by Olli Parviainen
Do you know how ants can solve complex path finding operations when trying to bring food to the hive? The ants leave a trail of pheromones that other ants can detect. This pheromone trail slowly evaporates unless it is reinforced by another ant. A single ant has little intelligence, it relies instead on few single guidelines: “bring food to the hive”, “follow the strongest pheromone trail” and finally “if no pheromone trail present, walk randomly”. These guidelines aggregated through couple of thousand ants delivers results that even modern computers struggle to achieve: they find the shortest path between two points.
What does this have to with the extended mind and semantics? If we view our written or spoken sentences as strings of words, we can form a network from them (below is the network structure of this post). Unless you don’t have couple of thousand ants at hand, network analysis is one of the methods used to calculate the shortest paths between two concepts. In our everyday conversations certain words or concepts co-occur more often than others, e.g. “cucumber” and “sandwich” co-occurs more often than “sandwich” and “concrete”. Some words or concepts have more in common: we can map this semantic network. Using certain words together is a bit similar to ants finding food: you don’t know where the concept is, all you have to is follow the trail of others.