A blog about problems in the field of psychology and attempts to fix them.

Thursday, December 15, 2011

Beyond the Brain: Anti-Anthropomorphism

As luck would have it, my semester is wrapping up, and my philosophy colleague just returned my copy of  Beyond the Brain. He agreed it was excellent and has ordered his own copy. I have two or three more topics I wanted to write about in my extended review of Louise Barrett's excellent book. The first two chapters of the book deal with the problem of anthropomorphizing the behavior of animals, and the topic reappears several times in the later chapters. Barrett is convincing that anthropomorphism causes serious trouble when we try to explain behavior (even human behavior) and she advocates a system in which anthropomorphic terms are acceptable when they refer to evolutionary processes rather than immediate happenings. This is all good, but I think psychology can push further, and use mental terms in a way that refers to immediate happenings, while still avoiding the pitfalls of naive anthropomorphism.


Anti-Anthropomorphism

Barrett begins the book with a brilliant set of anecdotes showing how quick we are to attribute sophisticated mental traits to animals based quite limited evidence. That is, we readily assume that the animals are doing whatever would best explain our behavior if we were acting in a similar way in a similar circumstance - and by "doing" we usually mean some elaborate, mental, conscious and rational, decision making and planning process. Of course, the question of whether such explanations are typically needed to account for human behavior will loom large throughout the book, but it is enough initially to show that it is questionable to assume that animals engage in such processes. There are more than a few reasons why anthropomorphism is problematic, Barrett highlights three:

"First, an anthropomorphic stance means that often we end up asking scientific questions that simply reflect our own concerns." (p. 3) That is, it draws us away from the hard work of determining the situation from the point of view of the animal in question - what is the challenge (or is there a challenge) for that animal, and what are the crucial factors that animal is sensitive to? This is a much more interesting and challenging question to answer than "What would I think if that happened to me?"

"[Second] it cuts both ways and can create errors in both directions." (p. 5) Even just the phrase "anthropomorphism" implies that there are traits that are uniquely human, but it has been remarkably hard to identify such traits (as an animal behaviorist, I'll note that the repeated attempts at identifying the uniquely human ability have wasted countless research lifetimes). That we often end up using terminology that simultaneously denies that animals have certain abilities while attributing those abilities to them causes nothing but confusion (e.g. to say that 'the dolphin has very human-like memory' is to both deny and attribute human memory).

"Finally, anthropmorphism is a problem because the attribution of human characteristics often results in confusion about what, exactly, we have explained about an animal's behavior and psychology." (p. 6) Here Barrett is particularly concerned about confusion between functional explanations and mechanistic explanations of behavior (more on that below).

There is critique of Frans de Waal and others who argue that we fully embrace anthropomorphism, and who try to shift the burden onto those who would deny animal's abilities. (Advocacy of this tradition as a basis for scientific work goes back to Darwin's disciple Romanes.) The first highlight of this section is a nice counter to the simplistic application of Occam's razor. Many proponents of anthropomorphism argue that, given what we know about humans, it would be silly to assume animals work in a completely different way. Putting aside how such arguments completely beg the question, Barrett counters that:
a simple explanation is not necessarily a virtue in and of itself. The real question is this: do we enhance our scientific understanding by adopting this stance? (p. 12)
The second highlight is a section in which the notion that we can determine the relative parsimony of cognitive explanations is exploded. This is an issue near to my heart. Parsimony arguments get bandied about fairly often in psychology and animal behavior circles, and if we ever really tried to take them seriously we'd quickly realize what a mess we were in: Are explanations based on elaborate histories of operant conditioning really more parsimonious than explanations based on 'mental rotation' ability? Are explanations based on the perception of affordances really more parsimonious than explanations based on imitation? (Incidentally, though I have not had time to read all the articles, the most recent issue of IJCP seems to be confronting some of these issues.)

Mental terms in functional explanations

Barrett advocates the use of mental terms as part of a functional explanation of behavior, especially when such terms are used to generate testable hypothesis about evolutionary processes. So far as I can tell, advocacy of this approach dates back to Williams, and was also heavily endorsed by Dawkins and Krebs in their influential work. Barrett adds a few more modern endorsers of this approach, including John Kennedy, who I haven't read, and Daniel Dennett, who I have mixed feelings about. There are two parts to this argument; part one is that it is alright for us to use anthropomorphic terms in this manner, part two is that this is what good scientists already do in biology.
For example, if we argue that male frogs sit and call by a pond all night because they "want" to attract a mate and "know" that calling will entice females, we are using the words "want" and "know" in a purely metaphorical sense. (p. 7)
What we mean (Barrett asserts) is that natural selection has selected for males who successfully attract mates and that those males of the past who have sat and called by ponds all night have successfully attracted mates and passed on their genes. In a sense, it is Natural Selection that "wants" and "knows" by this account - wants new good replicators and knows to get them through past good replicators. There are plenty of techniques to test the evolutionary hypothesis thus implied: Do frogs that call all night outperform those who only call part of the night? Does position near ponds make a difference? etc. Such "mock anthropomorphism" is deemed acceptable, as it makes no claims about any immediate frog-motivation or frog-knowledge. Such efforts at redefinition-into-empirically-measurable-terms is a version of the crude operationalization that most psychologists are be familiar with. Barrett goes on to suggest that some degree of anthropomorphism is inevitable, given our penchant for it and the limitations of human language, but that as scientists we can use the terms in this more technical ways, and ultimately that will be acceptable.

In my view this is a good, but not great solution.

Why not just state evolutionary hypotheses?

 The first reason I don't like this solution is because I do not think it is that hard to simply state evolutionary hypotheses, e.g., selection has favored male frogs that sing all night by pond's edge. In fact, writing that out explicitly demonstrates how incomplete the hypothesis is: sexual selection? natural selection? by what mechanism? due to what advantage? In any case, if it is scientists using the terms, it is not that hard to drop the anthropomorphic jargon and be state your hypotheses plainly. When it does help to use the anthropomorphic language in this context, it is usually because you are dealing with students, lay people, or the media. In such cases, it helps to use the language exactly because the listener will assume the terms mean things you do not want them to mean. Thus, endorsing anthropomorphic language in these cases seems like an attempt to have your cake and eat it too. The association of such a stance with Dawkins reinforces this view for me; his use of the term "selfish gene" is a classic example of trying to garner appeal through terms that imply everything you claim to be denying. (I actually really like Dawkin's early work, despite this flaw.)

Now, even with that critique, I'll admit that thinking in anthropological ways can sometimes lead to interesting, testable, evolutionary hypotheses... which is Barrett's main argument. However, I'm not sure why we would continue to use those terms when we report on our results. It would be one of the many aspects of research that, for better or worse, gets cleaned up in the process of article writing.


Mental terms as objective descriptors of behavior

A better solution to the problems of anthropomorphism is a critical evaluation of the behavioral patterns that our mental terms refer to. If we want to look at solid, biologically based traditions, the ethologists used mental terms frequently, and they meant them to refer to proximate aspects of animal behavior. The ethological tradition's overall agenda was to demonstrate the evolution of behavior, but an important part of this was establishing what, exactly, animals were motivated to do. Careful experiments were designed to determine how animals varied their behavior to achieve certain ends, which aspects of the environment they were responding to in which ways, etc. The point was that careful study of the behavior of the animals relative to their circumstances objectively revealed the minds of the animals. That is, when you examine the ethologist's methods, it is clear that mental terms never referred to inner entities, they referred to patterns of action. Barrett tacitly accepts this in one of her thesis statements:

 a reduced focus on the nature of animals "inner lives" and greater attention to how their brains, bodies and environments work together will give us a deeper understanding of how intelligent, adaptive behavior is produced. (p. 18, emphasis added)

"Intelligent" is a descriptor of some behaviors when compared with other behaviors, and to "have intelligence" is to act in intelligent ways, not to have some inner entity "intelligence" that causes your behaviors. To have a science of psychology, we can't abandon mental terms, nor foist them off onto evolutionary biology. Rather we need to determine what we are talking about when we use mental terms and the empirical evidence necessary to confirm or deny claims about animal minds. With such a basis, we can avoid both sides of the anthropomorphism problem that Barrett is rightly concerned with, we wouldn't accept intuitive generalizations to animals and we wouldn't naively attribute abilities to humans without empirical evidence. This is a recurring theme in this blog, and I'm still working on better and better ways to articulate the position, but I wanted to place it in this context.

Barrett could clearly use this approach in several places. A few examples:

Barrett has nice discussion of how, in the classic Heider-Simmel display and similar modern variants, people can be lead to claim that a moving triangle is sad, though of course the triangle does not feel sad. True, but this nicely avoids the issue of determining what, exactly, people are responding to when they experience a moving object as sad.

In the robotics section, Barrett has convincing arguments that it is important to distinguish between an organism's goal and the reliable outcomes of the organism's behavior. (I'm not sure she ever phrases it exactly that way.) This is true, at the least, it is clear that only some reliable outcomes occur because the organism intends to produce that outcome. It is also true that if we have only naturally occurring observations to go on, it is often impossible in practice to make this distinction. However, that is not because it is impossible to objectively determine an animal's goals, or because the goals are located somewhere "inside" the organism. It is because distinguishing an intended outcome from an incidental outcome requires experimentation, i.e. manipulation of the organisms conditions, to see how the organism redirects its behavior. Barrett recognizes this implicitly in some of her other discussions, for example where she shows that weaver birds do not actually intend to build a nest of a specific shape. This is evidenced by manipulating the nest mid-construction, and noting than rather than correct course, the bird produces nests of very different shapes. Rather, the intention of the bird seems to be limited to maintaining local relationships during the course of construction, e.g. if this part of the nest looks like X, attach the next part in relationship Y. This produces intelligible new structures in a manner very reminiscent of the discussion in The Plausibility of Life (also highly recommended).

At any rate, the point is that many points where Barrett seems to get into trouble would be solved by the argument that mental terms can be used in a proximate way, to refer to the structure of an animals behavior across different circumstances. We need not banish the use of such terms to discussion of evolutionary mechanisms. Given how little trouble there is in the whole book, I might even wager that this would smooth out the vast majority of sections where I felt some strain.

4 comments:

  1. Eric, I'm a little confused. You seem to be saying a) the mental terms stand in for rigorous and testable hypotheses, so we should really use the latter as good practice, but then b) we can actually use the mental terms. What have I missed?

    ReplyDelete
  2. Hmm.... I can see where I might have caused that confusion. Let me try to clarify. This is inherently a bit awkward, so let me know if I need to try again.

    Traditionally mental terms are used to refer to something about this organism, now. The approach I'm advocating expands that a little bit, in that I think mental terms refer to a pattern of behavior, of which the currently observed happening is an exemplar. For example, to say that a rat "chooses" to go left in a T-maze, is to contextualize this particular rat-going-left with the many behaviors you have seen in similar circumstances (including at least some going-rights). Thus, choice isn't a thing that happens in the head, nor in some ethereal mind-land, choice is a doing, and we can observe it objectively. It takes many observations over time, perhaps with experimental manipulation, but it is behavioral, and hence can be studied objectively. This addresses your point b.

    Now, about point a.... For quite a few years now, researchers coming out of a particular tradition of evolutionary research have advocated a very different use of mental terms. They have advocated that, to expedite theory formation and communication between researchers, we can use mental terms as a stand in for evolutionary hypothesis. If we followed this to its logical conclusion, we would say that mental terms do not refer to anything about the current organism, but rather something about the life and death of many generations that came before. For example, if these people say that a given female fish "prefers" males with long tales, they would want you to understand that as hypothesizing that - successful female members of past generations of that fish reproduced more effectively if they mated with long-tailed males than if they mated with small-tailed males. This risks confusing levels of analysis in really weird ways*, and is totally unnecessary, as it is just as easy to take the mental terms out of it and simply state the evolutionary hypothesis.

    I suppose a more neutral way to state that last part is: While the use of mental terms to refer to evolutionary processes might be fine in evolutionary biology (as long as they remember what they are doing). But it will not work as the basis for a scientific psychology, because psychologists are exactly interested in the proximate structure of behavior. Thus, we need mental terms to refer to proximate structure.



    -----
    * Worth noting, if we look at all the really good research about female tail-length preference, we find that it focuses on proximate behavior!

    ReplyDelete
  3. That works. The distinction between referring to processes over evolutionary vs individual organism time wasn't clear.

    So your suggestion is state your evolutionary hypotheses in detail, but feel free to (carefully) use mental terms to refer to immediate behaviour? That seems sensible, other than the obvious risk of not being careful.

    ReplyDelete
  4. That is my suggestion!

    Careful use of terms is a perpetual problem in psychology, and that is one of the major signs that we are not as advanced as we like to think as a science. When I teach my Psychology of Learning class, the hardest struggle for the students is that the words used in class mean something specific, and they cannot be used sloppily.

    ReplyDelete