1.  Simulation helps test ‘what-if’ scenarios that guide real-world design

In the paper “Simulating network intervention strategies: Implications for adoption of behaviour” the team used simulation on both real and artificial networks to compare different seed-selection strategies under simple and complex contagion models. They found that seeds placed at central network positions often achieved faster and more complete adoption than random selection.

This kind of modelling lets us test, in silico, different intervention designs, such as choosing peer leaders vs random peers, or altering the network topology, and to observe how diffusion might unfold under idealised conditions.

2. Seed selection matters, but the network structure and diffusion mechanism also matter

In “Effectiveness variation in simulated school‑based network interventions” the team applied simulations to 17 real school friendship networks, comparing seven different strategies for seeding adoption. They found that selecting the most popular students (nodes with high degree) was generally effective across many diffusion scenarios.

However, their findings also demonstrate that effectiveness varied significantly depending on network structure and whether the contagion process was ‘simple’ (one contact triggers adoption) or ‘complex’ (requires multiple adopted neighbours). For complex contagion especially, some seed strategies failed to trigger any spread at all.

This highlights that it is not enough to pick “popular” individuals, intervention developers must also account for how behaviours propagate (thresholds, peer reinforcement) and the underlying network topology (community structure, path-lengths, diameter, etc).

3. Network structural properties confound intervention performance

In a follow‐up study “Network structure influence on simulated network interventions for behaviour change” the team examined how structural network features (e.g., diameter, connectedness, community clustering) influence simulation outcomes and the relative effectiveness of different seeding strategies. The results showed that structural properties can confound expected results: for example, high-degree seeding did not always outperform random selection, depending on the network. Thus, simulation allows us to explore when and why certain network-intervention methods work or fail which is a critical step for translating into real-world applications.

4. Implications for practice and future research

From this set of simulation studies, we derive a few practical takeaways for interventions using network methods:

1. Whenever possible, map the social network or gather network proxy data to identify potential seed nodes (influencers, connectors).
2. Use simulation modelling (agent-based, diffusion models) in the design phase to compare alternative seeding strategies and estimate reach, speed, and coverage under different assumptions (transmission probability, threshold levels, network structure).
3. Recognise that behaviour adoption often requires peer reinforcement (complex contagion) rather than simple exposure; seeds must be selected and network configured accordingly.
4. Tailor seed selection strategies to network characteristics: in highly structured networks with strong communities, distributing seeds across communities (“community leaders”) may outperform simply picking the highest-degree nodes.
5. Understand limitations: simulations operate under simplifying assumptions (perfect knowledge of network, defined contagion rules, full participation). Real-world interventions introduce messiness (non-participation, missing ties, attrition, context variation).
6. Future work should integrate simulation with empirical implementation (e.g., using pilot data to calibrate models, simulate context‐specific network dynamics, and then design the intervention accordingly).