ConceptNet is a semantic network for common sense reasoning in AI systems, created by researchers at the MIT Media Lab. It’s a good idea, but as with all attempts at this sort of thing that I’ve seen, trying to collapse common sense knowledge into a simple many-to-many network doesn’t always work great. Especially when you start accepting contributions from the general internet population.

A paper written by the creators exists:

… To compare the correctness of the predicted statements, we translated the responses into an integer scale, giving a score of 0 for “doesn’t make sense,” 1 for “not true but amusing,” 2 for “not true,” 3 for “don’t know” or “opinion,” 4 for “sometimes true,” and 5 for “generally true.” …


Many tools have been developed to give computers better ways to understand people – via analysis of facial expressions, written language, speech, and internet activity – allowing for the prediction of intent and future action. How do the resulting digital ontologies and software representation express, mutate, or influence qualities of human experience? Is there a tension between the hypothetical outcomes of these tools and their practical, quotidian applications? Semantic networks are one such tool. This installation co-opts the topology of a semantic network called the ConceptNet to act as a metaphor for the particular outcome of commodification – via an aggressive, real-time, recursive depth-first graph search of the API.