Too Many Whats? Part I. An Investigation Of Problem-Like Concepts

There's a method of Concept Analysis I sometimes use when I have a big list of words (like an English spelling dictionary) and I want to sort it into narrower clusters: I come up with a sentence that has a blank spot which can be filled by some of the words on the list (maybe two or three select words to start the process) but not some of the others, and I use the sentence as a binary classifier. "Does this sentence scan naturally when I let its variable equal this value"? Often it works better to start with a few blanked-sentences which all fit well with the seed words: a cluster of tests to point your attention toward a cluster of concepts.

Here's an example: What words can stand in for X in all of these phrases?:

• I'm facing too many X at once.
• I've been dealing with X for too long without a break.
• I perceive the presence of important X too readily.
• I find myself systematically dealing with the wrong X.

The word "problems" fits in each sentence well, while "loaves of bread" makes for some really weird sentences. If you'd like to try your hand at finding other words or phrases that fit, before I share mine down below, now's the time.

It could be fun. .... All right.

I don't know a name for this technique. I've seen it operationalized in the lexicosemantics literature where, for example, a word is judged to have passed the test if the filled-in sentence returns a threshold number of search results when used as a Google search query. Maybe the lexicosemanticians have a name for it and I've just forgotten. It's a little bit like type checking, isn't it? "Error, line 1: "loaves of bread". Problem-Like entity expected. Received Unvalenced Material Object."

I once used this technique to sort a big list of random abstract concepts into categories for use in a natural language Context-Free Grammar of wise sentences. The categories of Virtue, Vice, Reward, and Punishment towards the bottom of my MoralityBot's CFG source JSON here, while not particularly narrow, are each fairly semantically regular, and this is hard to accomplish just by thinking about concepts without any blank-sentence tests.

Let's get back to the example. I was thinking about those Things that can overwhelm a person's attention. Terms that are fit include: problems, tricky details, threats, obstacles, messes, informal contradictions, hard jobs, defeaters, disasters, conflicts, shortages, emotional triggers, self-doubts, social expectations, deontological prescriptions, obligations, emergencies, and hazards.

Some terms that are less abstractly cognitive and more specific to a physical domain also pass, but I don't like them as much: epidemics, acts of god, droughts, and famines.

The terms "people" and "gods" (thank you for the entry, Grace), pass all of the tests, but they seem very different from the other terms and very similar to each other. I think other terms like "demons" could also go here, while "inner demons" fits better with the previous words.

Here are some words that pass some of the tests, but I'd say not all: duties (which is interesting, because the very similar word "obligations" does pass), distractions, addictions, changes, omens, and reminders of my dead loves. The same way that the semantically-fit "inner demons" is similar to less-fit "demons", and "obligations" is similar to the less-fit "duties", I think the other almost-fit examples are evidence of yet-undiscovered concepts that are similar and fully fit.

Anyway, now we've got a cluster: problems, tricky details, threats, obstacles, messes, informal contradictions, hard jobs, defeaters, disasters, conflicts, shortages, emotional triggers, self-doubts, social expectations, deontological prescriptions, obligations, emergencies, hazards, and inner demons. What are they? What do they have in common? What is their semantic type? Some of them seem necessarily internal to the mind, while threats, obstacles, messes, disasters, conflicts, and shortages don't. They're bad things, but I think you'll agree that "diseases" are bad and yet the word only passes some of the tests.

I have no good second method here to take care of the products of the first method. Clustering by similarity, making up more test sentences, examining guesses that are true of some list members to see if they're true of all, thinking about where the concepts fall among ontological categories like Top Types, Formal Roles, and Phased Sortals. I have no good second method, but I still have work before me, because this is the kind of thing my mind does in its off hours. Also, this is partly trying to get at an insight which I think will help me to compress a bunch of thoughts that I want to include in a future post on varieties and sources of executive dysfunction. Or a bunch of posts. There's nothing wrong with incremental installments. One post on serotonin, one on anhedonia, one on ambiguity aversion, one on effort discounting or whatever. It'll be grand. And they will feature a unified understanding of the features of problem-like concepts. Do let me know if you see that they have something in common.