The COWCULT model is built using System Dynamics in a commercial software programme (STELLA, v.7). The major available data sets used by COWCULT are:
Weekly record of number of BSE cases (1988 - 2004)
Yearly record of deaths from nvCJD (1988 - 2004)
Yearly spending on BSE and nvCJD Research (1988 - 2004)
Six monthly frequency of media articles on BSE/nvCJD (1988-1995)
Weekly ftequency of media articles on BSElnvCJD (1996-2004)
The 'same' number of BSE/nvCID cases are not translated into the same quantity of
risk perception by the different Cultural Theory Archetypes, however. Egalitarians will accumulate a much higher level of risk perception for a given number of events than is the case for an Individualist, with the Hierarchists typically falling somewhere in between. The Fatalists will tend to have a more unpredictable and discontinuous response. Following through the logic of each Cultural Theory Archetype (CT A), we allow the same events to be differently constructed by each CTA for each of seven variables that we have identified in the sociological and psychometric literature on risk perceptions: involuntariness, pollution of nature, unfamiliarity, dreadness, trusthworthiness, vulnerability and fairness. The output is a "risk perception" level for each CTA.
We then use the Theory of Surprise (a branch of Cultural Theory) to allow for a shift in the population between the CT As. Each CTA has an expectation of the risk which is provided exogenously (on the basis of our judgement). The extent to which expectations are thwarted by what 'actually' happens provides a measure of surprise. The greater the surprise, the more integration occurs from that CTA to another one. Hence, if the Individualists expected that BSE cases would fall away rapidly in the early 1990s, their surprise value would be high and adherents to Individualism would shift to one of the CTAs that did not experience any (or as much) surprise, e.g. Egalitarianism or Fatalism.
The resulting "risk contributions" take account of the relative size in the population of each CTA, i.e. if the number of egalitarians increases from 25% to 50% then the egalitarian CT A will double in calculating risk contributions.
We now briefly describe the 'risk amplification/dampening' components of the model, i.e. other processes which increase or decrease the risk perceptions held by each CTA. The media is a major source of amplification, though we define the pattern of media amplification differently for each CTA. Weekly data of media coverage from 1996 to 2004 is classified according to whether it is likely to be amplifying or dampening. The record of significant ‘events' such as scientific and governmental announcements is a further amplifier. The amplification produces 'amplified risk contributions'.
Finally, we introduce a Trust-Reliance sub-model, which includes the effect of trust or distrust in government by each of the CTAs and the extent to which different CT As are reassured or not by government efforts to reduce adverse risk perceptions. The submodel is driven by the extent of government R&D (which we assume to increase the reliance on science in BSE/nvCJD decision-making), by the individual CTA and collective or 'average' risk contributions. The trust-reliance model generates "modified CTA risk contributions". These are modifications of "amplified risk contributions" taking into account: the trust of each CTA in reliance on science and technology by the government in reducing risks, and each individual CTA's own risk contributions. This sub-model additionally allows us to change the exogenous level of trust in the government due to, e.g. a change of approach or a PR campaign, an unanticipated mistake, participatory decision-making, and so on.
We have undertaken some preliminary continuation of COWCULT using data on: public perceptions of BSE (1990-1998) and sales and prices of beef (both 1985 -1999).
There are some areas of correspondence though the data sets themselves are of questionable validity in attempting to confirm the model's outputs. We have also undertaken some testing and experimentation with. COWCULT to explore where some of the key sensitivities lie, and to begin to utilize the model to ask 'what if' questions.
We believe that COWCULT does succeed in simulating theoretically-oriented understandings of different risk perceptions, though much more testing is required to be confident that the simulation is occurring for the 'right' reasons. Throughout COWCULT we have provided speculative relationships between variables based upon our own judgement of the way that risks would be interpreted according to each CTA. As such the model is a 'thought experiment' and not calibrated or validated. In order to achieve better calibration it would be necessary for an extensive data collection exercise to be undertaken though this could be problematic due to the unavailability of relevant data from the past.
We hope to have contributed to simulation and complexity model by focusing upon beliefs and perceptions and by using CT (our first aim). We have provided a new methodological tool for STS which could be further developed in future research (our second aim). Finally, we have developed what can be turned into a user-friendly tool though we did not have sufficient resources to explore stakeholder and user issues in detail (our third aim), in part due to a budget cut. The COWCULT model, if it has any validity, shows the importance to government of focusing and customising its efforts at risk engagement and communication upon particular culturally-defined groups.
