Skip to main content
Search form
  • 00:00

    [MUSIC PLAYING][Researching Day Trading Teams Using Balance TheoryAnd Social Network Analysis]

  • 00:10

    PJ LAMBERSON: My name is PJ Lamberson,and I'm an assistant professor of communication at UCLA.The work that I do is mainly mathematical and computationalmodeling.I study social interactions and social influence.Social networks.How things spread from person to person.And how teams collaborate.

  • 00:33

    PJ LAMBERSON [continued]: It's kind of a long story.I used to be-- my PhD is in mathematics.But I realized that I wanted to study something thatwas more connected to human behavior and had,maybe, more of an impact--direct impact, at least-- on choices that peoplewere making or society.And I discovered that there was a political scientist who

  • 00:55

    PJ LAMBERSON [continued]: had done some work that was similar to some workthat I had done previously.I contacted him.I ended up going to work with him for a postdocat the University of Michigan.And that's what really launched me on this pathto studying social influence and social interaction.

  • 01:16

    PJ LAMBERSON [continued]: One of the recent projects that's actually ongoingthat I've been working on is a papertrying to understand how teams collaborate.And what the best network structure for communicationamong those team members would beto lead to them to be better at solving difficult problemsor finding new innovations.

  • 01:40

    PJ LAMBERSON [continued]: This is really a theoretical paper,but it was driven by two other papers that existed outin the research literature that had conflicting results.Basically, one paper was a computational modelthat implied that the best thing to do for teams was to keepthem as spread out as possible.And have as little slow down communication among the team

  • 02:03

    PJ LAMBERSON [continued]: members.So that allowed each individual team memberto explore and look for solutions on their ownbefore they were influenced by their collaborators.But there was another paper out therethat used an experiment that looked at almostthe exact same setup, but found the complete opposite result.That, instead, you should have team membersas connected as possible and communicating

  • 02:23

    PJ LAMBERSON [continued]: as quickly as possible.And so I had some ideas about maybewhat was driving the difference between these two results.And that led me to pursue this researchand come up with a model myself, based on some previous workthat I had done that allowed me to explainwhat was different between these two findings.

  • 02:48

    PJ LAMBERSON [continued]: The first thing I had to do was Iwanted to build a model that couldexplain both of these results and the differencebetween them.So I needed to make sure that the model Ihad under some conditions would actually replicate thesetwo extremes.So we put the model together.It took quite a while to figure outjust the details of exactly how we were going

  • 03:08

    PJ LAMBERSON [continued]: to model the team members.And one of the key ingredients in this model--and one of the reasons that I thoughtexplained part of this difference--was that in the original paper, individual team members wereall essentially the same, except for the information that theyhad.There weren't differences in how they approached

  • 03:28

    PJ LAMBERSON [continued]: the problems they were solving.And so, I wanted to model team membersas having diverse approaches to problem solvingor diverse specialties or expertise.Because really, that's what creates synergy between teammembers in real-life.And I thought that's what's drivingthe effective faster communicationin the real-life experimental results.So we had to figure out how we were

  • 03:49

    PJ LAMBERSON [continued]: going to model team members that have diverse expertiseand diverse specializations.And for that, I tried a lot of different things.And I talked to other people.I talked to people who are experts in this field, as well.And got different advice and eventually, wehit upon the right specification that allowed us to really get

  • 04:12

    PJ LAMBERSON [continued]: that synergistic effect in this effective collaborationthat we were looking for.The main tool that I use in this researchis what we call an agent-based model.It's a computational simulation of people workingto solve problems where we representeach individual on the team as like a it's a simulated person,

  • 04:36

    PJ LAMBERSON [continued]: like you would see in a computer game-- almost.And those simulated people interact.They have rules for how they are connected to one anotherin a network.They have rules for how they go about tryingto solve the problem.Rules for how they communicate.And then once we setup this simulated team,we let them go about trying to solve the problemand see what solution they get.And we do that many, many, many, many times.

  • 04:58

    PJ LAMBERSON [continued]: And we then gradually change parameters of the model.So it's almost as though we're running an experiment,but except in real life, you don'thave such good control over the individuals in your experimentby here.Because these are simulated people,we can just very precisely change just one little aspect.In this case, change the network structureor change exactly how they communicate with one another.

  • 05:21

    PJ LAMBERSON [continued]: And that allows us to run this computational experimentand see which factors drive how effective theyare at solving the problems.Well, to do this kind of research you'redefinitely going to have to have some programmingskills of some type.There's lots of different kinds of languages

  • 05:41

    PJ LAMBERSON [continued]: that you can write these computational models in.You can use the statistical package, R.You could use Python or C++.For this particular project, I havea postdoc who's working with me and he is an expert in Matlab.So he's using Matlab for a lot of the coding.The statistics that we're using aren't particularly complex,

  • 06:03

    PJ LAMBERSON [continued]: but you do have to have some basic statisticsunderstanding in the end to analyze the results.And ultimately, we'd like to be able to prove some theoremsmathematically about what we're seeingat the extreme cases, at least.And so there, we're moving up to some more difficultmathematical background-- linear algebra and beyond, stochastic

  • 06:27

    PJ LAMBERSON [continued]: systems and so forth.As far as unexpected challenges go,I think there is always a challenge whenany computational model of just how long it takesto run the simulations.And so one of the things that you have to work hard to dois figure out how you can optimize your code so that you

  • 06:50

    PJ LAMBERSON [continued]: can run the simulations more quickly because youwant to run a lot of them.And you want to be able to sweep over many different parametersettings.So figuring out the right way to do that is always a tricky bit.I think the best tip I have is, actually, not so much

  • 07:11

    PJ LAMBERSON [continued]: about writing optimal code because I'm actually nota computer scientist by training.I'm not the world's best coder.But I think one step before that,when you're thinking about the model itself--you want to come up with really, the simplest representationthat you can captures all of the relevant phenomena.

  • 07:31

    PJ LAMBERSON [continued]: And so sometimes, really in the design stagebefore you move on to the actual technical coding,there are choices you make that can lead to--oftentimes, the model that's most computationally efficientis also actually, the most understandable.And the best for shedding light on your results.So doing that a step ahead really helps.

  • 07:56

    PJ LAMBERSON [continued]: One thing that's just always important to keep in mindis just to have fun and look for projectsthat you find exciting and interesting yourself.Because whenever you start a research project,it always takes a long time to get through.And there will be challenges that come up.And if it's not something that you're really interested in,

  • 08:17

    PJ LAMBERSON [continued]: it can be daunting.But when you're having fun with it, then it makes it easy.We're at a stage that always happens in research,where we've formed a question.We had a hypothesis.We went through all of the work.We've done the modeling.

  • 08:37

    PJ LAMBERSON [continued]: And we have a lot of results.And at this point, we really just need to write it all upand make sure that we've dotted all our I's and crossedour T's.We need to check the robustness to different kindsof variations in the model and make surethat there's nothing specific about what wedid that is driving everything.And look at some other variations.

  • 08:58

    PJ LAMBERSON [continued]: But really, we're trying to just wrap this up in a nice packageand send it out.Some advice I would give to somebodywho's new-- first of all, if you'recoming in from another field, that's great.I came into this from another field, myself.And coming in from other field gives you, in some ways,

  • 09:19

    PJ LAMBERSON [continued]: an advantage of coming with a fresh set of eyes.And actually, research we've done shows that what we callknowledge transfer-- like bringing in new ideas fromoutside--can lead to some of the biggest discoveries.So don't feel afraid that you're some kind of an imposterin the wrong place.Because, actually, computational social science--

  • 09:40

    PJ LAMBERSON [continued]: in particular-- is a super interdisciplinary fieldof research.And it's a great place to be if you're coming from outside.And having said that-- another piece of adviceis to look to other fields for inspiration.So a lot of my work has from looking at research in biologyand transferring insights there into economics-- for example.

  • 10:03

    PJ LAMBERSON [continued]: Or taking results from economics and transferring theminto organizational theory.And so that's a great place to get inspirationfor something that you can--for a problem to solve.

  • 10:25

    PJ LAMBERSON [continued]: One of the things that's very exciting in computationalsocial science and that draws a lot of the attentionis this issue of what people call big data.There's all this new digital dataout there on human behavior.And that is driving a huge amount of exciting researchand discoveries.However, one of the things I think has happened recently

  • 10:46

    PJ LAMBERSON [continued]: and I think is really beneficial for our fieldis that we are starting to take a really serious closelook at exactly what we're doing in that research.And maybe tampering down our enthusiasm a littlebit to actually be really carefulthat the research that we do is not just blindly

  • 11:08

    PJ LAMBERSON [continued]: mining data for correlations.But actually following the scientific methodof starting with a theory, and generating a hypothesis,and testing that hypothesis.And being sure that we're really being extra careful.There's a few times in the literature where--I guess-- we've been burned a little bit, where there havebeen findings that didn't last or we realized maybe

  • 11:32

    PJ LAMBERSON [continued]: we were a little overenthusiasticas a discipline.And I'm really happy to see that we're now coming togetheraround this idea that we have to be more carefuland that big data is not just the answer to everythingon its own.But requires still a careful process,marrying that with theory and everything

  • 11:53

    PJ LAMBERSON [continued]: that we learned about small data, and sampling,and sample bias, and so forth.So if you're interested in studying teamsand how they collaborate, there are lots of great placesto look.One of my favorite books is a book called The Difference,by Scott Page.That's a book that talks really about how

  • 12:17

    PJ LAMBERSON [continued]: the diversity of the members of a teamcan drive its performance.And sometimes teams that have diverse memberscan actually outperform teams thathave individually more skilled members,but that team is not diverse.So that's a great book.And it really introduces this basic ideasof mathematical modeling.And how we can use models to understand that team

  • 12:40

    PJ LAMBERSON [continued]: performance.[MUSIC PLAYING]


P. J. Lamberson, PhD, Assistant Professor of Communication at UCLA, discusses his research using computational models to study effective communication in groups, including an example of a recent project, how the research was conducted, the main tools used, the key skills and techniques needed, unexpected challenges encountered and overcome, advice for doing computational modelling and for students new to the field, exciting prospects for computational social science, and recommended resources for student interested in the field.

Looks like you do not have access to this content.

Studying Effective Communication in Groups Using Computational Models

P. J. Lamberson, PhD, Assistant Professor of Communication at UCLA, discusses his research using computational models to study effective communication in groups, including an example of a recent project, how the research was conducted, the main tools used, the key skills and techniques needed, unexpected challenges encountered and overcome, advice for doing computational modelling and for students new to the field, exciting prospects for computational social science, and recommended resources for student interested in the field.

Copy and paste the following HTML into your website