Proceedings of the Annual Meeting of the Cognitive Science Society

The key insight in this work is that the events people associate with risk is under systematic change over the past200 years. Leveraging Latent Dirichlet Allocation (Topics Modeling) and the Google Ngram Corpus, we identified historicaland newly-emerging events associated with risk and tracked their relevance over time. We also computed the probability ofrisk co-occurring with words associated with those identified events to capture a more accurate trend. Several highlights ofthe findings include: attention on risk has been spreading from one general domain (about losing life and war/battle) to a setof wider, more specific events and activities such as cancer, sex, HIV, smoke, and finance; in addition, the concept of risk hasrecently become more differentiated, incorporating both cost and benefits, long and short-term consequences. This approachcould be extended to study semantic history of a number of other concepts of interest.