The generation of alternative policies is essential in complex decision tasks with multiple interests and stakeholders. Today such settings are common in the mitigation and management of environmental impacts by governments and industries. A diverse set of policies is typically desirable to cover the range of options and objectives. Decision modelling literature has often assumed that clearly defined decision alternatives are readily available. This is not a realistic assumption in practice. We present a structured process model for the generation of policy alternatives in settings that include non-quantifiable elements and where portfolio optimization approaches are not applicable. Behavioral issues and path dependence as well as heuristics and biases which can occur during the process are discussed. The experiment with the climate change mitigation game compares the results obtained by using two different generation techniques. The results show that the outcome can be process dependent and that cognitive biases can occur and that the use of heuristics is common. our conclusion is that modelling support in policy problems needs to be combined with processes for the generation of alternatives paying attention to the related behavioral effects.