How Faculty Director Kai Larsen went from designing banking systems to training tomorrow’s analytics leaders
For Faculty Director Kai Larsen, the future of business analytics presents exciting problem-solving possibilities and challenging questions about ethics. He’s been honing his expertise in analytics for over 30 years, and his unconventional journey lends a world perspective to his teachings. Through his unique professional experiences and insights on this dynamic discipline, he’s training students to drive value in the workplace.
The problem-solving possibilities of analytics
At the beginning of his analytics career, Professor Larsen served as a consultant in bringing Norwegian banking into the future. His work helped create the first Norwegian Internet banking system, as well as a central billing system—which altered the way people received bills in the mail and automated their payment process.
“Instead of a bill per day, all bills would arrive in one envelope every two weeks,” said Larsen. “If you did nothing, then they would just get paid. So that’s sort of a case of automating an existing process and making it easier.”
Using analytics to design automated systems was an exciting combination for Larsen, and it meant the ability to solve problems in a way that hadn’t been possible before.
“To me, it was always about innovating processes—automating and innovating,” said Larsen. “So, can you take something people are doing and make it way easier? Analytics and artificial intelligence is just one more really cool toolset for tackling processes and problems."
An introduction to machine learning
While a consultant in Norway, Larsen began working on expert systems and automating processes that normally require human expertise. By interviewing area experts, such as loan processing officers, he was then able to create enormous collections of ‘if–then’ statements based on their knowledge. This would serve as a foundation for automating these processes with machine learning.
“If someone walked in asking for a business loan, a loan processing officer would follow up with a set of questions. At the end of those questions, that officer would then say if you could get a loan or not. So, what we would focus on there was developing systems that built that expert’s knowledge into technology. These ‘expert systems’ were expensive to build, and it was never quite clear whether automating past processes would lead to fair and correct answers.”
Shortly after his consulting work on expert systems in Norway, Larsen moved to the U.S. to pursue a PhD. At that time, he became more familiar with machine learning, its role in cutting the individual expert out of the process, and it’s relationship with the human knowledge and biases it’s created from.
Defining systems from data
In using analytics to automate processes, Larsen explained that the method relies upon past evidence, or the history of data, that’s available. That data makes it possible to then determine the guidelines, or rules, that will drive an automated system.
“If you’ve already given out ten thousand loans, then we know how those people who received loans behaved—and we can set up a definition of success or failure,” said Larsen. “So maybe they didn’t make a payment in 60 days, or maybe they paid the whole loan on time. You can define what ‘success’ and ‘failure’ for a system looks like.”
In addition to people’s loan repayment behaviors, all of the collected information about those individuals is also fed into the system—such as purchase history, credit rating, demographic information and more. Using all of this data, an expert—or in automated cases, the algorithm—tries to figure out the ‘rules’ that will drive future decisions, like whether or not someone qualifies for a loan.
“But there’s enormous bias in that decision-making, right? We know for a fact that certain groups of people, often because of their race, ethnicity or gender, have traditionally had a hard time getting loans because of the biases of the loan officers and other systemic issues.”
For Larsen, this question of human biases in analytics and the consequent ethics of machine learning means it’s crucial for students to be properly trained on these issues before they enter this increasingly complex discipline.
“Initially, we were excited about machine learning because we would give that algorithm all the data and it would come up with a ‘model’ that behaved like an expert. But what people often forget in cases like that, is that the machine and its data is still just based off the biased intelligence and social systems it was created from.”
Training tomorrow’s leaders
Now an associate professor and faculty director at Leeds, Larsen puts his expertise to use by preparing future generations in the field. His comprehensive approach to teaching business analytics includes shaping students into experts while incorporating ethics and social issues into their coursework. From the undergraduate Business Analytics Track to the Master’s in Business Analytics program, students are challenged to think critically about data and to create models in ways that glean insights and drive value. The skills that Larsen’s students develop throughout the programs mean that, by the time they graduate, they’re prepared to operate in a variety of business landscapes and to always have an eye for ethical business.
“My classes tend to start very technical at the undergraduate level. First, I teach students all about transforming data and everything they can do. Then, we go into machine learning and how to develop models. They have to have these skills before we can teach them how to use this knowledge ethically. And for our ten-month graduate program, we first make students into business analytics experts and then get more into the social issues in the spring.”
Read more on Data, ethics and AI: Learning about the Leeds MS in Business Analytics to discover more insights on artificial intelligence from Faculty Director Kai Larsen.