Could it not be explained by an (incorrect) over-generalisation of the rules learnt by observing animate objects?

What positive evidence is there to support the claim of some kind of "agency detection module"? Most/all of the discussion in the article linked to in the OP just identify the fact that children attribute some kind of purpose and agency to things, but I'm not aware of any study which attempts to demonstrate that this is innate rather than learnt.

Slightly related: I think Hood's woeful understanding of behaviorism and laws of learning makes some of his claims in "Supersense" questionable, where he seems to reject possible learning explanations based on a superficial understanding of the subject, rather than an accurate review of the explanations.

seeker wrote:

Mr.Samsa wrote:Could it not be explained by an (incorrect) over-generalisation of the rules learnt by observing animate objects?

Maybe, but how would you test this? Of course, learning and overgeneralization would be a plausible hypothesis. But I guess innateness is also a plausible hypothesis, given that there's evidence of some innate complex behaviors in other species. I guess transcultural data could be an indirect evidence. Do you know any transcultural data?

I'm not aware of any solid studies on it no, but it would be a good initial test. If it's not cross-cultural, then it would be difficult to argue that it's innate, but if it is cross-cultural then we'd be back at square one since it could still be explained by both approaches.

I think to really test the learning hypothesis properly we'd have to come up with a concrete, detailed theory and then we could look for how such an idea would explain the phenomenon in more general terms. For example, in a similar way as Sperber, Cara, and Girotto (1995) proposed 'relevance theory' to explain the results generated by the "cheater-detection module" in a way that didn't require an innate mechanism (discussed in some detail here).

seeker wrote:

Mr.Samsa wrote:What positive evidence is there to support the claim of some kind of "agency detection module"? Most/all of the discussion in the article linked to in the OP just identify the fact that children attribute some kind of purpose and agency to things, but I'm not aware of any study which attempts to demonstrate that this is innate rather than learnt.

How would you test this? Could an experimental study be done about this issue?

It would be difficult, but what we've discussed above would be a start (cross-cultural studies, testing a generalised learning rule, etc).

seeker wrote:

Mr.Samsa wrote:Slightly related: I think Hood's woeful understanding of behaviorism and laws of learning makes some of his claims in "Supersense" questionable, where he seems to reject possible learning explanations based on a superficial understanding of the subject, rather than an accurate review of the explanations.

Yes, I agree that Hood doesn't understand learning, but some of the studies he mentioned seem to be interesting. Have you seen the third link I've mentioned? The author claims that religious behavior is an adaptation (that it was selected because of its effects on the reproductive success of ancestral humans), but he's doesn't claim that religion is currently adaptive (nor maladaptive). What do you think?

I haven't had a chance to read through it yet no, but it does look interesting. I think the problem would be that anything can be argued to be adaptive - we just need evidence that something is innate first, and once we've established that, we can look at whether it's reasonable to view it as an adaptive behavior or not.

seeker wrote:

Mr.Samsa wrote:I'm not aware of any solid studies on it no, but it would be a good initial test. If it's not cross-cultural, then it would be difficult to argue that it's innate, but if it is cross-cultural then we'd be back at square one since it could still be explained by both approaches.

I would expect that many human innate propensities are modified by learning, so we couldn't assume that finding different learned modifications excludes the possibility of an innate propensity. Learning can both lead to learned differences from similar innate tendencies and to learned similarities from different innate propensities.

True enough, but there would still need to be an invariable component which is common across cultures. It's okay if the final form it takes differs, or if learning affects the development of the behavior, but ultimately we do need to be able to point out a specific predisposition - otherwise it's a purely hypothetical construct that has no observable or measurable qualities.

seeker wrote:But finding that a propensity is highly probable and that current environmental influences don't seem to influence it, would be an indirect evidence for saying that the propensity is innate.

Yes, but the difficulty is in ruling out all environmental influences (which is impossible to control for in some cases, at least ethically). This is because there will always be environmental conditions common across cultures - things like gravity, day-night cycles, existence of other people, etc. So we get species-specific constraints producing universal behaviors entirely through learning in what are, superficially, vastly different cultures. I think we've discussed this before, but for example, all cultures in the world eat soup from a bowl. There doesn't seem to be any common environmental variable across the cultures - they all use different languages, different ages, etc. But the commonality is the existence of gravity, because obviously people who try to eat it off a plate burn their laps.

seeker wrote:

Mr.Samsa wrote:I think to really test the learning hypothesis properly we'd have to come up with a concrete, detailed theory and then we could look for how such an idea would explain the phenomenon in more general terms. For example, in a similar way as Sperber, Cara, and Girotto (1995) proposed 'relevance theory' to explain the results generated by the "cheater-detection module" in a way that didn't require an innate mechanism.

Sperber's theory will also require some innate mechanisms (at the very least, learning principles and some perceptual propensities). I think it's wrong to think that innateness and learning are dichotomic concepts: we should explore the contribution of both components in each behavioral class.

Certainly, I agree. Sperber et al.'s theory (as they mention in their paper) does not rule out the possibility that the fundamental logical rules at the base of their theory are innate. The point is just that the cheater-detection module is not innate itself.

And I agree that nature and nurture are not dichotomous concepts.

seeker wrote:

Mr.Samsa wrote:I haven't had a chance to read through it yet no, but it does look interesting. I think the problem would be that anything can be argued to be adaptive - we just need evidence that something is innate first, and once we've established that, we can look at whether it's reasonable to view it as an adaptive behavior or not.

I disagree with the claim that "anything can be argued to be adaptive". Many traits don't fulfill the requirements (causing an increase in the survival and reproductive success of individuals with the trait, being selected by an environmental pressure over another variations, being transmitted through genetic inheritance). Also, we should make a distinction between "current adaptiveness" and "historical adaptiveness" (a trait might be maladaptive in our times but adaptive in the times of our ancestors, or viceversa), because the concept of "adaptation" only refers to traits that were adaptive in the past and were selected for such historical adaptiveness (current adaptiveness is irrelevant).

The ease with which just-so stories can be invented is currently one of the major criticisms of evolutionary psychology - the problem is that even the well-accepted theories in the field don't have all the requirements, particularly the criterion that it can be transmitted through genetic inheritance (which necessarily includes some genetic evidence).

seeker wrote:Also, I have doubts with the criterion of "needing evidence that something is innate". Most human traits, including innate propensities, will be influenced by learning. Assuming that innateness and learning are dichotomic concepts, and searching for "pure" cases of innate versus learned traits doesn't seem to be a useful strategy. Instead, we should explore the contribution of both components in each behavioral class.

I agree, but we need evidence that a behavior is innate before we can start discussing its adaptive value meaningfully. For example, we could easily discuss how having red blood is an adaptive trait - it makes injuries clear to us so we can deal to them, there are obvious environmental pressures that would continue this (i.e. those who couldn't tell where they were injured wouldn't be able to tend to their wounds), and it has a clear genetic component. However, we know that obvious there is no adaptive component so it's a waste of time discussing these possibilities. Another example is tool-use in the New Caledonian crows - it's easy to think up adaptive explanations for how the ability to use tools could come about, but our ideas of what selection pressures would lead to this and the genetic component would be entirely speculative (and mostly wrong) if we didn't have the behavioral evidence we have. That is, we know that only part of the tool-use is genetic; the selection and cutting of leaves, the actual tool construction, however, is a product of cultural learning. If we built adaptive explanations on tool-use, then we'd be wrong, but after the evidence came in we can build adaptive explanations for the 'attraction' to the certain kind of leaf they use and their 'cutting' behavior. Look at the cheater-detection example: they had vast frameworks explaining the adaptive history and value of such a behavior, yet it turns out to be entirely useless and wrong.

palindnilap wrote:

Mr.Samsa wrote:What positive evidence is there to support the claim of some kind of "agency detection module"? Most/all of the discussion in the article linked to in the OP just identify the fact that children attribute some kind of purpose and agency to things, but I'm not aware of any study which attempts to demonstrate that this is innate rather than learnt.

To say the truth the "agency detection module" seems a so irresistible hypothesis to me that I haven't paid special attention to empirical support, and maybe I should have done. The extreme adaptiveness of agency detection is so obvious that it would seem odd that evolution has relied on it having to be learned.

Two important points here though:

1) As mentioned above, the "obviousness" of how something can be adaptive can be applied to near any trait or behavior that you can think up.
2) Evolution not only attempts to create "cool" and useful traits, but it also attempts to do it on a tight budget - if setting up basic learning rules can lead to things like self-control, altruism, memory, etc, and agency detection, then why specifically use energy to encode these into a more rigid, inflexible innate predisposition?

This isn't to say that it isn't innate, just that it's usefulness and practicality is not a reason in itself to preclude learning.

palindnilap wrote:But that also means that I am not very strong on evidence for innate agency detection. I have seen convincing examples of agency detection being well hidden in language subtleties, but since I know that you reject innate properties of language it will not say much to you (and anyway agency detection is clearly more basic than language).

I'm not really decided on the language issue - I tend to take a more "environmentalist" stance simply because everyone seems to be a nativist, and I find that position based on largely poor logic. I'm interested in what language subtleties you're referring to here though.

palindnilap wrote:A hint is also the difficulty with which we can overrule the attribution of agency to points moving on a screen in a agent-like manner.

We'd expect this with learning too though. If a discriminative stimulus has been associated with a particular thing (i.e. if movement has been associated with agency), then it will be difficult to retrain or "overrule" this initial association using the same discriminative stimulus. More simply: Imagine that the government decided that, starting from today, traffic lights will be reversed so that red means "GO" and green means "STOP". You will find that a lot of accidents will result because overruling the initial association will be difficult, but this doesn't prove that we innately view red to mean "stop" and green "go". A direct demonstration of this is called the stroop effect where participants are asked to name either the written word, or the colour the word is written in, and we find that response times increase when the written word is not the same as the colour. For example: RED GREEN BLUE

palindnilap wrote:A place to go would be to study individual differences, where we should see very small continuous differences and some rare all-or-nothing differences. Autism springs to mind, but it is not an area that I know well. Tell me what approach would seem most promising to you.

I'm not quite sure what different predictions would be made with individual differences? And autism could be an interesting example, but the problem would be that we obviously wouldn't know whether their development disorder uniquely affects some innate mechanism, or affects the learning of agency detection, and unless we could figure out what environmental conditions might produce agency detection, we wouldn't know whether the circumstances surrounding autism would sufficiently rule these out.

palindnilap wrote:

Mr.Samsa wrote:What positive evidence is there to support the claim of some kind of "agency detection module"? Most/all of the discussion in the article linked to in the OP just identify the fact that children attribute some kind of purpose and agency to things, but I'm not aware of any study which attempts to demonstrate that this is innate rather than learnt.

I forgot about the definitive argument, the one that thwarts all contradiction : the number of rational skeptics in the Philosophy forum who will stretch any possibility in order to rescue the fiction of free will. Err, me included.

Free will has no real relevance here though, since agency detection would still occur without "true" agency.