Reputational Risk and Data Transparency

Marie: Hello everybody and welcome to the Data Science Ethics Podcast. I am Marie Weber and today I’m here with Alexis Kassan. We are talking about a case where an algorithm was used by Chase Bank to deactivate some credit cards and the implications around that algorithm and those decisions. The reason why we’re talking about this is because it effects how people can access credit and the idea that somebody knows what parameters are allowed in terms of them being able to keep access to their credit. So a lot of times credit card companies will be looking at their list of clients and looking at the different attributes that make them a good cardholder.

Marie: So Lexy, can you talk through a little bit about what some of those different parameters are that credit card companies might use to look at their clients and identify who is a good credit card holder and who potentially might be a risk that they’d want to review?

Lexy: Sure. Before I do that, let’s back up and actually talk about what the case. So in this instance, Chase Bank has a premier tier credit card, called Sapphire Reserve, in which there are better benefits. It’s an expensive card. It’s got a pretty hefty annual fee. It’s for very high credit score individuals who are issued fairly substantial lines of credit on this card. It has accelerated earnings in particular categories and things like that.

Lexy: One of the areas where a number of card holders had voiced concerns was on Facebook and other online forums stating that they had received a letter from Chase saying that their card was being deactivated. Through sort of this group discussion, they determined that, in most cases, it was because they were shuffling their credit balance from card to card or they had made some large purchases and tried to game the point system or they had tried to use a very large number of the benefits of the card, which while they are there, there’s a certain amount of breakage or a certain amount of understanding from the bank that not everyone’s going to use every one of these benefits. So when somebody starts using a very large number of the benefits, they get a little skeptical.

Marie: I’ve also heard that from, for example, in this case, Chase’s perspective, they view those people as basically being free loaders and they’re not paying interest. They’re not bringing revenue into Chase. So then they look at it and be like, well they’re getting the benefits and that is how we set up the card. But they’re definitely utilizing it more than we expect an average cardholder to use it.

Lexy: Right. So for example, if they are constantly paying off that balance or especially if they’re balanced, transferring the balance away, that might be an indicator that they’re kind of gaming the points.

Lexy: There was one person though on these forums that didn’t seem to fit the same pattern, who had also received a letter stating that their card was going to be deactivated and they couldn’t figure out why. They were not shuffling their card balance. They weren’t doing a lot of points gaming. They weren’t really using a lot of the other benefits of the card. They were, in all respects, seemingly a great cardholder. They paid less than their balance, but more than their minimum every month. They were employed. Good track record of payment. All the things that typically you would think a card company would want. And yet they got this letter that they tried to call to figure out what was going on and they were given no information whatsoever. So Business Insider got involved to help investigate.

Lexy: What they found was that, completely outside of this person’s credit history, completely outside of their payment history, all of it, they had been involved in a lawsuit that was for the company they worked for. The company settled a claim with the federal government regarding misrepresentation of its software. Even though they’ve denied wrongdoing, they settled to avoid this big, expensive trial case, and some of the employees, including this person, had to personally pay a fee. The fact that this person’s name was associated with this firm was used as grounds to cancel their credit card because of reputational damage

Marie: So, when you say reputational damage, how is that defined?

Lexy: A number of very large fraud cases and very high profile class action lawsuits have made it more likely that banks invoke this reputational concerns reason for cancelling cards that they feel would reflect poorly on the bank. This was used to close accounts related to certain industries. For example, it might be related to it’s all entertainment or gun shops and things like that. It could be for other things where the bank feels like, if they’re seen as supporting this industry or supporting this business, it could cause reputational damage to the bank.

Marie: So from a data science ethics perspective, do you have a sense of when this reputational management may have started to be incorporated into some of these algorithms?

Lexy: It’s been around for awhile. There’s been an increased concern probably since at least 9/11 with regard to anti money laundering. The other big one from a fraud perspective was the Madoff scandal that happened several years ago. Any of these where there’s an opportunity that a bank would be involved in something that, even if it isn’t associated with terrorism or associated with some other bad cause, if it’s fraud and there is the possibility for a large impact, the bank will try to cut ties.

Lexy: In this case, it may not have been fraud. It may have just been kind of an overabundance of concern, but these types of algorithms now have more data available to them.

Lexy: More often than not, when you’re first applying for a credit card, a company will look at your credit history, your payment history, your current debt outstanding, your income… Try to calculate your debt to income ratio. If you’ve been sent to collections before, if you’ve had a bankruptcy in the last several years, so on and so forth,

Marie: How many accounts you have. How many inquiries there have been.

Lexy: Exactly all the things you can see on your credit score. That’s traditionally what they use to make a decision when they’re first giving you an account. However, they’re still keeping tabs on what you’re doing, so they’re still looking at your credit report. They are trying to determine if you’re still a good credit holder.

Lexy: They’re also looking for any signs that you might be involved money laundering. There are a lot of things that go into an anti money laundering program, but in addition to your transactional history, what this really brought to light is that they’re looking outside of their own collected data to what else is going on more broadly with their customer base to determine if that person is still a good customer. For example, scraping Lexis Nexus and identifying legal cases in which someone was named as a defendant in a lawsuit. They were able to see that that person was involved. They were named. They settled. They had to pay a penalty to the government. And they said, nope, we’re not going to continue to give you credit.

Lexy: As a consumer, we rely on what we think is going to be the information that they have, which is our credit report. We’re always kind of told the credit report is the be all end all of whether or not you get credit and everyone focuses on their credit score and so forth.

Marie: And like we were just mentioning earlier, there are certain things that you’re typically told our factor that they are looking at for your credit score and there are various different parameters. So trying to follow all of the best practices. Those are all different factors that people do have some control over and are trying to use to improve their credit score. And so it’s very interesting to find out that there are these other parameters even outside of that that can effect your ability to have or maintain a credit account.

Lexy: The reason that I think this ties into data science ethics is really one of data access. It’s the more general idea of privacy, but also the availability of information. That we as consumers of credit, we as customers of banks, presume that banks are going to use financial information and we give them access to that.

Marie: Right.

Lexy: But you don’t always think about what all the information is that they could potentially use about you from the broader internet and available data.

Lexy: This person was penalized by having their card, deactivated for a case. They thought that they were doing the right thing. They were told by their employer, we need to settle this. We need to just do it out of court. It’s going to be easier in the long run. We’re not saying that we did anything wrong, but we need to do this and they are still being penalized because that information is public domain. It’s publicly available. There’s a misconception that only certain information can be used against you by a corporation and that’s not the case. Anything that’s out there could potentially be brought in.

Marie: While that might sound far-fetched, one of our future episodes is actually going to go into how this is playing out in China where this type of thing, people’s activities both online and offline are being taken into account in terms of how people have access to credit and other systems. So the idea here is that one is consumers. It’s something to be aware of and then to as people that are involved in the data science community, thinking about the ethical implications of this and it goes back into the types of data that you include, the types of business decisions that you’re making and ultimately how you’re constructing the data and pulling everything together.

Lexy: Absolutely. I’m making sure that people are aware that you’re gathering this information. There’s a certain amount of informed consent that we give when we first sign up with a credit card that says, yes, I agree that you can pull my credit report. I don’t agree that you can go do a document search for everything I’ve ever, ever done.

Marie: This is an area where the boundaries haven’t been defined. There’s going to be, on the side of the consumer, an awareness that this is happening. And then, on the side of the companies, explaining how it’s happening and why it’s happening and being a little bit more transparent about.

Marie: The idea that somebody can just lose a credit card and not have any explanation can be very jarring for people. So even when the financial crash happened, there were a lot of people that all of a sudden started having credit cards, deactivated. They weren’t necessarily told why. That is part of the reason why now credit card companies have to say why a credit card is being cancelled. This is a type of situation where communicating that to the customers, and customers understanding how those decisions are made, is really important for everybody to be able to have confidence in the system and understand what’s going on.

Lexy: It also makes me think about some of the GDPR regulations that were put in place in Europe. In GDPR, one of the core tenets is that you have to have fully informed consent, meaning that you have to be told exactly what’s being collected, exactly how it’s being used. It has to not be in legalese. You have to agree. Now we can get into a lot as to whether or not people read it. Even though it’s not legalese. Even though it’s not pages upon pages of end user license agreement and so forth. The idea behind it is that you have more information about what the company is going to gather in terms of data and how they’re going to use it.

Marie: And even here in the US with the consumer protection laws that have been put into place, credit card agreements and things like that have been updated so they’re easier to understand. They’re clearer. Less legalese, that type of thing.

Lexy: Sure. However, I don’t think any of them specify we have the right to cancel your card based on a search for lawsuits pending with your name on them.

Marie: Oh absolutely agree. Absolutely agree.

Lexy: It probably says something along the lines of, we have the right to cancel your card at anytime for any reason. Which, while general, means that these kinds of questions can come up. But when this information comes to light about how someone has used data they found about a consumer in this manner, it really does shine a light as to how much these companies are gathering information. How much they’re gathering data. And how they’re using it for and against their customers.

Marie: Yeah, I think that’s probably a good thing to also touch on. It’s that, while it might sound like we’re being very critical of Chase, in this case, for using this type of publicly available information to close a credit card account, I think it is also important to understand the business case that they’re trying to solve. They’re trying to not let these reputation risks bring down the bank. With how much banks have had to deal with recently in terms of reputation on multiple different levels, there’s obviously sensitivity that the banks are trying to be proactive in terms of building their reputation and building trust back with the public.

Lexy: Yeah, and JP Morgan Chase has been called out numerous times and I’ve put in place some of the more stringent policies for reputational risk mitigation. They’ve had a number of times where they had made a decision to cancel accounts for a given business, for example, and then reverse course because of public outcry and so forth. When you’re one person, there’s only so much public outcry for you for a credit card. JP Morgan Chase, which is a bigger company, have chase bank, has a lot of reason to put in place these practices and to gather this information. There is sufficient reason though, also to be more transparent, to your point, about what they’re gathering, how they’re gathering, what they’re using it for, and so on.

Marie: Makes sense. Well, this has been our quick take on the evolution of the algorithms that Chase uses to evaluate the credit card holders and if they should be able to keep their accounts if they find something that would potentially have a damaging reputation associated with it. Thanks everybody for joining us for this episode of the Data Science Ethics Podcast and exploring this news article with us.

Lexy: Thanks so much.

We hope you’ve enjoyed listening to this episode of the Data Science Ethics podcast. If you have, please like and subscribe via your favorite podcast App.

Join in the conversation at datascienceethics.com, or on Facebook and Twitter at @DSEthics where we’re discussing model behavior. See you next time.

This podcast is copyright Alexis Kassan. All rights reserved. Music for this podcast is by DJ Shahmoney. Find him on Soundcloud or YouTube as DJShahMoneyBeatz.