Podcast Transcript

You're listening to the RSA conference podcast where the world talks security.

Hello, listeners. Welcome to this edition of our RSAC podcast series. Thank you for tuning in. I'm Tatyana Sanchez.

And I'm Kacy Zurkus.

And we are your RSAC podcast hosts. Kacy, what are we going to discuss today?

Well, we often discuss AI in the cybersecurity world as either a defense tool for organizations and users or an attack technique used by cybercriminals. However, we rarely take the time to examine how AI is affecting humans, especially our emotional intelligence. That's why we're excited to be joined today by Nancy Yoon, who will explore the gap between AI and emotional intelligence.

We'll talk about what individuals can do to bridge that gap and how to effectively use AI to assist us in our professional and personal lives while keeping our brains exercised. Ready to dive right in, Tatyana?

Yes. I'm really excited about this topic. But before we get started, we do wanna remind our listeners that here at RSAC, we host podcast twice a month, and we encourage you to subscribe, rate, and review us on your preferred podcast app so that you can be notified when new tracks are posted.

And now we would like to ask our guest to take a quick moment to introduce herself before we dive in. Nancy?

Sure. Hi, everyone. This is Nancy Yuen, and I am the head of SOX Financial Data and Reporting Regulatory Governance at SoFi Technologies. And thanks so much for having me. I am passionate about both topics, emotional intelligence and artificial intelligence and how they intersect.

And just passionate about just getting learning to everybody while learning myself, so I'm looking forward to this.

Thank you, and thank you for being here with us today. Nancy, in a world where AI can flag every anomaly, score every risk, and automate every control, why do our biggest failures still come down to the human behavior?

In my research and even personal use of AI, the biggest failures still come by taking that assumption, whatever we do with artificial intelligence is just gonna work itself out without a human involved. And that's what I call the bias and overreliance of automation.

We humans want to find the quickest pathway, just like an electrical circuit. You want to find the path of least resistance. And what we tend to do with AI is trust machines. And we trust machines to perform the work that we typically would do manually, but we also would accept their outputs blindly, and especially given that path of least resistance, we are always facing time pressures, and therefore it leads us to over rely on the critical judgment that we assume the AI will have, and turn that AI into a place of reliance.

There's also this concept of the black box problem. So many AI models that we all use daily, oftentimes in our work and in our personal life, they operate without any transparency. It's really hard to understand how they made a decision, how they produced an input or output. And this opacity really makes it impossible for you and I to double check their logic, to double check their challenges, and also to double check how they have integrated ethics within their decision making.

The other thing that we also want to talk about is contextual understanding. Does the AI understand nuance? Does it understand culture? Does it have a sense of bias? And what is the ultimate why driving the data that it's producing? And that's where humans like us, come in, like the three of us come in, so that we can go ahead and understand, interpret, and really direct the AI instead of having it direct us.

I think that's so interesting, and I love what you said about trust. Right? Because as we talked about when we were just sort of starting our conversation backstage, we're always onto the next thing. Right? We're always like, we're all busy all the time. And so it is so easy when a tool comes along that provides us with this opportunity to sort of trust what it's doing and say that, you know, it's saving us time.

But automation is really good at catching, like to your point, what's documented, how it's trained. But it does struggle with what's felt, what it means to be human. And if we over automate, are we disabling the sort of human relational controls that allow us to see the risks that AI simply isn't programmed to find?

That's a great question, Casey. And you are absolutely right. When we are stretched thin on resources, either at work or at home, we often turn it to AI. It excels.

Let me be clear, it can excel at handling your routine, high volume, data heavy tasks, but that human judgment is still really critical for managing exceptions, but also ethical dilemmas. That's what we're built for. We're built to live in community and to have human experiences so that we can determine what's right and what's wrong. That is something that AI cannot do, and that is where a human needs to feed it and it learns, But ultimately, all of our human experiences and the way that the neurons travel within our brain, it cannot do.

It can never do. It cannot have feelings, it cannot have faith, it cannot do all the things that we're built to do. So what I would say is over automation can disable the relational controls that we humans need to have. And this would include your human intuition, that gut feeling, the ability to take in context, what I say is read the room, and also that empathy of human, emotional intelligence, that empathy that's required to identify those risks that the AI is not programmed to recognize.

And while we deem automation as a keyword, and it does excel at processing all of that high volume data, what it fails to do is it fails to understand the nuance, the context and the human intent, or even ethics of certain data inputs and the outputs it generates. So this human reliance on automation, what we call automation bias, can really cause mutual atrophy to our human brain, to our human just instinct and response, and that's what we can't lose. We can't lose the concept of emotional intelligence, of having self awareness, because we're not always going to have AI in our back pocket, it's not going to drive those situations where you're at a coffee shop, and you're interacting with a barista, and you're perhaps at a hotel, and you're interacting with the door person, it's not gonna be able to be there at every single moment.

And it still requires us to have that emotional intelligence piece that we derive from the experiences, and also the challenges, the interruptions of that flow of path of least resistance, those challenges are the largest way that our body learns to heal itself and prepare itself for future challenges.

When we rely on AI so much that it interrupts that practice, that challenge, that muscle usage, we start to over rely and unfortunately our neurons get adapted to just a quick hit of dopamine, of that quick satisfaction of, oh I got the answer. But when we're placed in a real life situation without AI, that's when we need it most, all that human emotional intelligence.

And Nancy, you mentioned at the end of the day, AI can do a lot, but it's not programmed to do certain things. It doesn't have feelings. So it ultimately comes down to us humans determining what's right or wrong.

But AI can't find a risk on employees too scared to report. And as we know, when fear of blame or appearing incompetent leads a team to hide their mistakes, it can create a massive visibility risk.

So how can organizations build true psychological security, ensuring teams know that their honesty is their most important defense and that speaking up is always safer than staying silent?

That's exactly it.

We don't want to get to a place where we ourselves are becoming machines. And I know there's a lot of great movies out there, some recently released, some that have been, you know, I would say, classic novels at this point, But that's exactly it. We don't want to be machines. We were not built to be machines.

We are built human and we have mistakes and that's okay. And that's exactly, I love that question, Tatiana, because having a human nature means making mistakes, means being imperfect. And what we're talking about here is in this workplace culture where it is, you know, using automation AI, you're almost expected to perform in that perfect manner, and yet we will never be, I'll be clear about that, we will never be perfect, whether it's in workplace or personal life, but especially when your workplace and your management leadership, even peers, require you to be perfect, you're going to be scared.

And that's where psychological safety and what we call psychological security is requiring this transformation now within the workplace.

We go from a blame based system, right? We all know that, especially if we had brothers and sisters or if we went to, you know, primary school, and we automatically understand that blame based system in that system of tattling and pointing, right, because it feels good to correct mistakes. It feels especially good to correct mistakes of others.

But we need to move away, especially as we're using AI and expecting perfection, to a learning oriented one. And that's where we're able to speak up and be seen and heard and recognized and see that as a sign of being proactive, as a sign of adding value and contribution. And this admission of weakness becomes a superpower. If we don't know where we are weak, there is no chance that we are going to improve.

Just like, I'll compare this to the AI, if we don't tell it that it is wrong, it will keep learning off of the data that it feeds. Similar to in the workplace, if us in leadership keep accepting the wrongdoings of a blame based system, a doc on every time someone perhaps makes a mistake, and it's really up to us to change that culture and to go into what we call the psychologically safe environment where you can admit mistakes publicly and not have ramifications. You can normalize not knowing. Often my team hears, Oh no, I don't know that.

Tell me more. Or I've made a mistake. When leaders start to do that, and when leaders take a pause and really self assess what they know and don't know, but even more importantly, when they verbalize it for their team, you're normalizing that it's okay to make mistakes, and please make mistakes. The faster we break things, the faster we make mistakes, is the faster we find solutions.

Please normalize not knowing. The curiosity of a child is how it learns. We need to have that childlike curiosity throughout life. And also when we're speaking about curiosity, we need to as leaders respond to curiosity without judgment.

The typical phrase that you may hear in class is, oh, maybe everyone knows this already, but I'm going to ask, I wish more people did that. I wish more people could be brave and stand up. Because more often than not, you will hear a teacher, a great teacher, Ask your question, I'm sure others in the room have it. And if we're brave about doing that, we're creating this culture that allows the psychological safety, allows questions, and allows this kind of counter of anything that we're given.

And that's especially important in AI, to counter every single time we hear, you know, a response if it's given to us verbally or we challenge it and we challenge the prompts itself as well.

I think it's so important to think about everything that you're saying in the context of the development of these tools too. Right? And, you know, sometimes as we've seen in headlines recently, features are created that are intended for one purpose, but then when someone figures out how to misuse them, it creates a really big problem. Right?

And so maybe taking the time to sort of explore potential misuse and how could this go wrong once it goes to market before it actually is implemented as a feature and a tool that can be exploited that allows for this expansion of, like, this application of emotional intelligence. Right? And that's, I think, important for risk leaders to be thinking about because certainly you don't want to deal with the risk in the aftermath. Right?

So Exactly.

As AI becomes more embedded in our control environment, are we investing as intentionally in the one thing that it can't replace in emotional intelligence as risk leaders. So, you know, what are some of the primary gaps between artificial and emotional intelligence and how do we bridge those?

It's a great area to explore, Casey. And what we can do is just break down what is Emotional Intelligence. So Emotional Intelligence, or EI is what we call it for short. It has the components of self awareness.

So being aware of yourself, not just your surroundings, but also how you are feeling. How did you arrive to that room? How did you arrive to the office? How did you arrive to the grocery store?

How did you arrive back home after a really long day? Being really self aware is the key and the cornerstone to improve emotional intelligence. The second component of this EI concept is self regulation.

So you recognize yourself, you recognize that, Hey, I'm coming home after a really long day, I'm exhausted. I myself am an introvert, and so after a long day of say networking meetings, I am tired and it is hard to come back to three children that need you. And therefore I need to regulate my emotions and focus on what that priority is, which brings me to the next component of AI, which is motivation.

That motivation to change, to adapt to the new surroundings that you're in, to understand where you're at, that self awareness, how to regulate your own emotions, and the motivation to address what's in front of you. And that comes with the next component, which is empathy. This empathy of how are the others going to interpret my actions, my feelings, even the signals that my face give, right? It's the empathy of placing ourselves in someone else's shoes to see and really determine how they are going to react to us.

The last component is social skills. And this is something that I've talked about before with the RSAC is neurodiversity. I myself have a high functioning autism, social skills were not my strong suit, and this is an area I need to build up. Well, all of these areas are where I need to build up.

But social skills, we need to develop that social skills. And that's what I was talking about before with the AI. We need to get out into the world. As an introvert, that is really hard.

But we do, because those human interactions, those are building up our social skills and social muscles. And also our muscles within our brain that exercises empathy, all of the above, self awareness, self regulation, motivation, empathy, but that social skill is where it's put to the test. That is your final exam. Your final exam at the end of the day is really how do you respond and how do you react in social situations, because as humans, we're meant to live in community, that is why we have language, that is why we have the ability to embrace, to smile and have these emotions.

And when we take that away and we bridge this back to the development of AI, we need to start understanding, and this is where the, feedback loop concept comes in for AI, when we are instructing the machine to perform certain actions, it's really important for just like a parent and child, the most important thing a parent can do is to set boundaries, and to set rules, and to set these are our values, this is our mission and values as a family. These are things that we must do with AI. We need to teach it, the more especially, you know, you've heard cases of the mishandling of AI, and because it starts to learn who you are and it's learning primarily off of your data, what we need to do is to bring in, especially when it's being used for say, the medical area, we need to bring in the experts, and we need to ask what could go wrong?

Just like us, right, all of those listeners that are in risk and compliance and security, what could go wrong? How could this tool be used for good? How could it be used maliciously? And what are the loopholes in which it may falter, and the intended purpose of what it was built for is no longer met?

Those are the questions we need to start asking, but it all starts with us having that emotional intelligence to be able to ask those questions. Because when humans go to use AI, as we've spoken about before, they're gonna use it for that quick hit, for that quick response.

It is a responsibility as well for the user, but also for the designers of that tool to be mindful how it would be used. To be mindful of what outputs would be generated, and what those humans and the users would do with those outputs. It needs to take care in that. And we've become better, right, when you're putting in a prompt about a topic that the AI is not sanctioned to speak about, it will give you a response saying, I'm not, you know, I'm not able to opine on that.

And that is something that's improving, but we need to do it faster. The rate at which we use AI is now outpacing the rate of development. And that's something important, especially when we're considering it's humans at the other end. It needs to be humans at the designer end as well.

Nancy, thank you so much for being here today. I love these conversations that keep us focused on, you know, the human with AI rather than AI. And, you know, we might be this sort of tangential thing that's just a user on the side. So really appreciate your perspective that you bring.

Listeners, thank you for tuning in. Please keep the conversation going in our RSAC membership platform by visiting 1rsac.com/membership, and be sure to check out 1rsac.com for new content posted year round.

Finally, don't forget to register for RSAC twenty twenty six conference by visiting rsaconference.com/usa.