along one dimension, and internal and external organizational resources along another. Stakeholders in all sectors would be well served to consult these four options when designing their soft law implementation systems.
SECTION V.
Conclusion
Although the future of AI governance is undeniably uncertain, we have no doubt that soft law will play a decisive role in AI’s evolution. For the impact of soft law in AI governance to be trustworthy, effective, and sustainable, it is imperative that its creators heed Calo’s warning against the creation of empty instruments that serve no purpose or that do not consider the alignment of incentives. Instead, future programs should understand the AI governance context as described by Thierer, follow the strategies proposed by Marchant, Coglianese and Dor, and Cihon et al. in the development of program archetypes, and select the most appropriate implementation mechanisms as mentioned by Gutierrez. All of this is to ensure that beneficial impacts are maximized and that safe and useful AI is the more probable outcome for society.
Appendix: Related Articles
P. Cihon, M. J. Kleinaltenkamp, J. Schuett, and S. D. Baum, “AI certification: Advancing ethical practice by reducing information asymmetries,” IEEE Trans. Technol. Soc., vol. 2, no. 4, pp. 200–209, Dec. 2021, doi: 10.1109/TTS.2021.3077595.
References
1. G. E. Marchant and B. Allenby, “Soft law: New tools for governing emerging technologies,” Bull. Atom. Sci., vol. 73, no. 2, pp. 108–114, Mar. 2017, doi: 10.1080/00963402.2017.1288447.
2. G. Marchant, “Soft Law,” Governance of Artificial Intelligence, AI Pulse, Los Angeles, CA, USA, 2019. [Online]. Available: https://aipulse.org/soft-law-governance-of-artificial-intelligence/
3. S. A. Bini, “Artificial intelligence, machine learning, deep learning, and cognitive computing: What do these terms mean and how will they impact health care?,” J. Arthroplasty, vol. 33, no. 8, pp. 2358–2361, 2018.
4. D. H. Guston, “Understanding ‘anticipatory governance,”’ Social Studies Sci., vol. 44, no. 2, pp. 218–242, 2014.
5. E. Lander and A. Nelson, “Americans Need a Bill of Rights for an AI-Powered World,” WIRED. [Online]. Available: https://www.wired.com/story/opinion-bill-of-rights-artificial-intelligence/
6. K. W. Abbott and D. Snidal, “Hard and soft law in international governance,” Int. Org., vol. 54, no. 3, pp. 421–456, 2000.
7. D. M. Bowman, “The role of soft law in governing nanotechnologies,” Jurimetrics, vol. 61, no. 1, pp. 53–78, 2021.
8. C. Coglianese, “Environmental ‘soft law’ as a governance strategy,” Jurimetrics, vol. 61, no. 1, pp. 19–51, 2021.
9. A. Thierer, “Soft law in U.S. ICT sectors: Four case studies,” Jurimetrics, vol. 61, no. 1, pp. 79–119, 2021.
10. Y. A. Stevens, “Soft law governance: A historical perspective from life-science technologies,” Jurimetrics, vol. 61, no. 1, pp. 121–131, 2021.
11. C. I. Gutierrez and G. E. Marchant, “A global perspective of soft law programs for the governance of artificial intelligence,” SSRN Electron. J., to be published, doi: 10.2139/SSRN.3855171.
12. C. I. Gutierrez, “Identifying Incentives for the Enforcement of Artificial Intelligence Soft Law Programs,” SSRN Electron. J., to be published, doi: 10.2139/SSRN.3897486.
13. A. F. Winfield, K. Michael, J. Pitt, and V. Evers, “Machine ethics: The design and governance of ethical AI and autonomous systems,” Proc. IEEE, vol. 107, no. 3, pp. 509–517, Mar. 2019.
14. R. Abbas, J. Pitt, and K. Michael, “Socio-technical design for public interest technology,” IEEE Trans. Technol. Soc., vol. 2, no. 2, pp. 55–61, Jun. 2021, doi: 10.1109/TTS.2021.3086260.
15. K. Michael, “Can good standards propel unethical technologies?,” IEEE Technol. Soc. Mag., vol. 35, no. 3, pp. 6–9, Sep. 2016.
Authors
Sandra Day O’Connor College of Law, Arizona State University, Phoenix, AZ, USA
Carlos Ignacio Gutierrez received the B.B.A. degree from the University of Notre Dame, Notre Dame, IN, USA, in 2005, the M.A. degree from University College London, London, U.K., in 2013, the M.P.P. degree from the Pardee RAND Graduate School, Santa Monica, CA, USA, and the Ph.D. degree from the Pardee RAND Graduate School in 2020.
Sandra Day O’Connor College of Law, Arizona State University, Phoenix, AZ, USA
Gary E. Marchant (Member, IEEE) as born in Squamish, BC, Canada, in 1958. He received the B.Sc. and Ph.D. degrees from the University of British Columbia, Vancouver, BC, Canada, in 1980 and 1986, respectively, the J.D. degree from Harvard Law School, Cambridge, MA, USA, in 1990, and the M.P.P. degree from the Kennedy School of Government, Harvard University, Cambridge, in 1990.
School for the Future of Innovation in Society, School of Computing and Augmented Intelligence Arizona State University, Tempe, AZ, USA
Katina Michael (Senior Member, IEEE) received the B.Sc. degree from the University of Technology Sydney, Ultimo, NSW, Australia, and the Ph.D. and MTransCrimPrev degrees from the University of Wollongong, Wollongong, NSW, Australia.
Citation: C. I. Gutierrez, G. E. Marchant and K. Michael, "Effective and Trustworthy Implementation of AI Soft Law Governance," in IEEE Transactions on Technology and Society, vol. 2, no. 4, pp. 168-170, Dec. 2021, doi: 10.1109/TTS.2021.3121959.