AI-first company

Claim check

“Microsoft has a maturity model for responsible AI”

The model exists as described: 24 dimensions, five levels, built from interviews with more than 90 specialists. One limit: it covers responsible AI only, not AI readiness in general.

By
Last reviewed
Corrections
Report an error
Verdict Confirmed

The primary source says what the claim says.

The claim

Microsoft’s Responsible AI Maturity Model can be used to assess an organization’s AI maturity.

What the primary source says

Microsoft Research published the Responsible AI Maturity Model in May 2023. The authors are Mihaela Vorvoreanu, Amy Heger, Samir Passi, Shipi Dhanorkar, Zoe Kahn and Ruotong Wang.

  • It “contains 24 empirically derived dimensions”.
  • Each dimension has five levels: Latent, Emerging, Developing, Realizing and Leading.
  • The dimensions fall into three groups: Organizational Foundations, Team Approach, and RAI Practice. RAI stands for responsible AI.
  • It is “based on interviews and focus groups with over 90 RAI specialists” and AI practitioners.
  • Microsoft describes it as “a framework to help organizations plan how to achieve RAI maturity”, to be used in planning workshops.

What to keep in mind

The claim holds if “AI maturity” means maturity in responsible AI: fairness, transparency, accountability and similar practices. Three limits apply.

  • It is not a readiness check. The model does not cover data, infrastructure, skills in general or business value.
  • It was built from interviews. It describes what specialists said good practice looks like. Nobody has tested whether a higher level goes with fewer problems or better results. That gap is common to maturity models: see ten years of digital maturity models.
  • It has no open licence. The document is free to read. Translating or adapting it needs Microsoft’s permission.

What to cite instead

Microsoft Research’s Responsible AI Maturity Model (Vorvoreanu and colleagues, 2023) describes 24 dimensions of responsible AI practice, each at five levels from Latent to Leading. It is a planning framework for responsible AI, not a measure of general AI readiness.

Sources

  1. Vorvoreanu, Heger, Passi, Dhanorkar, Kahn and Wang (2023). Responsible AI Maturity Model. Microsoft Research, report MSR-TR-2023-26. microsoft.com
  2. Thordsen and Bick (2023). A decade of digital maturity models: much ado about nothing?. Information Systems and e-Business Management, 21(4), 947–976. What this study did and found