AI Risk Management in Practice: Why Algorithmic Audits?
Sri Krishnamurthy, CFA, CAP
CEO, QuantUniversity | Entrepreneur | AI Innovation | Education | Author | Speaker
I spend a week in the White Mountains in New Hampshire and one of the must dos here is climb Mt.Washington! 6000+ ft, and people climb, ride, take a train up or drive up for the bragging rights of being on top of this famous mountain. With a time crunch this year, I suggested driving up Mt.Washington to my 10 year-old daughter and she asked “Why”?. I said, well, we can get a sticker at the top “This car climbed Mt.Washington!”. That didn’t impress her.. She asked, can we just get a sticker???Sigh!
In the last edition, I talked about how Algorithmic audits are hard to define. In this edition, I want to start with the question
“Why Algorithmic audits?”
We know many regulators are calling for increased scrutiny for Algorithmic systems. Some of the goals here is to ensure compliance, protect vulnerable consumers, ensure equitable access to systems & tools, ensure systemic risks aren’t creeping up etc.
But will Algorithmic audits solve the problem? If so, what is the problem? How will Algorithmic audits address these issues? The discussion on Algorithmic audits need to be formally scoped and contextualized! Who is it for? What is the purpose? What is the scope? What are the claims? Who is asserting those claims? Who is verifying the claims are valid or not? What tools and methods are used to validate those claims? How independent is the auditing process? What standards define the tenets of an audit???
These are not easy questions. As an academic and someone who has studied and practiced computer science for two decades, formal verification is hard? Just using heuristics and looking for a needle in the haystack won’t address the potential adverse effects of algorithmic systems.?
Also, I have seen many reports being published (along with press releases) that companies have submitted themselves to algorithmic audits and have gotten a green sticker!??To tie back to my “This car climbed Mt.Washington!” sticker, this “Inspection sticker” of Algorithmic audits is more of a bragging rights at this point rather than a guarantee or formal claim of certification!?
More discussion is needed, and I am glad the community is engaged in this topic. But we need the right stake holders including practitioners at the discussion table. Let’s not rush to collect Algorithmic audit stickers! It will end up being decals on a rear-view window. Bragging rights and nothing more!
We will discuss more on this topic in the next few editions of the Algorithmic Audit newsletter.
??Algorithmic Audits in the News:
Stanford Launches AI Audit Challenge. This challenge is generously funded by the Rockefeller Foundation and will focus on tools to assess whether deployed AI systems illegally discriminate against protected categories. More details at:?https://hai.stanford.edu/policy/ai-audit-challenge
领英推荐
??Event of the week:?
Maya Murad from IBM will be speaking on the topic:?“A framework for involving people in enabling the responsible use of algorithmic decision-making systems”?at the Quantuniversity Machine Learning Summer School 2022. I will be moderating the session. Join in at 12.00pm ET if you want to be part of the discussion. The event is free !?
??Keep on learning!
?? Want to learn more formally? Join the AI Risk Management Certificate program developed in partnership with PRMIA ->?https://lnkd.in/eVEhyNSQ
??Many of these topics will be elaborated in the AI Risk Management Book published by Wiley. Check updates here ->?https://lnkd.in/gAcUPf_m
??Subscribe to this newsletter/share it with your network -> https://www.dhirubhai.net/newsletters/ai-risk-management-newsletter-6951868127286636544/
I am constantly learning too :) Please share your feedback and reach out if you have any interesting product news, updates or requests so we can add it to our pipeline.
Sri Krishnamurthy?