Risk Domain and Perceived Similarity
Perceptual Mapping
I collected the data via an online random sample. Respondents rated the similarity they perceived between pairs of risky activities. One risk (the target risk) was evaluated relative to another risk (the referent risk). Target and referent risks could either be from the same domain or different domains (there are five possible risk domains that have been identified in previous academic research: financial, health/safety, ethical, social, and recreational).
Overall, the results demonstrate that individuals see varying degrees of similarity between risks and that this similarity largely coincides with established risk domains. To visualize the similarity-domain correlation, a perceptual map was created via Multidimensional Scaling (MDS).
General Results
Respondents from a random online sample were asked to rate similarity between risky behaviors—a measurement that has not been mapped onto domain definitions before—thereby showing that behavior and similarity judgments vary by risk domain.
Respondents were shown twenty risky activities (four from each of five domains) and were asked to rate similarity on a scale from zero (“Completely Different”) to 100 (“Identical—No Difference”) for all possible pairs of activities within the set. No information about the risk domain was conveyed to the respondents.
To understand how perceived similarity relates to domain, I compared similarity ratings within and across each domain. The average similarity ratings are significantly higher for comparisons within the same domain than for comparisons across different domains. This pattern of results holds for all five risk domains. When making an overall comparison of similarity ratings between same-domain risks and different-domain risks, I find that generally, when activities are from the same domain, they are seen as more similar to each other than to activities from different domains (MSame = 40.52 vs. MDifferent = 18.60, t(142.98) = 8.70, p < 0.001).
Perceptual Map
I originally created the perceptual map in R, but it has been reformatted below for ease of exposition.
​
To create the perceptual map, I first defined the plot I wanted to create from the MDS (Multidimensional Scaling) object. After loading the data, I made the lower triangular matrix of average differences. Then, I created a symmetric differences matrix. After this, I found the MDS solution, which was mapped using the MDSPlot function.
​
The map below shows individual risky behaviors by risk domain. For example, financial risks are all identified in light blue. Each light blue point below represents a different risky financial behavior (e.g., betting on a sporting event or gambling at a casino). Points that are closer together are perceived as more similar on average. While there isn't a perfect 1:1 mapping between domain and similarity, most behaviors from a given domain cluster together in the perceptual map.
