A caveat is to be offered at this time that relates directly to the matter at hand: the psychology of nothingness. While it is certainly of benefit to identify and study the paradigms and models that underly our daily practices, it is also important to remind ourselves that these paradigms and models don’t really exist. We can’t touch them or measure them. We can only infer that they exist by observing the patterns to be found in the practices we engage and the words we use to describe these practices. It is certainly “fair game” to declare that paradigms don’t really exist, but also only “figments” of Thomas Kuhn’s historical imagination. One can even declare that the notion of paradigms (and models) is itself just an arbitrary “paradigmatic” fiction that might be convenient but certainly isn’t scientifically verifiable as a “real” entity. Put simply, paradigms (along with models and perhaps even practices) are actually nothing.
Semantic Patterning
A second framework has recently been identified. This concerns the way in which words tend to cluster together. Known as semantic patterning or semantic signaling, this framework concerns the constructs formed by our often-unacknowledged interconnecting of words in our speech and writing. We identify and separate meaningful patterns and structures of words (and other forms of data), map these patterns and structures into something called semantic spaces that help us uncover latent topics and agenda. Furthermore, we find that individual people and clusters of people (for example in a family, tribe or organization) create and engage their own unique cluster of words (called semantic signatures).
I find that this second framework is of great value when reviewing the text of a meeting that has been transcribed (which is now automatic with many digital meeting devices). Powerful tools of analysis now allow us to determine not only frequency in the use of certain words, but also the relationship that exists between the use of these words. Clustering Methodologies are concerned with the identification of patterns and relationships among words and (more broadly) concepts.
These methods range from Pattern-Based Clustering which is least accurate to Search Engine Results Pages (SERPs) which are the most accurate. `The simplest methods focus on shared words or text patterns, while more complex methods move beyond word-matching to the consideration of meaning and context. They group keywords based on semantic similarity rather than just shared words. Even higher order, AI/LLM-based clustering use large language models to group keywords by semantic similarity rather than search engine behavior. Tools like ChatGPT, Answer Socrates, Writersonic, and Search Atlas fall into this category.
Finally, we consider SERP-Based Clustering, which is the gold standard, in that it focuses on the point where “the rubber hits the road”. Namely, it focuses on how specific key words produce pages on Google. These tools analyze Google’s actual search results (SERP Overlap) to group keywords. If two keywords show similar top-ranking pages, they will be grouped in the same cluster. When two keywords show similar websites ranking in Google’s top 10 results, they have “SERP overlap.” High overlap (70%+ shared URLs) means keywords can likely be targeted on the same page. Low overlap means they need separate pages.
