UID Everything!

I often joke that we should assign unique identifiers (UID) to everything. The important points in my articles are labeled with UIDs, not numbered 1.2.3 like theorems in a paper. This is not a fetish; it has an epistemic reason. Here is a very simple example. Someone can be known as A and B on different occasions; for many tasks, this name is not a correct identifier. For some tasks, even a Social Security Number cannot be a correct identifier. But we can still find a coherent description of the situation, viz. “A, B, and the person holding the Social Security Number 987-65-4321 are the same biological person (whatever ‘biological’ means)”. At this point we have shifted the senses to track the same object in a more desirable way. But as long as the identifier goes, the best we can do is to assign a UID to that object in each small interaction session as soon as we can conceptualize and identify it, attach a label in natural language, then attach a proof or belief that these objects are identical in some sense across the sessions. Fortunately in many cases we can afford to weaken this standard and automate a bit. Interestingly, when I attempted to assign UIDs to my friends they appeared to not understand the motivation and actively disagreed with the practice. There seems to be an emotional response to the technicality of this practice: I can feel once significant objects being apparently demoted to UIDs, stripped away of any metaphysical significance. But such an emotional response is premature.

In the way of UID, the status of the information captured is different from that of natural language text. With natural language we tend to believe that a text tells something concrete and ready to be understood. Contrarily, the information corresponding to the UID is a sense evoked by its natural language labeling; it can only be truly recovered by those sufficiently engaged with the subject. This is a fundamental setting: no sense is by default widely recognized, but whenever it is used it must have been understood very well. Alternatively speaking, this permits labeling to be levitated from literal natural language when needed.

Extending the idea, I am going to write as if these senses in human thinking are reflected in computer data in a rather direct way. Any linguistic objects we consider identical, i.e. cannot further distinguish as different senses, are considered the same object. I might call this system “pseudo-formal”. The objective here is not to perform computation on this pseudo-formal data, but manage, distribute, and facilitate intelligent senses in an effective way. Only after these senses become formal enough may a computer process them symbolically.

Thesis 51879e1b. The initial motivation of UID as an accurate tracker of sense maintained by human users. This thesis.

Part of Project 262c86ec. This idea is the start of an approach to information where humans maintain structured senses “pseudo-formally”, in a format that is more structural than natural language yet represent informal senses.