Conceptual Background

Organizational routines are an extremely complex phenomenon that involves individual psychology, physical tools, social norms, formal rules and a host of other factors. In the action network simulator, we have created an abstract model that steps back from those details to capture the core idea of routines as recognizable, repetitive patterns of action.

Here is how the model works. For help on the simulation controls, click here.

  • The model is initialized with a set of random sequences for either a "formed routine" or "clean slate". The sequences are on the right, and a matrix that summarizes these sequences as a network of action is on the left. We call that the history matrix.
  • When the model runs, it generates new sequences based on the probabilities in the history matrix. This models the path dependence within the routine, as one action influences the probability of the next. This is what makes the pattern of action recognizable.
  • Variation. In real routines, variations can occur due to error, exceptions, improvisation, and so on. In the model, you can control these variations by setting V, the probability that each action in the sequence may be changed. Variation is what allows the routine to drift.
  • Retention. At each iteration, the transition matrix is recalculated based on the most recent R sequences. This simulates path dependence between iterations. This is what makes the pattern of action repetitive. When R is low, the routine can change very fast (low inertia); when R is high, it changes very slowly (high inertia).

If you work with the simulation, you will see that the network of action will become more or less complex depending on the settings of variation and retention.