Causal Loop Diagrams (CLDs) document and visualize the dynamics of complex systems, describing their relevant factors, called variables, and causal relationships between them. The most interesting features for exploring CLDs arecausal loops, i.e., circuits of causal relationships, which can be characterized as balancing or reinforcing; other interesting features of CLDs caaruesal routes, i.e., chains of relationships connecting any two pairs of nodes, recognized as increasing or decreasing. We hereby introCdLuDc-eExplorer, a prototype interactive system for inputting new CLD diagrams and analysing their features, reviewing their causal loops and routes. The system supports simple interfaces for systematically extracting and exploring loops and routes, which can be used by systemic designers in identifying the ”areas of intervention” within a complex system, empowering organized reasoning on CLDs and improving the related decision-making processes.
Systemic Design is driven by the ambitious objective of addressing complex problems in terms of
“relationship and global dynamics”, rather than isolated components. An important instrument of
systemic design is the description of complex systems’ dynamics by means of Causal Loop Diagrams
(CLDs, [
In a companion paper2[], we showed that CLDs’ analysis takes advantage of exploring fceaedubsaaclk loops, i.e., cyclic paths looping from one variable back to the same variabclaeu,salnrdoutes, consisting of paths of edges connecting two nodes, denoted as source and destination. Loops are characterized as balancing (B) orreinforcing (R) - simply by counting edges with negative polarity along a loop, and denoting a loop as balancing if the count is odd, reinforcing if the count is even. Similarly, causal routes are associated wi tinhcreasing anddecreasing polarity - again by counting edges with negative polarity along a route, and denoting a route as increasing when the count is even, decreasing when the count is odd.
CEUR Workshop
ISSN1613-0073
We designed and implemented a prototypeCaoufsal Loop Diagram Explorer, a web-based application designed for the interactive construction and analysis of CLDs1. Fililguusrterates the application’s workflow diagram and sketches some of the pages of the web-based user interface.
The workflow consists of four phases, introduced by a welcome page, which can be progressively
executed. Each phase is associated with a Web page; pages can be accessed from the sidebar of the
interface. In the welcome page, a brief guide with all the main elements and characteristics of CLDs is
presented to the user. TDhesign your CLD page (Figure1A) embeds LOOPY [
After adding and editing all elements, it is necessary to export a LOOPY diagram and save it for
reuse. The loading mechanism verifies that the diagram is a valid CLD and uploads the CLD to a
Neo4j [
In theExplore Loops (Figure1B-C) andExplore Routes pages (not shown), users can browse, select, and visualize these CLD features using interactive forms. To help search, pages initially present statistics about the number of loops/routes existing in the CLD; loops are divided into balancing and reinforcing. The page also includes the list of all nodes of the CLD, and the total number of loops/routes in which the node is involved (see Figu1rBe). Users can search for specific loops/routes by selecting, within a mask, the starting node, the polarity (balancing or reinforcing), the length (1seCe),Faingudrcehoose to save them for inclusion in a final report. Loops/routes can also be paired for direct comparison.
At the end of the explorative process, users can find an exportable report of their activities in the View Report and Export page, which ofers a visual summary of the features they have discovered in the process. Each report includes all the loops/routes that were saved during the current design session.
The system combines a typical stack of web technologies with the N4]epor4oj p[erty graph database, to automatically extract quantitative characteristics of CLDs. The user interface is implemented with Streamlit6][, a low-code Python web framework, and proposes a process organized in pages for constructing and exploring the CLDs, as shown in Fi1g.ure
According to the metamodel discussed2]i,nt[he Neo4j graph database representsVtahreiables andCausalityLinks; each causality link is directed from a ”tail” variable to a ”head” variable, hence the Neo4j graph consists of uniquely identified nodes (with their names) and directed edges (with their polarity). We then use Cypher quer7ie]st[o extract aFleledbackLoops (identified as a chain of nodes connecting an initial variable back to itself)CaanudsaallRloutes (as a chain connecting any two nodes); the output of the query is, in both cases, a list of intermediate node identifiers and binary attributes indicating either balancing/reinforcing for loops or increasing/decreasing for routes. This operation is almost instantaneous for CLDs of reasonable sizes, up to hundreds of nodes and thousands of edges.
The extracted characteristics are then uploaded to an instance o5f]MtyoSQquLic[kly enable the following exploratory pages of the analysis without recomputing them.
Limitations. The current version of the tool presents several limitations. First, integration with LOOPY is not seamless, as users must manually export and paste the diagram link to begin exploration. Additionally, starting a new session or exploring a diferent CLD requires the user to refresh or reopen the page, which disrupts the user-interaction workflow. Finally, while users can explore built-in diagrams, previewing custom CLDs before loading them into the system is not yet supported.
The tool includes three predefined examples, whose CLDs can be loaded at any time. We consider the COVID-19 and the Fashion use cases presented2i]n, a[nd the complete CLD designed i8n],[describing the adoption of renewable energy technology in Australian hotels, the largest, most documented CLD that we found in the literature. For these predefined cases,1Traebploerts the number Voafriables, CausalityLinks, FeedbackLoops, andCausalRoutes.
Use cases Variables CausalityLinks (a) (b) (c)
The CLD-Explorer can be used to support a case-based analysis for extracting quantitative insights that complement the visual modeling of CLDs. This proved particularly valuable in making explicit the structural properties of causal models that are often left implicit in qualitative practice. For example, in the case study on the environmental footprint of the fashion industry, the CLD was constructed with the aim of exploring the feedback structures surroundinfagshtihoen industry footprint variable, which was found to participate in 359 feedback loops and initiate 875 distinct causal pathways. Interestingly, the variablepurchases of new clothes appeared in an even greater number of loops—411 in total—underscoring its pervasive involvement across the system and highlighting the centrality of consumer behavior in shaping sustainability dynamics within the sector.
Figure2 shows a comparison of alternative routes betpwuercehnases of new clothes andconsumers’ desire to buy (i.e., with reversed polarity). The first is increasing, simply going thrcoluotghhing production andmarket competitiveness. The second, instead, is decreasing and takes a wider perspective, as the purchase of new clothes leads to increase of owned clothes and washes, resulting into chemical pollution and industry footprint; however –with some delay– this causes a raise in consumers awareness and consequent investments in sustainability, thereby decreasing (note the dashed arrow) production intensity. Here, profits diminish, decreasing marketing actions and consumers’ desire to buy. The tool allows for placing pairs of loops/routes side by side (in carts) for visual comparison.
The application also enabled the comparison of structural positions of diferent actors. A clear asymmetry was observed betweeconnsumer behavior andgovernment policies; while the former is connected to the footprint through several short/direct paths, the latter — despite being involved in many more distinct routes — tends to influence the system via longer chains in≥vo8livnintegrmediate variables. This suggests a more mediated and potentially less immediate systemic efect of governmental action when compared to consumer-driven dynamics.
A parallel can be drawn between the CLDs formalism a ni*dftrhaemework9[
Preliminary workshops with expert systemic designers have been used to understand their needs for analytical support in their daily practice; our prototype supports their main requirements related to searching all loops and routes and characterizing their properties. We plan to extend the CLD-Explorer platform by turning it into a more user-friendly tool for supporting the design process, i.e,. various versions of the same CLD, thereby modeling the same system across diferent design stages and allowing the comparison of successive or alternative versions. Classical queries comparing versions will be used to monitor how models evolve—e.g., how the introduction of new variables or links leads to new feedback loops or reveals causal routes. The platform will support robustness analysis, help reveal structural vulnerabilities and redundancies, and foster interpretability by tightly linking visu representation and systemic logic. The design process will also be supported by a new UI for inputting CLDs.
The authors have not employed any Generative AI tools.