Policy and Complex Systems
Possible Explanations
Finally , we want to provide some discussion of the potential reasons for the patterns that were demonstrated in the simulation results . Under the no information treatment , the adoption of the technology is largely negatively related to the size of the farm . A possible explanation for this observation is that since the cost of adopting the technology is proportionally related to the size of the farm , participants may follow some heuristic decision rules that attribute significant weight to the cost of adopting in the processes . This clearly demonstrates that as opposed to always following profit maximizing decision rules , human behavior is often limited in their calculating ability and may be affected by various cognitive reasons and therefore demonstrate bounded rationality in terms of forming some rather heuristic decision rules . Furthermore , both information treatments seem to provide anchors for the participants . Knowing what people like them have done in the past and what others in their group have done provide people with a reference point in their decision process . Since individual level information provides people with tailored information , it helps people make better decisions compared to the myopic baseline case . Under the group level information where a group average is provided , we can observe that the absolute adoption and production decisions for farms with different sizes tend to be very close . This suggests that people might be anchored to the group level averages , or peer actions , even though it might not be in their best interest to do so . an ABM should come from complex agent interactions and adaptions in a concise model rather than from complex assumptions about individual behavior and free parameters ( Axelrod , 1997 ). Most of the parameters that influence the observed results in the ABM are calibrated and validated based on experimental data . Therefore , the uncertainty only results from realization of the randomness in each simulation experiment , which is stochastic in nature and should not generate any systematic biases . Meanwhile , if an uncertain variable affects each scenario of the simulation in an equal magnitude , the relative comparisons between the scenarios will not be affected . Therefore , one uncertain parameter that would possibly affect the result is how many farms the participants consider part of their group . This parameter affects the grouping structure and the group level information that is shown to the participants . In our baseline scenario presented before , we assume five people are considered to be in one group . We increase this parameter to 10 , 15 , and 20 in this part and the result is shown in Figure 12 .
As shown in Figure 12 , as the number of people that the participants consider themselves to be in the same group with increase , the deviation from the target pollution level and the simulated pollution level is not largely affected under individual level information treatment , but increases under the group level information treatment . This suggests that individual level information not only generates highest policy efficiency , but also is more robust to participant perceptions on their group size .
Sensitivity Analysis
In this part , we discuss how our results would be affected by uncertain parameters in our ABM . Ideally , the result of
183