Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bridging the Unavoidable A Priori: A Framework for Comparative Causal Modeling

Peter S. Hovmand, Kari O'Donnell, Callie Ogland-Hand, Brian Biroscak, Douglas D. Gunzler

TL;DR

The paper addresses the challenge of reconciling system dynamics (SD) and structural equation modeling (SEM) to enable robust causal modeling for data science and AI/ML. It introduces a general SD–SEM framework that unifies SD's nonlinear dynamic systems with SEM's latent-variable structure by decomposing models into dynamic, static, and measurement subsystems connected through a common mathematical form. The approach is demonstrated with four examples (two pure SD/SEM models and two hybrids) to show how reference modes, diagram conventions, and tripartite equations can represent both methodologies within a single space. This framework supports systematic generation and comparison of causal models across distributions, facilitating better understanding of model assumptions and reducing risks of unintended algorithmic bias in AI/ML systems. The work lays groundwork for integrating advanced statistical tools into SD and for using SD insights to inform SEM-based inference and AI applications.

Abstract

AI/ML models have rapidly gained prominence as innovations for solving previously unsolved problems and their unintended consequences from amplifying human biases. Advocates for responsible AI/ML have sought ways to draw on the richer causal models of system dynamics to better inform the development of responsible AI/ML. However, a major barrier to advancing this work is the difficulty of bringing together methods rooted in different underlying assumptions (i.e., Dana Meadow's "the unavoidable a priori"). This paper brings system dynamics and structural equation modeling together into a common mathematical framework that can be used to generate systems from distributions, develop methods, and compare results to inform the underlying epistemology of system dynamics for data science and AI/ML applications.

Bridging the Unavoidable A Priori: A Framework for Comparative Causal Modeling

TL;DR

The paper addresses the challenge of reconciling system dynamics (SD) and structural equation modeling (SEM) to enable robust causal modeling for data science and AI/ML. It introduces a general SD–SEM framework that unifies SD's nonlinear dynamic systems with SEM's latent-variable structure by decomposing models into dynamic, static, and measurement subsystems connected through a common mathematical form. The approach is demonstrated with four examples (two pure SD/SEM models and two hybrids) to show how reference modes, diagram conventions, and tripartite equations can represent both methodologies within a single space. This framework supports systematic generation and comparison of causal models across distributions, facilitating better understanding of model assumptions and reducing risks of unintended algorithmic bias in AI/ML systems. The work lays groundwork for integrating advanced statistical tools into SD and for using SD insights to inform SEM-based inference and AI applications.

Abstract

AI/ML models have rapidly gained prominence as innovations for solving previously unsolved problems and their unintended consequences from amplifying human biases. Advocates for responsible AI/ML have sought ways to draw on the richer causal models of system dynamics to better inform the development of responsible AI/ML. However, a major barrier to advancing this work is the difficulty of bringing together methods rooted in different underlying assumptions (i.e., Dana Meadow's "the unavoidable a priori"). This paper brings system dynamics and structural equation modeling together into a common mathematical framework that can be used to generate systems from distributions, develop methods, and compare results to inform the underlying epistemology of system dynamics for data science and AI/ML applications.

Paper Structure

This paper contains 35 sections, 25 equations, 8 figures.

Table of Contents

  1. Introduction
  2. Background
  3. Brief history of structural equation modeling
  4. Efforts to link SD and SEM
  5. General frameworks for computer simulation
  6. Differences between SD and SEM
  7. Goals
  8. Problem type
  9. Diagramming conventions
  10. Equations
  11. SD - systems of nonlinear ordinary differential equations
  12. SEM - systems of linear equations
  13. Implied covariance matrix
  14. Dynamics
  15. Continuity
  16. ...and 20 more sections

Figures (8)

  • Figure 1: System dynamics diagraming conventions of (a) causal loop diagram, (b) stock and flow diagram, and (c) a path diagram from structural equation modeling
  • Figure 2: Observed behavior pattern over time $b(t)$
  • Figure 3: Reference mode $r(t)$ for $New User Adoption$
  • Figure 4: Diagramming conventions for SD-SEM framework
  • Figure 5: nonrecursive relationship between $y_i$ and $y_j$
  • ...and 3 more figures