Sunday, July 14, 2019

Evening Note for Sunday, July 14

Thought for the Evening: Hayes on Liquids

One of the most interesting works in what could be called analytic philosophy in the twentieth century was not a work in analytic philosophy but in computer science. Pat Hayes in 1979 wrote a work called "The Naive Physics Manifesto", which criticized a great deal of artificial intelligence research for playing around with toy models built simply to be toy models. He proposed that researchers instead should be focused on the formal representation of actual common-sense knowledge. No more theories made for artificial models made for theories; rather, deal with the real world, the everyday world. Hayes himself gave a demonstration of how this might work, using common-sense knowledge about liquids, "Naive Physics I: Ontology for Liquids", and the result is, I think, far superior to most work in analytic philosophy. How can you represent, in a formal, logical way (Hayes prefers to use just ordinary first-order logic with an occasional bell or whistle), the way we reason about liquids?

Standard first-order logic works as well as it does because it tracks relations between individuals. Liquids obviously pose something of an initial puzzle for this, because liquid, as we find it in everyday life, does not come in obvious discrete units. It's difficult to pin down what a piece of liquid would be (although Hayes eventually does so). But liquids can have a spatiotemporal continuity, a unified history, and we can and do make sense of this in terms of contained quantities of liquid: water in a lake, tea in a cup, and even a raindrop is contained by its own surface cohesion. So we can think of containers (it doesn't matter what kind), which we can call c, and then talking about the inside of them, inside(c), and this contained space can harbor an amount of a liquid, amount(l,s). Even with just this much we can represent quite a few things. Amounts are partially ordered; there is a zero amount, which we can call none. For instance, we only need this much and 'greater than' to say that there is tea in the cup:

amount(tea,inside(cup)) > none

There are plenty of other things that we might want to add: a way to compare the capacities of different containers, metric units, definitions for things like 'full' or 'channel'. Some of this can potentially be complicated, and requires hard thinking about things like measurement or surfaces, as well as the kinds of activities or processes liquids undergo.

One of the things he makes in order to try to make the work easier is a taxonomy, an 'ontology', of liquids. In everyday reasoning there are features of liquids that have a particularly important role to play in distinguishing different kinds of liquid situations. Some liquid is bulk, some finely divided into drops. Some is lazy (normal behavior of water on its own), some is energetic (requires some activity to maintain). Some is supported, either inside a space or on a surface, some is unsupported. Some is moving, some is still. These have various relations that can be traced out; for instance, while not all lazy water is still (for instance, falling water is lazy and moving), all still water is lazy. And, Hayes says, "Of the 32 logical possibilities, only 15 are physically possible, even allowing souch outrè possibilities as mist being blown along a tube" (p. 86). These 15 can be put in a table that looks something like this, with examples:

wet surface liquid flowing down a sloping surface waves on a shore? SUPPORTED ON SURFACE BULK
liquid in a container river flowing along a channel liquid pumped through pipe SUPPORTED IN SPACE
falling column, as in a waterfall rising column, as in a waterspout UNSUPPORTED
mist in valley? mist rolling down valley? mist blown through tube? SUPPORTED IN SPACE
cloud of mist falling rain splashing spray UNSUPPORTED

I have sometimes wondered how Hayes came to his conclusion that there were only fifteen possible cases here. Hayes recognizes that there are other states for water -- he mentions liquid soaked up by something, liquid, suspended across a mesh, and free-floating bubbles -- but I take it that he thinks that the fifteen capture all the physical possible cases that we get if we only look at these possible features of liquid. Presumably it's right to rule out LAZY STILL UNSUPPORTED BULK, which we might perhaps get with blobs of liquid on a space station but not in any everyday circumstance. (I have somewhere a children's book from the space shuttle days in which liquid blobs floating in space are highlighted as a weird and new thing that astronauts deal with, which can be taken as evidence that there seems something fantastic rather than everyday about it.) Are there really no ordinary cases of LAZY MOVING SURFACE-SUPPORTED DIVIDED and ENERGETIC MOVING SURFACE-SUPPORTED DIVIDED? If drops on a surface are LAZY STILL, there seems an obvious possibility for ENERGETIC MOVING -- drops skittering on a hot surface. And if that would count, then LAZY MOVING would obviously be single drops rolling off a sloped surface. A single falling tear is LAZY MOVING, and it is DIVIDED, and it seems to be SURFACE-SUPPORTED.

In any case, the idea is that for each of these you can formalize some basic principles that govern common-sense reasoning about them; Hayes himself only looks at LAZY BULK, suggesting that at least a lot of the principles would carry over to the other cases. To do this he has to work through questions like, "How should you characterize a liquid's wetting something?" (obviously this requires looking at how surfaces work) and "How should you characterize change in the liquid?" (for which Hayes suggests we should consider not just the liquid at a time but the kinds of histories a liquid can have). It takes some work to come up with the principles, but it turns out that you can describe a lot of situations involving liquids with relatively few of them.

One of the important things that Hayes notes -- I think it is probably the single most important idea in an article full of important ideas -- is that the taxonomies are not a secondary matter. They do significant work in the reasoning. In many cases they are what make the axioms or formal principles even usable to begin with, and they also serve a function in ruling out possibilities, which lets you draw more conclusions from your formal principles than you otherwise would be able to draw. Classification is a central part of reasoning itself.

Hayes' work touched off an interest in 'ontologies' in computer science, some of which has been very worthwhile and interesting, and some of which has not been so, but it's an admirable bit of work. One could wish that more 'naive' work of the sort had been done in more fields.

[Patrick J. Hayes, "Naive Physics I: Ontology for Liquids", Formal Theories of the Commonsense World, Hobbs & Moore, eds., Ablex Publishing Corporation (Norwood, NJ: 1988) 71-107.]

Various Links of Interest

* Natalja Deng, What is temporal ontology? (PDF)

* Daniel A. Kaufman, Feeling Like a Man

* Richard Marshall interviews Christopher Shields on Aristotle and metaphysics at "3:16".

* Lisa Shapiro on the history of philosophy

* The Beast of Gévaudan

* Matthew Wills looks at Sor Juana Inés de la Cruz, at the JSTOR blog.

* Ofra Magidor, Category Mistakes, at the SEP

* Brett T. Feger, The Importance of Good Posture, looks at what Aquinas says about the subject.

* Rob Alspaugh, Biblical Weaponry and Josiah's Failure

* Arend Smilde, Horrid Red Herrings: C. S. Lewis and the "Argument from Desire"

* Amy Olberding on the problem of incivility

* Eduard Habsburg on the dissolution of Austrian monasteries in the early modern period

* If you are an American wondering how to contribute to constructive handling of current immigration problems, I have heard very good things about both Annunciation House and CLINIC.

* Willis Renuart, In Praise of Religion's Dark Side

* Chateaubriand on the “Dangers Facing the United States” (1846)

* Bl. John Henry Newman is due to be canonized on October 13.

* Mark Spencer, Beauty, First and Last of All Transcendentals

* Merlin looks like an interesting approach to public philosophy

* Undergrads learn about humanity first-hand by studying philosophy with incarcerated individuals

* I have been spending way too much time watching the videos at the Townsends YouTube channel (there's an auto-play video). Here's one on how to make barley water:

Currently Reading

Oscar Wilde, The Complete Plays
Oscar Wilde, De Profundis (Unabridged)
St. Peter Damian, The Letters of Peter Damian, 1-30
St. Augustine, On the Trinity
Hegel, Philosophy of Right

No comments:

Post a Comment

Please understand that this weblog runs on a third-party comment system, not on Blogger's comment system. If you have come by way of a mobile device and can see this message, you may have landed on the Blogger comment page, or the third party commenting system has not yet completely loaded; your comments will only be shown on this page and not on the page most people will see, and it is much more likely that your comment will be missed.