Over the past set of posts I have been discussing how implicit semantics can affect one’s interpretation of data. We looked at a literary example and tried to deconstruct the meaning. But when we adapted the inferred hierarchy to a different context, we introduced meaning into the model without accounting for it.

That suggests two things: first, our perception of what the original model represented and meant was skewed by our own interpretation. Second, we augmented that interpretation when adapting it to a different context, potentially limiting that description somewhat.

But there is a deeper implication as well: meaning is somewhat disconnected from intent. Those tasked with creating data sets for publication and sharing are by their own devices imbuing the data with biases based on their knowledge of the context in which the mode is created. And those using the data are imbuing the data with biases based on what they expect to see in the data. This realization suggests that there will always be some inconsistencies in interpretation that will become more acute as the distance between data creator and data consumer widens.