On (Knowledge) Science vs. Engineering | Victor Charpenay

On (Knowledge) Science vs. Engineering

September 2019

Is there such a thing as 'knowledge science'?

The concept is easy to define as the opposite of 'knowledge engineering' but one question remains: does knowledge science as a concept even make sense? Knowledge engineering is a well-established field but it is somewhat confusing that most knowledge engineers are also researchers. The Semantic Web, as the strongest example of this phenomenon, is mostly maintained, developed and used by the same researchers in a semi-porous environment. Is the big toolbox they have been creating (a.k.a. RDF) the product of research, is it also a theory of knowledge?

The "Sciences of Engineering"

Here is a cheap dialectics to help answer the question: engineers do, scientists observe. The difference lies in the fact that the action of engineers impact our world while scientists always tend to extract themselves from it to seek 'objectivity'. In that respect, what the Semantic Web community is doing—what a wise wording...—is somewhat amgibiguous. On the one hand, datasets as they appear on the Linked Open Data (LOD) cloud come from monitoring and curating information that is already on the Web (Wikipedia, OpenStreetMap, etc.), thus tending towards scientific work. But on the other hand, the Semantic Web community likes to think that the very result of that curation work is actually used by high-tech companies. It seems to be the case for IBM Watson (DBpedia and YAGO are used as sources) and it would hardly be a surprise if Google's Knowledge Graph tapped into Wikidata, since many of Wikidata's claims (the atomic elements of its knowledge base) have been kindly contributed by Freebase, a Google acquisition in 2016. Besides, it is precisely the purpose of these large-scale knowledge repositories to be used by... everyone.

To summarize: building the LOD cloud is a scientific endeavor, it is about observing what engineers publish on the Web; building large-scale knowledge graphs (DBpedia, Wikidata, YAGO) looks more like engineering/doing. Wherever the border science and engineering tends to blur, the French talk about 'sciences de l'ingénieur' or about sciences that consist in observing what engineers are doing. There is no direct translation but one can think of these "sciences of engineering" as whatever the MIT might be interested in. This kind of research has more to do with technologies than with observing nature. As engineers feel observed, they start implementing what researchers formalize and there occurs a shift in our dialectics: engineers do (or reproduce) what scientists observe.

Science as Knowledge

It is not too hard to recognize how mechanical engineering, computing and control systems may relate to science. Scientists simply take engineers' problems as the starting point of their theories, which engineers then turn into new technologies. However, things get a little bit more complex when it comes to knwoledge.

Indeed, the actual purpose of any kind of science is to produce new knowledge. The german word for 'science' ('Wissenschaft') is literally a concatenation of 'knowledge' ('Wissen') and 'whatever-is-derived-from' ('-schaft'). So, what is the specific purpose of a 'knowledge science'? Is it some kind of 'science of science'? That discipline already has a name. It is called 'epistemology'. In fact, what most Semantic Web researcher is doing has more to do with knowledge representation than anything else. That means, while other scientists create their own language to devise theories about the world (or to help create new technologies, as above), other scientists strive to unify them, the same way linguists search for similarities between Swahili and Finnish. (Quite naturally, one ends up with a 'Babel network' to represent knowledge.)

For what it is worth, first-order logic seems to establish itself as the reference language for all sorts of theories. But there is still a long path until engineers actually do anything with that observation.

Note (left as an exercise): given the following Oxford dictionary definition, how would you classify the various activities carried on in the field of artificial intelligence: 'intelligence is the ability to acquire and apply knowledge and skills'?