by Summary:
Problems are Timothy's passion and this passion led to a career of two arcs. The first arc spanned twenty-five years of working with clients in corporate, government, military and non-profit environments to solve their most challenging problems. The second arc was his personal journey along the way studying problems and the best methods to solve them. That journey led to a deployment with the US military in Afghanistan where he first encountered the Systems Thinking aspect of the science of System Dynamics. Three years later he left his position at IBM as Chief Methodologist of Lean, Six Sigma and Agile to fully commit himself to system dynamics. He pursued a Master’s and PhD in System Dynamics publishing research ranging from the reduction of violence and instability to business
Topics:
Steve Keen considers the following as important:
This could be interesting, too:
Problems are Timothy's passion and this passion led to a career of two arcs. The first arc spanned twenty-five years of working with clients in corporate, government, military and non-profit environments to solve their most challenging problems. The second arc was his personal journey along the way studying problems and the best methods to solve them. That journey led to a deployment with the US military in Afghanistan where he first encountered the Systems Thinking aspect of the science of System Dynamics. Three years later he left his position at IBM as Chief Methodologist of Lean, Six Sigma and Agile to fully commit himself to system dynamics. He pursued a Master’s and PhD in System Dynamics publishing research ranging from the reduction of violence and instability to business
Topics:
Steve Keen considers the following as important:
This could be interesting, too:
New Economics Foundation writes Moving forward
Dean Baker writes Health insurance killing: Economics does have something to say
NewDealdemocrat writes Retail Real Sales
Angry Bear writes Planned Tariffs, An Economy Argument with Political Implications
| Problems are Timothy's passion and this passion led to a career of two arcs. The first arc spanned twenty-five years of working with clients in corporate, government, military and non-profit environments to solve their most challenging problems. The second arc was his personal journey along the way studying problems and the best methods to solve them. That journey led to a deployment with the US military in Afghanistan where he first encountered the Systems Thinking aspect of the science of System Dynamics. Three years later he left his position at IBM as Chief Methodologist of Lean, Six Sigma and Agile to fully commit himself to system dynamics. He pursued a Master’s and PhD in System Dynamics publishing research ranging from the reduction of violence and instability to business application. Timothy believes Structured Systems Thinking is one of the most powerful problem-solving methods available today. Website: https://dialecticsims.com/ |
Thanks for coming on the show Timothy! Great discussion as always
Thanks to Steve, Ty, Daniel & Mike for having me! Had a good time!
Jessie Ventura vibe meets thriller novelist name. Tim is a superlative communicator.
Thank you for not mistaking me for Tom Clancy, that's a rarity I appreciate! 🙂
@Dialectic Simulation Consulting, LLC It would be hilarious though. Dynamic modeling by day, thriller novelist by night.
Plus Mike wants to gamify modeling. It could be another call of poopy
Ty — what the heck. On your own podcast, your audio is much better. Must be the swearing.
@Stephanie Hughes asked: "How do you link disparate models given that assumptions and data sets are different? Run Linear or Multivariate regressions?"
Entity transfer between models is a good way to handle differences in aggregation. For example E-SAM runs at the level of an operational theatre (e.g. Iraq & Syria, Ukraine, Myanmar etc.) so it's not going to be useful for simulating an individual city or even an individual neighborhood. That'd be better handled by a different model built to different purpose. However with entity transfer you can take the values at a given point in time in the regional model, transfer those to the city-based model to instantiate it, run that city simulation, and then entity-transfer the outcomes back up to the regional model.
As for evaluation of forecast accuracy regression is challenging in oscillating or cyclical patterns. MAE, MSE (RMSE) or Thiel's Inequality statistics decomposing MSE into Us, Um, Uc are better tools; though there is no one "right" answer. Depends on circumstance and level of confidence your audience needs.
@Owen Wall: "I have a further question about whether Tim has been able to model how the US could step back from occupying Syria without a resurgence of ISIS"
We proposed a policy (see paper below) that involved very limited US involvement relying instead on indigenous local actors (e.g. Kurdish SDF) while staying at arms length from others (e.g. Iraqi PMC's). We warned however that if the underlying grievance between state actors (Syria and Iraq) didn't reconcile the tensions between Arab Shia majorities and Arab Sunni and Kurdish Sunni majorities that ISIS would likely return as a clandestine terrorist group within 5-8 years, exploiting the grievance. The policy we proposed, or something close to that, worked well in terms of pushing ISIS back through the recapture of ar-Raqqah and Mosul. However President Trump escalated the air campaign beyond reason, and the subsequently abandoned the Syrian Kurds in 2019 when he ceded a 'security buffer' to Erdogan of Turkey. As we forecast, ISIS has returned in a clandestine terrorist capacity over much of the area, and even has begun operating as an insurgency again in certain localities. (See more of that on this blog article: https://infomullet.com/2021/01/21/isis_returns_to_baghdad/ )
Paper: “Application of Emerging-State Actor Theory: Analysis of Intervention and Containment Policies †,” Systems, vol. 6, no. 2, p. 17, May 2018, doi: 10.3390/systems6020017
@SlugiuesRex "Are you using something like Markov Chain or Monte Carlo structures to encapsulate each grouping ??"
Markov chains and monte carlo are probablistic (statistics), system dynamic models are deterministic(calculus). For the most part system dynamic models assert causal relationships and then model what happens when those relationships interact over time. So we wouldn't use markov chains or monte carlo when constructing a model. However both play a role in completed simulations.
We can use monte carlo analysis to brute-force a field search over policy spaces quickly, identifying outlier outcomes and then reverse engineering the causal steps that led that outlier to occur. (You still need to translate this into real-world processes but it can give you a hint. Hidden markov chain analysis is used to compare system dynamic runs over time to real world behavior, and suggest what structure might be missing. So they do play a role in the range of tools we use – but we don't use them generally to build the initial simulations.
I found the guest didn't fit well with the show's theme. Second hour better than the first which I found flat.
Though Timothy is not economic focused per se, from a system thinking perspective he operates in the same manner as Mike, Steve, or myself. So I enjoyed the whole two hours, and was grateful Tim took the time on his Saturday to come join us. That said I'm glad you got something out of the second half show!
@10:00 way too blokes-world tone guys. And not for the first time,
I came here for Tim but stayed for the conversation.
I hope there was at least some insight provided. We mainly just have fun and talk about complex systems. Thanks for coming for Tim, and sticking around for the full conversation.
This is so fascinating! I'd love to explore the use of system dynamics in schools and communities to reduce violence.
@Hugh N asked: "was there a team in IBM that did system dynamics at that point?"
Not that I was aware of. IBM in those days was all-in on Watson and getting corporate oxygen for alternative approaches to was very difficult. There may be individual practitioners here or there, but nothing formal. Several of us were trying to bring system dynamics in, pointing out the flaws of Watson and how simulations could help mitigate those. But the corporate tide was against us. I left IBM in 2015 to focus on system dynamics, so I have no idea whether there are any groups today. The concerns over Watson I think proved themselves to be valid over time, and I see similarities in the current hype cycle regarding AGI/AI.
You know your guests models are being used to overthrow sovereign governments, right?