Tuesday, September 10, 2019
Michael Hatfield (University of Washington) presents Professionally Responsible Artificial Intelligence at the 28th Annual Tax Research Network Conference today at the University of Central Lancashire, England:
As artificial intelligence (AI) developers produce more applications for professional use, how will we determine when the use is professionally responsible? One way to answer the question is to determine whether the AI augments the professional’s intelligence or whether it is used as a substitute for it. To augment the professional’s intelligence would be to make it greater, that is, to increase and improve the professional’s expertise. But a professional who substitutes artificial intelligence for his or her own puts both the professional role and the client at risk. The problem is developing guidance that encourages professionals to use AI when it can reliably improve expertise but discourages substitution that undermines expertise.
This Article proposes a solution, using tax professionals as a case study. There are several reasons tax professionals provide a good case study, including that tax practice has a long history of computerization and that AI is already being developed for tax professionals. Tax professionals, including not only lawyers but certified public accountants are directly regulated by the Internal Revenue Service, in addition to their regulation by professional bodies.
This Article proposes a public-private cooperation in regulating the use of AI by professionals in ex ante tax planning. On the private side would be panels of experts testing new AI applications for reliability by running experiments. The panels would certify AI products determined to be substantively sound and designed to educate and engage the professional. On the public side, the IRS would provide a presumptive defense to professional responsibility-related penalties against professionals who used the certified AI. This should motivate tax planners to prefer purchasing certified tax planning AI applications, and thereby motivate tax AI application developers to seek certification.
Though this Article’s proposal is specific for the use of AI by tax professionals, it illuminates a way forward for regulating AI use by other professions. The way would be for third parties such as government agencies, professional associations, or malpractice insurers to stimulate demand for certified AI products to be used by professionals. In general, these certifications should be provided to AI that augments the professional’s intelligence, increasing his or her professional competence. By keeping professionals involved in the certification process, space is opened to shape the transformation AI is bringing to the professions, and by stimulating product demand for certified products, the odds of successfully shaping that transformation are improved.