Skip to main content
Search form
  • 00:00

    [MUSIC PLAYING][Research Methods Case Study][Investigating Discrimination in the Service Industry usingAgent Based Modeling]

  • 00:12

    ROLAND RUST: Hi.I'm Roland Rust, and I'm a Professor at the Universityof Maryland, College Park.[Roland Rust, PhD, Distinguished University Professor,University of Maryland][How did you become interested in Agent Based Modeling?]I really started out in mathematics.And that was something I really enjoyed, but realized

  • 00:32

    ROLAND RUST [continued]: that I was really gravitating toward more practical problems.And so what I did was I carefully investigatedwhat fields would be good to go into whereI could use my mathematical knowledge,and business was the field that seemed best.And then I looked at the fields within business.

  • 00:54

    ROLAND RUST [continued]: And I thought, well, accounting, that's so boring.Why should I want to do accounting?And then there was finance.And I said, well, finance seems like a lotof the mathematical modeling has really already been done.Maybe I could only be incremental there.But marketing seemed like a fieldthat was a growth field in the quantitative area.And so I thought I could apply my mathematical ability

  • 01:17

    ROLAND RUST [continued]: to marketing modeling, and have that really work.And it turns out, I think, to have beena very good decision for me.[What was the purpose of this research?]Well, I can talk about one piece of research that's, I think,particularly interesting.This is something that I'm doing with a doctoral student

  • 01:38

    ROLAND RUST [continued]: of mine, Kalinda Ukanwa.And we're investigating discrimination in service.A lot of my research has to do with servicebecause it's such a large and growing part of the economy.It's about 80% of the economy in terms of GDP.So one of the things that we're looking atis what happens in bank lending.

  • 02:03

    ROLAND RUST [continued]: How can discrimination emerge even if the loanofficer is completely rational?And so we've built these models to modelhow the decision-maker works and showwhat the results of those models are.And it's really building off of some Nobel Prize-winning work

  • 02:25

    ROLAND RUST [continued]: that was done in economics.We're taking a look at the situation whereyou have an advantaged group--let's say, for example, it might be white people--versus a less advantaged group, let's saymight be black people.And we're taking a look at what happens when the decision-maker

  • 02:48

    ROLAND RUST [continued]: knows that information versus when they don't.And what we are able to show is that if youhave the information of what group the person is from,you take that information into accountand you end up discriminating against the less advantagedgroup, even if all the decision-maker is trying

  • 03:12

    ROLAND RUST [continued]: to do is maximize profits.And if you have a group-blind decision, where you cover upwhat group the person is from, thenyou have a different kind of decision that doesn'tdiscriminate, but at the same time, in the short run,you actually make less money.But what we're able to show by taking

  • 03:34

    ROLAND RUST [continued]: a look at the dynamics of this is that, over time, thiscan reverse, that by discriminatingagainst a less advantaged group that is gaining over time,such as we see with many disadvantaged groupsin the United States, for example,if you have that situation, then actually the profits

  • 03:58

    ROLAND RUST [continued]: can reverse, and you end up with the group-blind banks becomingmore profitable than the ones thattake the groups into account.But we address this problem with multiple methodologies.We have behavioral experiments that we run with real people,

  • 04:20

    ROLAND RUST [continued]: and we also have analytical modeling,where we're basically creating mathematical derivationsand proofs to show how things work.But then for the dynamics, we're using agent-based modeling,which is a computer simulation approach that basicallysimulates individual agents.

  • 04:42

    ROLAND RUST [continued]: And the agents in our case are the lenders and the customers.And by computer simulating that, youcan play things out over time and see what happensin terms of the profitability.[Were there any unexpected methodological challenges?]

  • 05:03

    ROLAND RUST [continued]: Yeah, well, there are definitely methodological challengeswith all of those three things, but the biggest issues,I think, are with the agent-based modeling.It's really a computational method,and the reason we do the agent-based modeling is that,if we tried to do all this with analytical modeling--in other words, just deriving equations--

  • 05:25

    ROLAND RUST [continued]: you can't really do it.It's an intractable problem.It's too complicated.Doing a computational approach, whichis the agent-based modeling, we can address a more complicatedsituation.So one of the things that, of course, you have to dois when you're building a computer simulation is you haveto have realistic assumptions.And so coming up with the correct assumption

  • 05:48

    ROLAND RUST [continued]: set, the ones that would really be insightful,is the biggest problem that we have.And in fact, right now, this paper that I'm talking aboutis in the review process, and the reviewersare asking us to test even more stuff.You know, that's what reviewers do.So we have to look at different kinds of networks,

  • 06:08

    ROLAND RUST [continued]: because a lot of what makes this goand a lot of what makes the group-blind banks moreprofitable is the effect of word of mouth.And so the reviewers are asking usto look at a lot of different ways in which word of mouthcan work.And there's a lot of literature in that,in the agent-based modeling literature in marketing.

  • 06:29

    ROLAND RUST [continued]: And so that's what we're really doing right now,is we're investigating different kinds of networks,and does that have an effect on the results.[What recommendations do you have for a studentor researcher looking to do something similar?]The one thing to realize about doing this kind of research,

  • 06:52

    ROLAND RUST [continued]: there is definitely a decision-makingthat has to be made between the analytical modelingand the computational modeling.And the reason that computational modelingis becoming more prevalent is that you have a--

  • 07:14

    ROLAND RUST [continued]: to use an economic metaphor--a cost of input situation where the cost of analytical modelinginput basically is constant over time,because people have always been able to domathematical equations.Sure, maybe some methods are advancing a little bit in that,

  • 07:35

    ROLAND RUST [continued]: but not very much, really.Whereas computation, the cost of computationhas gone down dramatically in the last 30, 40 years.And it's continuing to go down.So you know, the thing about the computational modelingis that you have, first of all, computation speed

  • 07:56

    ROLAND RUST [continued]: is so much faster, you have the abilityto store much more data.All this is just improving dramatically.And so the result should be that if you're a young researcher,I would say definitely you need to be in a computational partof the field.You know, you definitely need to bedoing more computational work with computer

  • 08:22

    ROLAND RUST [continued]: modeling of dynamic systems.You know, there's really not that much donewith dynamic systems in analytical modelingbecause it's so intractable.People who are doing dynamic modelinghave to do very, very simple models.And with agent-based modeling, you can relax a lot of that.A lot of those assumptions that make

  • 08:45

    ROLAND RUST [continued]: things intractable for an analytical model you cando with agent-based modeling.So you can have a much richer system.We refer to it as complexity modeling,because we're modeling complex systems.And that's relatively new.

  • 09:05

    ROLAND RUST [continued]: You know, it really hasn't been around that long.[MUSIC PLAYING]

Video Info

Publisher: SAGE Publications, Ltd.

Publication Year: 2020

Video Type:Case Study

Methods: Agent-based simulation, Computational modelling, Marketing research

Keywords: agent-based models; complex systems models; computer simulation approach; decision making in business; discrimination; network analysis; Service industry ... Show More

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:



University of Maryland Professor, Roland Rust, PhD, discusses researching discrimination in the service industry using agent-based modeling and computational methods.

Looks like you do not have access to this content.

Investigating Discrimination in the Service Industry using Agent-Based Modeling

University of Maryland Professor, Roland Rust, PhD, discusses researching discrimination in the service industry using agent-based modeling and computational methods.

Copy and paste the following HTML into your website