Professionally Responsible Artificial
Intelligence
Michael Hatfield*
ABSTRACT
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 (IRS), 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
* Dean Emeritus Roland L. Hjorth Professor of Law, University of Washington School of
Law. For their extraordinary research assistance, I would like to thank Mary Whisner, Maya
Swanes, and Crystal Alberthal of the University of Washington Gallagher Law Library and my
student research assistants Allexia Arnold and Mark Noel. I also would like to thank Shannon
McCormack and the participants in the 28th Annual United Kingdom Tax Research Conference
at the University of Central Lancashire, Preston, England.
