What makes World Modeler different?

In anticipation of Mark Zangari’s upcoming talk on Agency Theory at MLConf Seattle, a question appeared on Quora yesterday asking  “What is the World Modeler platform and how does it compare to similar platforms?”  I thought I’d answer the question here.

(If you’re coming here from Quora, then skip to the bottom for the new stuff)

As explained in two kinds of software, some software systems have a “world model” at their core, which is a “cartoon” simplification of something in the real world.

Familiar examples include Microsoft Project (which models tasks and resources), Amazon collaborative filtering (which models user preferences), Google advertising (which models the likelihood that users will click through on certain ads), and telecom and credit card fraud systems (which model the behavior of fraudulent persons).

The picture at the top of this article shows a high-level view of the World Modeler approach.  Any software that includes a World Model (whether using World Modeler or some other approach) includes four components:

  1. Data: from the usual sources (SQL, Excel, Hadoop, IaaS, and so on)
  2. A Model, in two parts:
    • An analytic, ML, or statistical model built in a modeling tool like R, SPSS, or SAS or a special-purpose tool like Theano, H2O, Caffe, or Torch.
    • A systems model-often in the form of a decision model-that links the multiple analytic models together
  3. A Decision-making component. This can either be an automated system, like we might see in churn or fraud systems or as in Amazon collaborative filtering or Netflix; or it might be a human.  If the decision-maker is human, we need two additional components:
    • Data visualization: to allow the data to be effectively “uploaded” into the human brain, as you can hear about in Mark’s talk from USF last year.
    • Decision visualization: to allow the model to be understood by the stakeholders in the decision.  Mark and I invented this part.
  4. Surrounding infrastructure: traditional code (javascript, C#, asp, python, php, and so forth) that connects it all together into a complete enterprise app.

World Modeler™ was built to “fill in the gap” between the technology, culture, and expertise of modelers, and others within an organization who need to bring those models to life in enterprise-grade, deployed and agile systems.  It contains about two million lines of general-purpose code that integrates with other tools to provide a framework to supply the above four functions.  This way, our customers (or partners) only need to write a thin layer of code to get a fully customized, enterprise-grade system at the low cost, low risk, and short time of COTS (commercial off-the-shelf) systems.

So to answer the specific question about how World Modeler compares to other kinds of software, here are some examples:

  • Like analytics tools like SPSS, or SAS or a special-purpose tool like Theano, H2O, Caffe, or Torch, we deliver models, but we are the “bones” to their “meat”, providing a rapid-deployment, risk-reduction infrastructure to deploy systems that have these technologies at their heart.
  • Like systems modeling tools like Stella and decision analysis tools like Analytica, we allow the user to use a drag-and-drop interface to build complex models.  Unlike those tools, though, our simulation interface allows the user to visualize the parts of the system moving as the simulation progresses (see this page for a web-based example and this part of this video for an explanation and a view of the desktop version).  We also continue with the user beyond the analysis phase of a project into full production.
  • Like complex adaptive systems modeling tools, our systems model includes feedback effects and other complex dynamics, which are visualized as the model runs.   We can do Monte Carlo simulation and various kinds of optimization.   We don’t include agent-based models at this time, but it would be an easy addition if needed.
  • Like Decision Intelligence consulting-focused firms such as AbsolutData and Informed Decisions, we offer Decision Intelligence consulting services as well.
  • Like AHP tools like Transparent Choice, we use computer support to help with making better decisions, but again, as described here, we are the infrastructure to their algorithms.
  • Like domain-specific models like the World Model at the heart of the International Futures (IFs) model maintained by the Frederick S. Pardee Center for International Futures, World Modeler can be used for modeling the future of many interacting complex systems worldwide. However, World Modeler is a modeling tool, into which users add their models and data, rather than the model and data themselves, which is one of the great values of IFs.   Both systems work in a browser, so that they can be used collaboratively by teams worldwide.  Unlike IFs, however, World Modeler adds agency modeling capabilities, to answer the question “How will the world change if we alter this policy?” .  In this way, World Modeler is complementary to IFs, because IFs forecasts could be used within the World Modeler framework to predict important global trends.

Have I missed any categories above? Feel free to make suggestions in the comments below and I’ll try to add them to the article.

World Modeler is a new kind of SW, integrating ML/more into a deployed enterprise app Click To Tweet

There’s an emerging ecosystem around World Modeler and Decision Intelligence/Decision Engineering.

Here are a few additional resources if you’d like to learn more:

Thank you for asking!

Lorien Pratt

Pratt has been delivering AI and DI solutions for her clients for over 30 years. These include the Human Genome Project, the Colorado Bureau of Investigation, the US Department of Energy, and the Administrative Office of the US Courts. Formerly a computer science professor, Pratt is a popular international speaker and has given two TEDx talks. Her Quantellia team offers AI, DI, and full-stack software solutions to clients worldwide. Previously a leading technology analyst, Pratt has also authored dozens of academic papers, co-edited the book: Learning to Learn, and co-authored the Decision Engineering Primer. Her next book: Link: How Decision Intelligence makes the Invisible Visible (Emerald Press), is in production. With media appearances such as on TechEmergence and GigaOm, Pratt is also listed on the Women Inventors and Innovator’s Mural. Pratt blogs at www.lorienpratt.com.

You may also like...