The model investigates how perception, habit, and social influence shape individual decisions about milk type. It reproduces adoption behaviors of milk consumption by the British public from 1974 to 2005, specifically the substitution of whole milk for skimmed and semi-skimmed varieties. Agents (n = 1,000) represent adult consumers placed randomly in an information environment and embedded in a directed social network. Each agent has a disposition to consider changing milk type under either a threshold-based or a probability-based mechanism. The simulation tracks average weekly consumption per person in milliliters of whole and skimmed/semi-skimmed milk, along with internal variables including habit factors, choice function values, and health and environmental value deviations. The parameter set used is: memory-lifetime = 5, alt-health-mean-initial = 1.69, habit-threshold = 2, p-interact = 0.3, social-susceptibility = 0.2, incumbent-initial-habit = 10, alt-env-mean-initial = 1.12, social-blindness = 0.51, justification = 0.61, cognitive-dissonance-threshold = 0.15, network-parameter = 8, number-of-agents = 1000, habit-on? = True, network-type = "watts-strogatz", norms = True, and social-conformity = 0.28.
In the early phase (steps 0–10), average weekly whole milk consumption starts at approximately 2,650–2,700 ml per person and declines to about 1,000 ml by step 10. Average weekly skimmed/semi-skimmed milk consumption starts near 0 ml and rises to about 1,200 ml by step 8. The mean habit factor for whole milk begins at 10, matching the incumbent-initial-habit parameter. The mean habit factor for skimmed/semi-skimmed milk starts near zero. During this period, the mean habit factor for whole milk decreases slightly and the mean habit factor for skimmed/semi-skimmed milk increases from near zero. Choice function deviations and health and environmental value deviations remain small. However, they do begin moving away from their initial levels.
In the middle phase (steps 10–26), skimmed/semi-skimmed milk consumption peaks around 1,250–1,300 ml between steps 10 and 12, then fluctuates slightly between approximately 1,200 and 1,250 ml through step 26. Whole milk consumption continues to decline after step 10, falling from about 1,000 ml to roughly 600 ml by step 26. Total skimmed/semi-skimmed milk choices increase from near 0 at step 0 to about 16,000 by step 26. Total whole milk choices follow a linear trend across runs. They rise from 0 to roughly 1,300 by step 26. Total choices made increase linearly from 0 at step 0 to about 20,000 by step 26. A crossover in average consumption occurs around step 18 (1992) in the empirical data, around step 14 (1988) in the probability-based variant, and around step 22 (1996) in the threshold-based variant. The mean habit factor for skimmed/semi-skimmed milk rises and exceeds the whole milk habit factor during this period. The mean choice function deviation and both health and environmental value deviations reach higher values than in the early phase.
In the later phase (steps 26–32), whole milk consumption drops more sharply from about 600 ml at step 26, with a brief pause and small bump around steps 26–27. They stabilize at roughly 250–300 ml by step 32. Skimmed/semi-skimmed milk consumption rises again after the mid-run fluctuations, climbing from about 1,250 ml at step 26 to approximately 1,450 ml by step 32. Total skimmed/semi-skimmed milk choices reach about 26,000 by step 32. Total whole milk choices reach roughly 2,200 by step 32. Skimmed/semi-skimmed choices remain dominant through the end of the run. Total choices made reach approximately 32,000 by step 32. Mean habit factors for both milk types converge to stable values. The mean choice function deviation levels off. Both health and environmental value deviations stabilize.