Description
Writing a verbal theory of behaviour as a few differential equations, then numerically finding its stable and unstable equilibria and how they appear, disappear or switch as a parameter changes.
Many behaviour change theories describe feedback loops in words: anxiety raises arousal which raises perceived threat, use builds skill which builds use. The book shows how to turn such theories into small systems of differential equations, simulate them, and analyse them rather than only run them. Phase-plane plots show the direction of change for two variables at once and locate fixed points; a Newton search finds equilibria from different starting states, and parameter continuation traces how each equilibrium moves as a parameter varies, producing a bifurcation diagram that reveals bistability, hysteresis and tipping points. Adding stochastic noise shows spontaneous jumps between states. The book separates dynamic noise, which changes the trajectory, from measurement error, which does not, and warns that fitting such models to data is sensitive to starting values and model choices; seemingly small changes in a term can transform the qualitative behaviour. For UX, the main value is formalising a theory of a habit loop, engagement spiral or abandonment so that its assumptions, time scales and tipping points become explicit and testable, even before data are collected.
Attributes
- Discipline
- Systems & Complexity
- Loss type
- Siloed
- Data
- Analytical / expert
- Effort
- Weeks
Steps
- State the phenomena to be explained and the verbal mechanism, naming the variables and the feedback loops between them.
- Write each loop as a differential (or difference) equation term, choosing functional forms such as linear, logistic or saturating and justifying each.
- Simulate the model from several starting states, with and without noise, and draw phase-plane plots for pairs of variables.
- Find equilibria and classify them as stable or unstable, then continue them across a key parameter to draw a bifurcation diagram.
- Identify tipping points, bistable regions and hysteresis, and derive qualitative predictions that data could falsify.
- Where time-series data exist, fit parameters and check sensitivity to start values and to the distinction between dynamic and measurement noise.
Sources
- van der Maas, H., Complex Systems Research in Psychology (2024)
Also known as
- Dynamical systems modelling
- Numerical bifurcation analysis
- Parameter continuation
- Phase portrait analysis
Categories
- Strategy & Collaboration
- Systems & Complexity