Class 4

Everything Is Connected: Thinking in Webs, Not Lists

Moving from flat risk registers to causal networks that reveal how workplace hazards interact, amplify, and cascade

The Problem with Lists

Traditional risk management treats hazards as independent items on a register. Each hazard gets a row, a rating, and an owner. This approach has intuitive appeal - it is tidy, auditable, and easy to delegate. But as Lin and colleagues (2024) argue directly, "risk events do not exist in a vacuum: the consequences of one risk may cause others, or the controls for one risk may prevent other risks." When we flatten a dynamic, interconnected workplace into a list of isolated items, we lose the very information we most need to act effectively.

Consider the hospital restructure. A list-based approach might rate "role ambiguity" as moderate risk and assign it to HR for updated position descriptions. Separately, it might rate "low job control" as moderate risk and assign it to operational managers. But what if role ambiguity is causing the drop in job control - because workers who don't understand their role boundaries cannot exercise meaningful autonomy within them? Fixing role ambiguity might simultaneously fix job control, but the list never reveals this connection. Worse, addressing job control independently (say, through a new "flexible scheduling" policy) may fail entirely because the upstream cause - ambiguity - remains untouched.

We need a different representation. We need a web.

---

Introducing the Causal Network

A causal network is a visual and conceptual tool for representing how variables influence one another. Instead of a flat list, imagine a diagram in which each workplace factor - each hazard, mediator, or outcome - appears as a circle, and arrows connect them to show the direction of causal influence. In the language of graph theory, and in the framework formalised by Pearl (2009), these diagrams are called directed graphs.

Three foundational ideas form the vocabulary of these graphs:

• Nodes are the circles. Each node represents a variable in the system - a psychosocial hazard (e.g., high job demands), a mediating factor (e.g., role ambiguity), or an outcome (e.g., psychological distress, absenteeism).
• Edges are the arrows connecting nodes. An edge represents a causal connection - the claim that a change in one variable tends to produce a change in another.
• Directionality is what makes the arrows meaningful. The arrow points from cause to effect. "Role ambiguity → Psychological distress" means ambiguity causally influences distress, not the other way around. Direction matters enormously: reversing an arrow changes what the diagram says about the world.

Schwartz, Gatto, and Campbell (2012) argue that constructing such diagrams allows the "inter-relationships of different factors to be readily visualised and then analysed," offering practical advantages not only for understanding complex systems but for designing interventions. The diagram makes the invisible structure of the workplace visible.

---

Tracing a Ripple: Organisational Change as a Cascade

Let us return to the hospital restructure and trace the causal chain as a network, not a list. Holten and colleagues (2019) conducted a longitudinal study examining precisely this question: what happens to psychosocial working conditions when organisations undergo large-scale change? Their findings map almost perfectly onto a causal network.

Start with a single upstream node: poor organisational change management. When change is managed poorly - communicated late, imposed without consultation, rolled out without adequate training - it creates uncertainty. That uncertainty flows along causal arrows to multiple downstream nodes simultaneously. Holten et al. (2019) found that organisational change was associated with increased job demands, decreased job control, increased role conflict and role ambiguity, and decreased social support.

Now trace the ripples further. Role ambiguity doesn't just sit there as a static hazard. Inoue and colleagues (2021), in a large longitudinal study of nearly 14,000 Japanese workers, demonstrated that role ambiguity acts as an amplifier - it not only causes psychological distress directly but also magnifies the harmful effects of other stressors like high demands and low control. In network terms, role ambiguity is not just a node with an outgoing arrow to distress; it is a node that strengthens the arrows running from other hazards to distress.

Follow the chain to its end: psychological distress drives absenteeism and presenteeism, which increase workload on remaining staff, which further degrades job control and supervisor capacity, which deepens distress. Van Dongen and colleagues (2022) mapped exactly these kinds of feedback dynamics using causal loop diagrams built with HR professionals, managers, and employees across twelve organisations. Their work confirms what the hospital OHS manager intuited: workplace well-being is a system of reinforcing loops, not a collection of independent symptoms.

The flat list sees six problems. The network sees one cascade with a single origin point.

---

Conditional Relationships: When Context Changes Everything

Not all causal arrows are equally strong in all circumstances. Sometimes the influence of one node on another depends on the state of a third node. This is the idea of a conditional relationship, and it is one of the most powerful insights that network thinking provides.

The classic example - one you encountered in Chapter 3 - is Karasek's (1979) demand-control model. Karasek proposed that high job demands do not, by themselves, inevitably produce psychological strain. The critical question is: how much decision latitude (job control) does the worker have? When control is high, demands can be energising - they become challenges rather than threats. When control is low, the same demands become overwhelming. The causal arrow from demands to strain is conditional on the state of the control node.

Van der Doef and Maes (1999) reviewed twenty years of empirical research testing this model and distinguished two versions of it. The "strain hypothesis" treats demands and control as having independent, additive effects on well-being. The "buffer hypothesis" - which is the conditional relationship version - proposes that control moderates the effect of demands. Their review also introduced a third node: social support. When demands are high, control is low, and support is absent, workers experience what the literature calls iso-strain - the most toxic combination in the model. This is a conditional relationship involving three nodes simultaneously.

In a list, "high demands," "low control," and "low support" are three separate hazards rated independently. In a network, they form a conditional structure where the real danger emerges only from their combination - a pattern invisible to any risk register.

In the formal language of causal graphs, Pearl (2009) provides tools for reasoning about exactly these kinds of dependencies - how information and influence flow (or are blocked) depending on the states of other nodes. We will not touch the mathematics until Chapter 5, but the intuition is already here: context modulates causation.

---

Leverage Points: Why Webs Are More Actionable

If the network view were only more accurate, it might still be a luxury - interesting for researchers but impractical for busy managers. But the network view is not just more accurate. It is more actionable, because it reveals leverage points - nodes where a single intervention can cascade benefits through the entire system.

Lin and colleagues (2024) formalised this idea by applying network topology analysis to workplace risk registers. By measuring properties like centrality (how connected a node is) and betweenness (how many causal pathways flow through a node), they identified which risks functioned as structural hubs in the network. These hubs are leverage points: intervening on a highly central node disrupts more causal pathways than intervening on a peripheral one, even if the peripheral hazard has a higher individual risk rating.

Return one final time to the hospital. On a flat list, "poor change management" might be rated as moderate risk - it's an organisational process failure, not an immediate psychosocial hazard. But in the network, it is the single upstream node from which five downstream hazards cascade. It is the highest-leverage intervention point in the system. Improving change management - through earlier communication, genuine consultation, and staged implementation - doesn't just address one line item. It loosens the causal pressure on role ambiguity, job control, demands, support, and ultimately on distress and absenteeism simultaneously.

This is the fundamental promise of network thinking: it shows you where to push so that the whole web moves.

---

Build Your Own Web

The best way to internalise network thinking is to practise it. In the interactive exercise below, you will encounter six common psychosocial variables and be asked to draw the causal arrows between them. There are no trick questions - just apply the logic you have built throughout this chapter. Which factors cause which? In what direction do the arrows flow? After you submit your network, you'll be able to compare it against an expert-validated structure and see where your intuitions align - and where the connections you missed reveal something new about how these hazards interact.

---