A scene from the 1970’s novel and 1980’s film The Princess Bride was the inspiration behind this Center for Security and Emerging Technology (CSET) report, Inigo Montoya Problem for Trustworthy AI: The Use of Keywords in Policy and Research. According to the report, the creation and assessment of trustworthy AI relies on detailed framework from academic and research communities. The progress in the development of trustworthy AI is complicated due to several reasons outlined in the report. The artificial intelligence and machine learning field is growing rapidly, accompanied by new tools and techniques. The very term “trustworthy AI” is complex and “multifaceted.” A third complication arises when policymakers and technical researchers unknowingly assign different meanings to the same terms in describing trustworthy AI, creating what the report describes as the “Inigo Montoya” problem.

Inigo Montoya is a character from the aforementioned literary work, and this paper references a scene where Inigo misunderstands the word “inconceivable”–a word frequently used by the antagonist, Vizzini. CSET offers this scene as a lesson for the research and policy communities. These communities use keywords drawn from the National Institute of Standards and Technologies’ (NIST’s) AI Risk Management Framework. NIST defines characteristics of trustworthy AI, and these characteristic lists serve as a foundation for “policy-relevant trustworthy AI terms.” In place of the term “inconceivable,” this study analyzed the frequency of use of trustworthy AI terms, including reliability, robustness, safety, security, bias, interpretability, privacy, trust, explainability, fairness, transparency, resiliency, and accountability. This paper aims to assist technology policymakers in understanding how research and development communities use trustworthy AI keywords and how overlaps in term usage can cause problems.

For more information, check out HSDL’s In Focus topic on Artificial Intelligence.