Sunday, September 25, 2016

Why does the hawk screech?

The red-tailed hawk has a great cry, or call, or whatever it is, which is something of a mystery. We hear it fairly often and call it a shriek, but it is generally known as a scream or a screech. You can hear it here:

The interesting thing is that no one is sure why the hawk is doing it. In fact the explanations tend to be contradictory. Sometimes it is said to be territorial, warning other hawks away. But then it is said to be a warning cry, but warning who? Other hawks? It seems to have been observed in both uses.

Another possibility is that it is a sort of a warning to something it might attack, that is a threat of sorts. My wife once came too close to an eagle's nest and was warned off by the eagle screaming while hovering over her. It was very effective.

But recently I observed a hawk screeching that seemed to fit none of these modes. I first saw it flying low over the horse field, shrieking as it went. It then landed in a tree, where it continued to call. This calling then went on, off and on, for over two hours, from various locations.

It is late September so there is no nest to guard. Nor does this seem like driving off a threat, because the hawk was all over the place, as it were. We only see or hear hawks very intermittently, so defending a territory seems unlikely. I suppose one possibility is that this was a juvenile hawk, marking out a territory for the first time, or some such, but that is a stretch since the migration is nigh.

So I am inclined to think that we just do not know what it was doing. The primary study on hawks appears to be this one from 1946:

But as the authors point out, this is a very specialized environment, where the hawks are closely packed. Territorial fighting is common, often carried out by mating pairs. We are more at the opposite extreme, in a national forest where the only open ground is the scattered valley farm. We only see hawks from time to time and seldom see two red-tails at once. I doubt there are any territories here, but in any case the protracted behavior I observed the other day was nothing like what the authors describe.

So we have yet another mystery, in this case a rather loud and entertaining one.

Saturday, September 3, 2016

How to do animal cognition research our way

One need not know anything about robotics to do the kind of research that I am describing here on this blog. The basic questions are these:

1. What decisions does the animal make in doing what it does?

2. What does the animal have to know or believe in order to make these decisions?

3. What concepts does this knowledge include, especially concepts that humans may not have?

Note that many human concepts are associated with language, but in animals most are likely perceptual in nature. For example, the animal sees that X, or looks for X, thus has the concept of X.

Just as I can search for something, animals do not have to perceive something at the time in order to use the concept of it. Horses going to a different field or a bird retrieving nest building material, for example. These animals are thinking about things that they cannot see at the time. Thinking ahead, as it were.

So our kind of animal cognition research is just a matter of thinking about the specific decisions and knowledge that the animal's behavior requires. A cognitive time and motion study, if you like, breaking the behavior down into its components and analyzing them. What are the steps? What decisions does each step require?

Of course one first has to know what the animal is doing and this can be a challenge. We often see the crows doing things that we do not understand, such as flying in a group from one place to another. But this is true of any science; one must correctly characterize a phenomenon in order to explain it.

In any case the point is that just because the behavior involves an instinct, this does not mean that no thought is required, quite the opposite. Instinctive behavior often involves expert knowledge, including complex concepts. It has to.

There is nothing anthropomorphic about this. A given behavior requires specific decisions and knowledge on the animal's part. It is as simple as that.

Saturday, July 30, 2016

Animal cognition research is not looking our way

 I have been poking around in the scientific literature on animal cognition and found several good reference collections. Here is the first:

This lengthy article on "Animal Cognition" recently came out in the Stanford Encyclopedia of Philosophy. It has an extensive bibliography, including a lot of recent items, indicating that animal cognition is a lively topic. In fact this is a revision of an earlier article, revised to incorporate a lot of new work.

In addition to referencing journal articles, there are even a number of books, although some of these are also collections of what amount to articles. Interestingly, many of the books are published by university presses, suggesting that they are not meant for a general audience.

Then there is Cogprints, which is what is called an eprint archive, where authors post what are typically pre-publication copies of scientific journal articles, or conference presentations. Cogprints is all about cognition, which involves a wide variety of disciplines, from biology to philosophy (my field) and artificial intelligence (which I am drawing on). Fortunately these contents are organized by categories and subcategories, including one for animal cognition.

They presently have 125 papers on animal cognition here:

The important thing is that I see nothing in these extensive collections to even suggest that anyone is pursuing the approach that we are developing here. No one seems to be talking about an instinct being a body of knowledge that is then used in complex decision making. We appear to be pioneering a new line of research, a new way to think about animal behavior.

Friday, June 17, 2016

Underestimating the flexibility of an instinct

We ran across the news report for an interesting study of bird nest building, which illustrates the common misconception that an instinct requires a rigid pattern of behavior.

The title is "New study says birds learn how to build nests."

I am certainly prepared to believe that a bird can improve its nest building with practice. In fact I would expect that to happen. How could it not?

The problem is that the researcher's evidence for this conclusion seems to be merely that individual birds built nests that varied over time. It is not that these successive nests got better over time, just that they varied, or so it seems from the news account.

One of the researchers -- Dr Patrick Walsh of Edinburgh University -- is quoted as saying "If birds built their nests according to a genetic template, you would expect all birds to build their nests the same way each time. However, this was not the case."

Dr. Walsh is simply mistaken about the supposed lack of flexibility in genetic templates. Instinct is expertise applied to a given situation. As the situation varies, so may the outcome.

For example, I have a bird's nest made completely of horse hair. I have another made from unwinding the filaments that coat Christmas bulbs, a very colorful example of the use of materials. I am sure there are no birds that depend on either of these materials. They merely used what was readily available.

I have also seen beaver dams made from very different materials. In one case our beavers built many dams using the materials from a grove of mature aspen trees, which they first felled. This dam included quite a few branches up to three inches in diameter. In the same area another set of dams, possibly built by the same pair of beavers, consisted entirely of saplings, as there were no mature poplar trees in the area. These constructions were very different, because the situations were different.

The point is that these instinctive activities allow for a great deal of local reasoning. That a bird, or a beaver, should build the same thing, the same way, every time is an incorrect view of how instincts work. An instinct is a body of expertise, not a specific set of behaviors. It is how to do something, not what to do in every different situation.

Tuesday, May 31, 2016

Baby behavior

It is spring here in West Virginia, with the usual coming of lots of baby critters. Babies are a good place to look for complex and knowledgeable instinctive behavior because they have not had time to learn much. Recently I was struck by two such observations.

In one case the critter was a baby starling that had left the nest for some reason but was still too young to fly. I was going to get on my riding mower when I saw it nearby. It saw me and ran to hide under the mower. This concerned me but when I got onto the mower it ran out and hid under my nearby truck. So I started the mower and left to mow.

Clearly this baby knew both when to hide and how to hide. It even knew when to leave one hiding place for another. From a robotic perspective this is very complex behavior. It involves a lot of different perceptions, which in turn requires requires complex concepts like "something to hide from" and "someplace to hide." it also involves a lot of decision making that uses these perceptions and concepts.

Of course we see this sort of behavior all the time, in all animals. But because we take it for granted, we fail to see just how complex and instinctive knowledge based the animal's reasoning must be.

As an aside, note that the running away decision may be much more complex than simply fleeing big moving critters. Our horses are presently sometimes followed by small flocks of cowbirds when they graze. Presumably the birds are feeding on insects, or something, disturbed by the horses. These birds often work within inches of a horse's hooves, even though the horse moves them every few seconds, grazing in motion as I call it.

But if I approach the cowbirds all fly away. Clearly the cowbirds recognize the horses as something not to be feared, while fearing (or at least avoiding) me. They must know the difference between a horse and a human. Note that this need not require them to distinguish horses from cows. They may have a generic concept that we do not understand.

The second case was a groundhog moving her babies from one den to another, over a distance of several hundred feet. She did this by carrying each youngster, one at a time. They were rather large so there was some difficulty.

Unlike cats, groundhogs grasp their baby by the base of the tail, not the scruff of the neck. The baby allows this passively. It also looks like the baby actually curls forward, hence upward, which would facilitate the carry. So not only does the mother know how to carry; the baby knows how to be picked up and then carried. This is instinctive knowledge at work.

Note too that the mother's decision to move the babies to another existing den is impressive. One wonders what specific knowledge, reasons and reasoning led to this decision, as well as what concepts were used in that reasoning?

The point, as always with our examples, is that an instinct is a body of expert knowledge, one that uses concepts that are specific to the critter's nature and needs. The robotic perspective helps us see just how complex these behaviors are.

Sunday, April 17, 2016

Instinctive specialized learning (in Phoebes and other critters)

We are watching a Phoebe rebuilding last year's nest on our Observatory (in a nice safe place by the way, another instinct). This rebuilding may well require different materials than building a nest from scratch, but the amazing thing we saw was where the bird went to get these materials.

We are on a small rise, facing a field where the horses graze. Beyond the field, directly in front of us, grows a row of dense bushes. The same is true for the field border to our right, which is at 90 degrees to the far border. Both borders are well over a hundred yards away, a considerable distance indeed.

First the Phoebe made two successive trips to the bushes across from us, bringing back materials each time. It went to roughly the same place and may well have gone to exactly the same place. Then it abruptly changed behavior and went to the bushes on our right, far away from where it had gone before, again returning with materials. Moreover, on each trip the bird spent only about 10 to 20 seconds finding and collecting the material.

Clearly this bird knew what it wanted and where to get it, despite it being (1) a long way away and (2) in two very different places. The bird must have had detailed advanced knowledge in order to do this. It must have spent some time, perhaps a lot of time, surveying the surrounding area and locating suitable materials, before beginning this round of nest building.

This pre-knowledge behavior implies an instinct to learn. Perhaps something of the form "go forth and locate nest building materials."

And of course the bird must remember what it has learned. I recall noting this long distance behavior last year. Perhaps the bird located the suitable materials then.

Note that this concept of locating suitable nest materials in advance is more abstract than simply needing something specific for a nest that is under construction. Nests can be built from a variety of materials. In fact it may be that the Phoebe builds its nest based on the materials it has previously located.

I am inclined to think that the instinct to learn like this is widespread in animals. What is especially interesting here is that it is an instinct to learn something highly specialized. This too may well be common. For example, I earlier discussed the horses deciding to explore a new field, that is to learn about it, rather than begin grazing. Then too there is the case of horses changing fields to graze on a particular plant, which they clearly already knew about.

The instinct to learn is certainly a concept worth exploring further.

Saturday, March 26, 2016

Horse face communication is complex

The University of Sussex has done some great research on how horses communicate using facial expressions, just as humans do. In fact some of these expressions are similar between humans and horses. Others are different, such as the horse's use of its very mobile ears.

The research has been done by Professor Karen McComb and her students at the Mammal Vocal Communication and Cognition Research Group. Here is part of a Science Daily news article about some of their work:

"Horses are sensitive to the facial expressions and attention of other horses, including the direction of the eyes and ears. The findings are a reminder for us humans to look beyond our own limitations and recognize that other species may communicate in ways that we can't, the researchers say. After all, human ears aren't mobile."

They even found that horses are able to distinguish between angry and happy human facial expressions. Here is an article about that work:

The point for us here is that the horse's understanding of other horse's (or people's) expressions is instinctive, not learned. For that matter it is instinctive in people as well. There are a lot of different expressions, so this is a complex body of instinctive knowledge.

McComb's group has even done impressive anatomical work toward cataloging these expressions. See this article:

We have certainly observed a lot of facial expressions in observing our horses, especially the ears, which can be quite active. Given this research we will pay more attention to these expressions and the behaviors they are involved with.

It would be very interesting to know, in some sense, how many expressions there are and what they mean. Again, the point is not that the horses know how to do these expressions, but that they know what they mean when other horses do them.

The expression itself may not involve knowledge, but the communication certainly does. This seems like a sizable body of instinctive knowledge. As with our other examples, imagine trying to build a robotic horse that responds correctly to the facial expressions of real horses. No mean feat that would be.

Sunday, February 7, 2016

Behavior based cognition: the Research Program

The previous three posts suggest that even trees exhibit cognition. In a way what we are doing here is a form of behaviorism. We are looking at behavior and asking what concepts and decision rules this behavior requires?

Classical behaviorism, which is now almost a hundred years old, posited stimulus and response mechanisms to explain behavior. In contrast, we are drawing on much more recent advances in artificial intelligence, especially robotics and expert systems.

These technologies have led scientists and engineers to think deeply about what is required to produce even seemingly simple human behaviors. What I am proposing is a straightforward extension of this research, to include animal behavior. Perhaps even plant behavior as well, but that is really more of an aside.

I am not concerned with thinking or mental processes in general. The point is simply that a given behavior requires certain concepts and decision making.

For example, a bird cannot build a nest without first finding and choosing a nest site. This behavior requires seeking, recognition and decision. The bird has to know what a good nest site looks like. Moreover it gets this knowledge via instinct. As we have said from the beginning, instinct is a way of knowing such things, an alternative to learning.

If we look closely and think about it, everywhere we look we see animals exhibiting this sort of behavior. The research question is how to unpack it? That is, how do we describe the concepts and decision rules that underlie this behavior?