How neuro-symbolic AI is becoming the new guardrails to reduce risk for brokers and borrowers
The rapid growth and usage of artificial intelligence in everyday life is both a technological marvel and a growing concern.
While all the new bells and whistles have made both personal and business life more efficient, there are concerns about both the accuracy of AI output and the potential impact if that output is incorrect.
Some of the effects are very low-stakes, and the penalty for an AI getting the answer wrong is negligible. However, in the world of finance, and specifically mortgage lending, the penalty for getting the wrong outcome can be severe. It can be a major financial penalty, and it can even cost a broker their job.
One executive whose company is working to help innovate with AI in the mortgage space understands those concerns.
Dave Parker (pictured top), CEO of LoanLogics, said as AI evolves, new guardrails will be both needed and developed to make sure brokers, lenders, and customers are protected.
“If you’re going to be in business outcome automation, then you better be right,” Parker told Mortgage Professional America. “We look at the world as a four-quadrant type of graph, where on the bottom you have these costs of errors. On the left-hand dimension, we have the type of knowledge that’s required, either very explicit or a tacit type of knowledge. And so if it’s low-cost and explicit knowledge, then there’s one type of AI solution.
“However, if it’s a high cost and tacit knowledge, then you'd better be able to justify how you came to those conclusions. You can’t have any of this drifting that you get with LLMs and some of these different models.”
The growth of neuro-symbolic AI
The reason this issue is so critical is because of the high regulatory standards in the mortgage industry. Not to mention the impact on human lives if AI gets it wrong.
Peter Idziak, a senior associate and mortgage attorney at Polunsky Beitel Green, told Mortgage Professional America in November that whatever new tech is used must comply with all laws, including the Equal Credit Opportunity Act (ECOA) and the Fair Housing Act (FHA).
“Whenever you're engaging vendors, what's their training data?” he said. “From a fair lending perspective, we have ECOA, so you can't discriminate based on sex, race, or national origin. But if you have a vendor from outside the space that's come in saying, ‘Hey, I'm going to help you underwrite your loans.’ One, what data do they train on? Two, how does it do its thinking, and how is it producing the result?
“Because, for ECOA adverse action notices, you need to have a reason. You can't just say ‘The AI said so.’ Well, why did it say so? ‘I don't know. It's a black box.’ So that's what people consider the more generative side.”
This is part of the reason that companies are turning to neuro-symbolic AI. Instead of just a black‑box model, neuro‑symbolic AI can say, “Here’s what I see in the data and here’s the rule or guideline that drives my decision.” It learns patterns from data, and it also follows explicit rules and regulations. It’s like an underwriter and a compliance officer sharing one brain.
Parker said this technology can avoid the drifting common in some AI models and be forced to follow strict guidelines.
“Neuro-symbolic stays within the confines of the logic that I’m feeding it,” he said. “And I’m telling it not to go get creative with more logic. Neuro-symbolic is nice because it gives me all the audit trails. It stays in the confines of the parameters that are the guidelines and the rules that we manage the work by today. But it interprets it faster, and it’s more comprehensive than what some of the traditional technologies and systems are.”
Challenging existing systems
One additional use for this type of AI is finding patterns in borrower behavior, similar to what credit card companies use when you make a big purchase. They will analyze the borrower, and if anything seems out of place, they will take note of it.
“When you look at the sophistication that credit card companies have now, where they can analyze your behaviors, where you use your card, when you use your card, and they know when you leave that happy path, and they’ll reach out and communicate with you,” Parker said. “They’ll ask, ‘Did you just buy a big TV from this place?’ I do believe the same will get applied to mortgage. And I think the sophistication goes up.”
While the technology will need to be good to convince those who might hesitate to use it, Parker believes it could change many of the longstanding systems in place in the mortgage industry. He wonders if some of the existing frameworks will be able to keep pace with all of the new data received from neuro-symbolic AI.
“I think it’s really going to challenge a lot of the traditional systems in the mortgage space to be able to keep up with the AI capabilities,” he said. “Where AI enables me to do a lot more than maybe my system of record supports, and how do I handle that?”
Stay updated with the freshest mortgage news. Get exclusive interviews, breaking news, and industry events in your inbox, and always be the first to know by subscribing to our FREE daily newsletter.
