Innovation and Governance and the FHFA

From research into AI to collaborative events encouraging the development of automation, Federal Housing Finance Agency Associate Director Anne Marie Pippin has been involved in several projects that given her broad insight into what the next wave of technology development in the industry may look like. In this discussion at NMN's Digital Mortgage Conference, she talks about the dual role she has with two offices within the agency's division of Conservatorship Oversight and Readiness, and the view that's given her of what lies ahead for the mortgage industry.


Transcription:
Transcripts are generated using a combination of speech recognition software and human transcribers, and may contain errors. Please check the corresponding audio for the authoritative record.

Bonnie Sinnock (00:09):
Welcome to our Arizent Leaders interview with Anne Marie Pippin, associate director at the Federal Housing Finance Agency, at Digital Mortgage 2023 in Las Vegas. Today we're going to do a short five-question style interview with Anne Marie about her work at the FHFA and the view it gives her of fintech trends in the market.



Anne Marie, I wanted to start by talking about your role in the FHFA's Office of Financial Technology. Can you tell us a little about that and the two hats you wear there, if that's a fair way to phrase it, in terms of innovation in general governance?



Anne Marie Pippin (00:44):
Sure. Sounds great. Thanks Bonnie. And then, thanks so much for inviting me to do this session. I remember being at Digital Mortgage about the same time last year when the FHFA Office of Financial Technology had just first been established. So we have been about a year now in terms of understanding where technology is driving innovation in the mortgage ecosystem, who the players are, what the challenges and the barriers are to innovation into greater adoption, and to really thinking about this from a responsible innovation standpoint. And to us at FHFA, that really means understanding innovation from the standpoint of doing innovation within guardrails, having good governance and controls in place to address the various risks that various innovations present. Now, as far as my role at FHFA, I'm an associate director in our fintech office. I also lead our Office of Governance and Strategic Initiatives, which falls underneath our division of Conservatorship Oversight and Readiness. So I wear those two hats. As far as governance and strategic initiatives, I really look very closely at Fannie and Freddie in terms of all of their corporate governance practices. And then within the fintech office, we work with our regulated entities to understand both Fannie and Freddie as well as the Federal Home Loan Banks where innovation is really driving within the mortgage ecosystem



Bonnie Sinnock (02:11):
Well, speaking of innovation, we talked earlier about digital data validations on another panel here, and I was thinking after your comments there about how transformative those partnerships are with aggregators of that information. Could you talk about what future discussions you might have with stakeholders about that?



Anne Marie Pippin (02:30):
So one of the things that we're really focused on in the fintech office at FHFA is engaging with stakeholders, engaging with the ecosystem to understand what are the challenges, what are the barriers, and what are the opportunities that are out there? So one of the things that we first did when the office was established is we issued an RFI. We engaged through a public listening session, and then just a few months ago, we held our first technology sprint or tech sprint, as we like to call it. This is an ideation focused sprint to understand where data digitization can drive for greater transparency within the mortgage process, can increase access and adoption, and where it can increase affordability, equity access as well as sustainability. And we really focused in on a couple of core areas for this tech sprint. We focused in on those direct source verification processes to understand the thinking about the existing tools that the enterprises have around AIM and Day 1 Certainty to really understand where those processes might lend themselves for greater adoption.



(03:37)
We also focused in on the need for data standards and where we should be focusing, thinking about that. We also focused in on the need for data trust. So trust from the perspective of the lending community as well as from the perspective of the borrower when thinking about a more digitized mortgage experience. So we had about 80 participants come to Washington, D.C., in July and participate. And that included stakeholders from across the fintech ecosystem, so data aggregators, technology service providers. We had lenders of all sizes also participate in the tech sprint. We had a few academics, we had trade groups and we had a few consultants. We brought everybody together and we put everyone onto teams that had a diverse makeup. And what does that mean? We wanted to be able to understand and drive towards creative solution making, think about solutions that were outside of the box, things that haven't already been developed.



(04:35)
And so we put fintechs and lenders and academics all blended onto different teams together so that they'd have an opportunity to really collaborate with their peers and across the mortgage ecosystem to drive towards some solution making. And so we had, I think a very successful first tech sprint. We were looking for ideas. It was an ideation-based tech sprint, and we're actually hoping to be able to engage with the teams again this fall. And to go a little bit deeper into the solutions that were presented, the teams only had about five minutes or so to present their solutions on demo day. The tech sprint in total was four days. So it was about three days of driving towards a particular solution. And then we had our demonstration day where the teams had five minutes to present to a panel of judges, and then about three minutes of Q&A. So we really, really think it's worth going in a little bit deeper to really understand where the themes are in terms of what the team has presented, and then think about next steps.



Bonnie Sinnock (05:36):
One of the themes I saw when I was looking over some of the takeaways from that tech sprint was on venture capital and on startups and some of the challenges there. I wondered if you could tell me a little bit about some of the takeaways from that.



Anne Marie Pippin (05:51):
Yeah. Well, I think some of the solutions really focus in on the need for greater verification and trust in mortgage processes and driving towards that through innovation. A couple of other things, really focusing on the need to understand emerging technology such as artificial intelligence, distributed ledger technology. And these are areas where startups too certainly play a very important role. Another key theme was around financial education and really understanding what could drive towards innovative processes to engage with prospective borrowers, with renters, to really educate how the mortgage process goes to build transparency into that. And then of course, consumer empowerment was also another key theme.



Bonnie Sinnock (06:39):
I was looking at the advisory bulletin on artificial intelligence that the FHFA has, and it looked like there were some core principles and ethics. There were kind of guidelines for some of the guardrails, if you will, for artificial intelligence and lending. I wondered how you came up with those, and if you could summarize those briefly here.



Anne Marie Pippin (06:57):
Yeah. So FHFA issued its advisory bulletin and advisory bulletins are supervision guidance to our regulated entities in terms of how to think about risk management best practices, and operating in a safe and sound manner in a particular area of risk. And FHFA, I had the opportunity to co-write an advisory bulletin on artificial intelligence. And machine learning was issued in February of 2022, so almost two years ago now, it's hard to believe. And it was about two years in the making actually putting that advisory bulletin together. We really drew from a lot of different sources, such as the work that NIST, the National Institute of Standards and Technology, was doing in the artificial intelligence space at the time around understanding the need for trustworthy AI and what that looks like, addressing bias and what that looks like, what considerations we should be thinking about. We also did a lot of outreach with our peer regulators, the OCC at the FDIC, the Federal Reserve, the CFPB, to understand, well, what are your thoughts around artificial intelligence and machine learning, the need to manage that risk within their supervision guidance-development processes?



(08:04)
So we took all of that together. We monitored where the enterprises Fannie and Freddie were at the time in terms of their exploration of artificial intelligence and machine learning. And that fed into the development of the AB. One of the things that we thought was really critical to include in the AB was this understanding and thinking about the use of AI from an ethical lens and the need to build in ethical principles into the governance structures that the enterprises were setting up to manage. The risks associated with AI/ML risks are not just model risks. They really cut across the enterprise. AI/ML can heighten operational risks, it can heighten model risk, can heighten compliance risks for certain types of usages. And so we have different aspects of that sort of built into the AB around considerations of thinking about those risks, thinking about the control structures that need to be in place and thinking about how that cuts across.



(09:01)
But the ethical lens and the ethical principles are really important. So there are a couple of core principles that I'll just highlight, and that's the need for transparency. And that's this concept of a black box risk that AI/ML models present. Some methods are more transparent than others, I will say. And when more complex algorithms are used, there definitely is a need to understand, and it can be able to explain and interpret how the model goes from the data inputs that come in, how it translates those into a prediction, into a classification, and then how the model output should be interpreted. So that's one of our core ethical principles. Another core ethical principle is addressing bias, right? And this is addressing bias from the perspective of thinking about the bias that's inherent within the data that the model's ingesting, as well as the bias that might be presented from the type of algorithm that's being used to solve for that particular business challenge.



Bonnie Sinnock (10:02):
So my final question for you was, and this came up on one of the panels today, or actually yesterday, about the role of regulators generally. And I don't know if you can speak on this in terms of regulation in general or your own role, but the question is, with innovation, should regulators be taking more a reactive role or a proactive role in innovation? Do you have any thoughts on that?



Anne Marie Pippin (10:26):
Yeah, I mean, I think it's interesting to see the innovation offices that have been established over the past several years across the federal regulatory landscape. It's really critical to go in eyes wide open with an understanding of how innovation is really driving changes in mortgage processes, and to really understand where the risks may be presented with those types of solution, the different innovations that are driving solutions, understand the risks, and then of course that feeds into the need for potential regulation, guidance, etc. development. But I think it's very important to kind of go eyes wide open into that.



Bonnie Sinnock (11:13):
Alright, thank you. Those are all the questions I have for you today. Thanks for coming to this panel and thanks to everybody who helped make this Leaders session possible today.



Anne Marie Pippin (11:22):
Thanks, Bonnie.