Skip to main content
Search form
  • 00:01

    [MUSIC PLAYING]Hi.I'm Roland Rust, and I'm a Professor at the Universityof Maryland, College Park.[How did you become interested in Agent Based Modeling?]

  • 00:24

    ROLAND RUST: Agent-Based Modelingis a method that really has become prominentin recent years.And I became interested in Agent-Based Modelingbecause I was aware of work on cellular automata, whichis something that computer scientists have dealt with.

  • 00:47

    ROLAND RUST [continued]: And there's a wonderful book called A New Kind of Scienceby Stephen Wolfram that goes into great detailabout cellular automata, and all the power that you have,and using that to model things.Agent-Based Modeling is really a broader setof tools that of which cellular automata is a special case.

  • 01:10

    ROLAND RUST [continued]: The idea behind Agent-Based Modelingis, really, that you're doing bottom-up modeling rather thantop-down.In the old way of modeling complex systems,you would set things up in a kind of top-down wayin which you would determine what the system should be like,and you would write the equationsfor the models that way.

  • 01:35

    ROLAND RUST [continued]: With Agent-Based Modeling, what you do is sort of the opposite.You start with very, very simple agents.You say, OK, we've got these agents,and they interact with each other,and here's the simple ways they make decisions.And by allowing the system to just run,complex results can actually result from that.

  • 01:59

    ROLAND RUST [continued]: So even though you start out with individual agents makingvery, very simple decisions, the outcome of the systemcan be very complex.That's a change in how people model complex systems, morebottom-up rather than top-down.And that's, really, also the way artificial intelligencehas gone.

  • 02:24

    ROLAND RUST [continued]: The old way of doing artificial intelligencewas, really, top-down.You would have experts saying, OK, hereare the kinds of rules that we should have,and let's just have a big, long set of rules.And now, artificial intelligence is oftenquite different than that.

  • 02:46

    ROLAND RUST [continued]: It's, again, looking at it in more a bottom-up approach.And there is a very, very interesting bookcalled The Society of Mind by oneof the pioneers of artificial intelligence, Marvin Minsky.And that book basically explains the brainas being a lot of autonomous agentsthat are competing for attention in the brain with each other.

  • 03:17

    ROLAND RUST [continued]: So imagine you have hundreds of these sittingaround your brain, trying to control decision making,and they're battling it out with each other.And that is a very revolutionary wayto think about how things work, but that is alsohow you can build a system.

  • 03:38

    ROLAND RUST [continued]: You can build a system with autonomous agents thathave individual goals that then interactwith each other in a way that produces resultsat the aggregate level.[What kind of research questions would this method be suitablefor?]Well, one of the things that Agent-Based Modeling can dois it can model complex systems dynamically.

  • 04:09

    ROLAND RUST [continued]: And that is not something that's easy for other methods to do.And one of the reasons that that's really importantis, first of all, we are in a dynamic world,you know things do change over time,and they're changing faster and faster.So if you can't model the dynamics of the system,you're going to lose out.

  • 04:32

    ROLAND RUST [continued]: There are a lot of things that you're going to miss.So that's something that's really important.Another thing that's happening in the world rightnow is that everything is networked more and more.So you're dealing with these big networks.And networks can be thought of as individual nodes all makingindividual decisions, and then theyinteract with each other in certain ways.

  • 05:02

    ROLAND RUST [continued]: And that is exactly what Agent-Based Modeling does.These nodes are like agents, and their interactionsare modeled in the system by modelingthe individual behavior of individual agents.And so Agent-Based Modeling is really well suited to modelthe networked world in a dynamic way, which is, as we know,of course, from social media and all these other ways in whichthe internet--everything's networked now.

  • 05:36

    ROLAND RUST [continued]: And so if you don't model the system as a network system,you're really losing out.[What kind of data can be collected and analyzed usingthis method?]For the most part, this is a computer-simulation method.And so one of the things that youhave to do to make sure that you getvalid results with your simulationis you have to make sure that the beginning assumptions arecorrect.

  • 06:04

    ROLAND RUST [continued]: And so often, that means that youhave to validate the simulation by using actual data.And in fact, in the review processfor Agent-Based Modeling papers, that'swhat you see so that the reviewers are expectingthe particular parameters that are set up in the modelto be validated on real life.

  • 06:29

    ROLAND RUST [continued]: And so that's what you end up doing.You start with the real world and use the real worldto parameterize the model, and then youlet the simulation run.And that's how that works.[Can you give an example of where you have used Agent BasedModeling and what you learned in the process?]One of the examples that we've done with Agent-Based Modelinghas to do with speed of innovation.

  • 06:57

    ROLAND RUST [continued]: In other words, we set up the decision makers suchthat they have particular speed.And, of course, with speed of innovation,you also have a trade with quality.The faster you are, the less quality you're going to have,on average.And so the question is, when should you be fast,and when should you be slow?

  • 07:22

    ROLAND RUST [continued]: Or the way we put it is--and this is the title of one of our papers--"Don't Do it Right.Do it Fast."And if you take a look at what is happening in Silicon Valley,that's really a lot of what you see.You see these beta versions coming outthat really are crap.And they are fast, but they don't really work that well.

  • 07:49

    ROLAND RUST [continued]: But you have to do it.Otherwise, you lose out in that market.And that is the case with a lot of markets.Sometimes you have to be fast, and sometimes you don't.So we can investigate that by setting updifferent agents that have different speed-qualitytrade-offs and letting them competein the marketplace in this Agent-Based Modeling simulationand figure out, under what conditions is bestto be fast versus high quality?

  • 08:21

    ROLAND RUST [continued]: [What tools and resources are helpful for students lookingto use this method?]Well, I think that, in general, what studentsneed to be good at these days more and moreis computer-science techniques.Agent-Based Modeling is a computer-science approach.It came out of computer science.

  • 08:42

    ROLAND RUST [continued]: And that is what students need to be workingon more than anything because again, with the world becomingmore computational, and with the academic world becoming morecomputational, you have to have the abilityto handle computational methods.

  • 09:05

    ROLAND RUST [continued]: And the field that's good at that is computer science.Computer science is, by and large,a little bit quicker and dirtier than economics.Economics is kind of, let's do everything exactly perfect,and we will polish everything to a fine sheenand make sure that it's absolutely totally, totally,totally rigorous.

  • 09:27

    ROLAND RUST [continued]: Well, computer science at a different point of view.For example, people don't even really pay that much attentionto journal articles in computer science.They pay attention to conference papers.And you send a paper to a conference in computer science,you're going to get a very quick decision.And sure, that means sometimes, those decisionsare going to be wrong, but that way, the word gets outvery, very fast, whereas in the more economics-based fields,such as marketing has tended to becomein the quantitative area, you don't really have that.

  • 10:08

    ROLAND RUST [continued]: You send a paper into a journal and it might be therefor years, you know, polishing everything upjust a little bit more, little bit more, round after round.And as a result, a lot of the marketing fieldis years behind, and the stuff that's publishedis sort of out of date.So I think we need to pay much more attention to time,and we need to be much faster in our way of doing research.

Video Info

Publisher: SAGE Publications Ltd.

Publication Year: 2020

Video Type:Tutorial

Methods: Agent-based simulation, Marketing research

Keywords: agent-based models; artificial intelligence; bottom-up approaches; complex systems models; computer science; computer simulation approach; conferences; journal articles; network analysis; network data model ... Show More

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:



University of Maryland Professor, Roland Rust, PhD, discusses types of research questions, data collection and analysis, and tools and resources suitable for agent-based modeling and computational methods.

Video Info

Publication Info

SAGE Publications Ltd.
Publication Year:
SAGE Research Methods Video: Market Research
Publication Place:
United Kingdom
SAGE Original Production Type:
SAGE Tutorials
Copyright Statement:
(c) SAGE Publications Ltd., 2020


Roland Rust

Segment Info


Segment Num: 1


Segment Start Time:

Segment End Time:


Things Discussed

Organizations Discussed:

Events Discussed:

Places Discussed:

Persons Discussed:




    Huang, M, Rand, W, & Rust, R.T(2016).Don't do it right, do it fast? Speed and quality of innovation as an emergent process.. Paper presented at ICIS, Dublin,


    Minksy, M(1988).The Society of Mind. New York, NY:Simon and Schuster,


    Rand, W, & Rust, R.T(2011).Agent-based modeling in marketing: Guidelines for rigor. International Journal of Research in Marketing, 28(3),181-193.


    Wolfram, S(2002).A New Kind of Science. Champaign, IL: Wolfram Media,

Methods Map

Agent-based simulation

An application of computer simulation to the social sciences useful for studying the complexity of social systems. Agents are software objects programmed to have autonomy and goals, and to act and react towards their environment and other agents within it. Running the simulation allows emerging patterns of activity to be observed and compared against those found in the real world.
Agent-based simulation
An Introduction to Agent-Based Modeling & Computational Methods

University of Maryland Professor, Roland Rust, PhD, discusses types of research questions, data collection and analysis, and tools and resources suitable for agent-based modeling and computational methods.

Copy and paste the following HTML into your website