0 likes | 15 Views
The field of System Dynamics faces the challenge of staying relevant by embracing new methods and practices. Various commentators discuss the need for rigor, hybrid modeling approaches, professional networks, and emphasis on implementation. Responses from John Sterman highlight the importance of new data acquisition methods, hybrid modeling, learning from other disciplines, and caution against normalizing low standards.
E N D
SYSTEM DYNAMICS AT SIXTY SYSTEM DYNAMICS AT SIXTY O ORIGINAL , COMMENTARIES RIGINAL P PAPER APER, C OMMENTARIES, , AND AND R RESPONSE ESPONSE James Paine System Dynamics Group MIT Sloan School of Management
SYSTEM DYNAMICS AT 60: THE PATH FORWARD John Sterman (2018) John Sterman (2018) “Despite many successes, too many in the field continue to develop models the way Jay did in the 1950s, 60s and 70s, adhering to outdated methods for model development, testing and communication, and closing themselves off from important developments in other fields. The result is a gap between best practice and what is often done.” “..we should not hold fast to methods for the development and analysis of system dynamics models that have outlived their useful life and for which better methods now exist.” Development of SD was result of synthesis of multiple fields and state-of-the-art computing What is ‘state-of-the-art’, and the very fields from which this art emerges continues to change Challenge is to stay relevant. SD is not just a static set of methods • • • 1
COMMENTARY System Dynamics Review: Vol 35, Issue 1, 2019 System Dynamics Review: Vol 35, Issue 1, 2019 Homer • • Applaud call to rigor, but concern that proposed new bar is too high Start small: encourage collecting of multiple historical time series data. Many papers have 1 or even no historical data. Anderson • Expands on and presents concepts of hybrid models • “If a single modeling methodology can prove to be a reasonable approximation for addressing a problem, then using a single methodology may be best. If not, a hybrid approach may provide a better option.” Schwaninger • Emphasizes need for professional networks and support to build the new and extra competency that Sterman suggests • “The problem is, the above ideas are espoused easily, but rarely put into effect in modeling practice” … “we need more training to achieve modeling excellence—I mean in-depth training” 2
COMMENTARY Randers • Argues for need for more emphasis on implementation • Proposes review that specifically focuses on implementation methods – Focus on impact (i.e. policy or behavior change). Separate review of why implementations fail Morrison • General agreement but raises concern that paper places to much emphasis on big data and statistical methods at the risk of restricting the art of System Dynamics modeling • Seems to stand in contrast, partially, with Homer -> Do not privilege calibrated and data centric models as it may silence other styles of SD work • Three Traps: 1. Forgetting that Models are Models 2. Numerical data bias 3. The model speaks for itself • Simple models that can be clearly understood will often be more effective and persuasive than large, numerically sophisticated models • My opinion: The trap here comes from a potential misunderstanding of Sterman’s message 3
RESPONSE TO THE COMMENTARY System Dynamics Review: Vol 35, Issue 1, 2019 System Dynamics Review: Vol 35, Issue 1, 2019 Sterman’s Response • Opens with summary of TPF • Emphasizes that over past 60 years new methods for both quantitative and qualitative data acquisition and confidence building have been developed • Full agreement on Ed Anderson's points on Hybrid modeling • Agree with Homer on use of multiple time series to build confidence • ‘Looks good to me’ test is not acceptable. • This is necessary, but not sufficient contention for good work • Rander’s agenda on focus on implementation is agreed with but with caveats • Take the time to learn from other disciplines on implementation • Some polite disagreement that there has been significant progress in modeling methodologies. • Caution against normalization of low standards 4
RESPONSE TO THE COMMENTARY System Dynamics Review: Vol 35, Issue 1 System Dynamics Review: Vol 35, Issue 1 Sterman’s Response (continued) Argues that tension present in Brad Morrison's piece is not present Strong disagreement with the idea that raising the bar for SD work discourages novices to the field • Clarifies that the concerns he raised are not necessarily present in John’s original piece • Important response, as Morrison likely is not the only one who had the concerns raised in his commentary • Commentary allowed response to clarify that goal is to stay rooted in whatever the state-of-the-art is, in the founding spirit of System Dynamics • • 5
RELATED WORK Chapter 21 of Business Chapter 21 of Business Dynamics Dynamics Industrial Industrial Dynamics Dynamics Looks Good Enough test isn’t enough: Original basis for SD and quoted throughout the above papers and commentaries • “…to uncover flaws in the structure or parameters of the model and assess whether they matter relative to the purpose. Instead of showing how well your model fits, you should point out to your clients all the places it doesn’t. These discrepancies mark the trails that can guide you to erroneous parameter estimates and inappropriate assumptions you should revise before using the model for policy analysis” 6