I wanted to model engagement dynamics using a framework inspired by infectious disease modeling - treating engagement as something that can “spread,” decay, and revive.

I started with a compartmental model with 3 categories:

• New (N): Forecasters who’ve joined but haven’t made forecasts yet.
• Engaged (E): Actively forecasting, updating, or interacting.
• Inactive (I): Made some forecasts, dropped off the platform temporarily or permanently.

Transitions: (add them)

ODE Sim Results

• Basic simulation as a start

Makes sense. No one is actually new at the end of a few time steps. Either everyone is engaged or inactive. And interestingly most are inactive in the long run. Not sure how to validate this but this seems to be true in general. There is a list of users on GJ Open with at least one upvote. Maybe scrape profile data to validate this? (Check TOS).

• Policy Shocks

What will happen if we increase the reactivation rate after introducing the platform to a community? Like introduce a competition (give incentives like money or something…), there are more engaged than inactive people.

• Transient Policy Shock

Early Insights

• The model can explain why admins on RANGE make early/in between forecasts - to “seed” engagement (a ‘reproduction’ threshold must be crossed). This could help newcomers feel less intimidated. This is kind of common sense as well but feels nice that a model validated this.