| |
Le texte est disponible en version anglaise.
Abstract
En admettant une ontologie réaliste du monde des objets physiques et une épistémologie de la connaissance singulière dépendante des objets, nous comparons la poursuite (‘tracking’) des objets dépourvus d’états mentaux avec la poursuite des agents intentionnels. Pour étudier la poursuite située des agents, nous défendons ‘l’approche de la dépendance à l’égard de l’objet’ qui affirme que, dans le cas le plus rudimentaire, les humains poursuivent les agents intentionnels en poursuivant leur corps. Toutefois, nous soutenons que la poursuite des agents intentionnels comme agents intentionnels requière des capacités spécifiques pour détecter et comprendre le mouvement biologique et les états intentionnels. |
 |
|
|
A few remarks 
David NICOLAS
29 mars 2005 17:16 UT
A simple, deflationary thesis would go like this. Agents are a specific case of material objects: material objects with specific properties. Hence « agent files » are simply « object files » with additional information: information about properties specific to agents.
Bullot & Rysiew argue against theses of this kind by presenting three types of experimental evidence and an argument of computational efficiency:
1) « Humans can efficiently detect and track biological motions which are specific to agents. »
2) « Certain types of motion lead to the attribution of specific types of intentional states » (e.g. triangles perceived as being chased, attacked, caressed, etc.).
3) « The visual system does not rely on the same resources for recognizing non-face objects as opposed to faces. »
4) If agent files are tagged as agent files (as opposed to object files), then there is « no need to access the descriptive content of a file so as to target one fo the two kinds of entity for any attentional of motor routine. » It is therefore less costly to have files tagged as agent files than not.
They conclude that agent files are not merely object files with additional information. They are files marked as dedicated to the tracking of agents. But of course, these files include information about the agents' bodies, and the tracking of agents is often made by tracking their bodies.
I have a lot of sympathy for the authors' position. Here are a few questions and remarks.
Would they agree with the following, more general picture? Human beings have files about individual things. Any type of information can enter into such files. Indeed, « individual files » can be about many different types of individuals, like material objects, sounds, human beings, gods, electrons, etc. However, among these files, at least one type of file is distinguished because of its importance: files about agents are tagged as such for easier access and use. This kind of picture raises several questions: which types of files are distinguished for special use, why exactly, and what does this « distinguishing » amount to? For instance, it is conceivable that, instead of « agent files », there be files for conspecifcs on the one hand, and files for members of other species on the other.
It would therefore be interesting if the authors could bring to bear some evolutionary arguments in favor of their thesis. It does seem plausible that evolution selected specific mechanisms for dealing with agents, be they conspecifics or member of other species. Conspecifics will often be mates or competitors for sexual reproduction. While members of other species will often be preys or predators. So accurate monitoring of both conspecifics and members of other species is particularly important for survival. But can more detail be given about how exactly an evolutionary story would go? For instance, would it types of files that would get selected, and if so, which?
By the way, which organisms can be said to have « object files » but no « agent files »? Povinelli argues that much of primates' behavior can be explained with simple, non-intentional mechanisms. Would these non-intentional mechanisms involve only « object files »?
Nota bene: Some of the experimental evidence adduced above does not necessarily imply the existence of intentional information in the « agent files ». Thus, recognition of certain patterns as corresponding to human motion and recognition of human faces does not, by itself, require the existence of intentional information in the files.
|
| |
|
2 réponses à A few remarks:
|
| |
|
|
Reply (part 2) to question (1): The hypothesis of the Universal File for a Token (UFT)
Maria Rossi, 18 avr. 2005 10:41 UT
|
| |
|
|
Reply (part 1) to question (1): The hypothesis of the Universal File for a Token (UFT
Maria Rossi
18 avr. 2005 10:37 UT
Thank you very much, David, for this comment and these interesting questions. Let me focus on your first question about the following hypothesis: “Human beings have files about individual things. Any type of information can enter into such files. Indeed, ‘individual files’ can be about many different types of individuals, like material objects, sounds, human beings, gods, electrons, etc.” Let me name this hypothesis the ‘hypothesis of the Universal File for a Token (UFT)’, and formulate it in the following way:
UFT Hypothesis: Each human mind obtains information and singular knowledge about an individual (object, agent, event, sound, fictional entity etc.) because it uses as standard a Universal File for a Token (UFT) (as opposed to an object file, agent file, event file etc.).
Here are some sketchy and skeptical remarks about this hypothesis. The UFT hypothesis is a speculative but interesting version of the file concept/hypothesis, which is found in some conceptual analyses. It postulates a file which would be about any kind of unique entity – a sort of all-purpose and ecumenical uniqueness mechanism. (UFT might be akin to the theoretical construct of “ ‘chunks’ of short-term memory” in early cognitive psychology (Miller, 1956; Simon, 1975) or some uses of the notion of ‘pointer’ in the same contexts.) Intuitively, it seems plausible that one can entertain singular thoughts about many kinds of entity, so why not suppose that we use only such a kind of abstract file? Authors in semantics might even argue that this kind of ability is required by the linguistic processing of indexical and anaphoric terms.
However, as stated, this hypothesis is not so clear. If you admit and use UFT in your theory, you have to deal with a very (too?) abstract notion of ‘file’, which is difficult to assess and to map onto a particular skill or cognitive ability. In fact, the notion of UFT file seems to lack a clear-cut explanatory goal. If its explanatory goal is to give a unified and systematic account of how one keeps track of things in perception, memory, reasoning and conception (Perry, 2001: pp. 50-51) – it is in need of complement so as to make explicit the relationships between perception, memory, reasoning and conception. In this case, Perry (2001) seems right consider this notion as a sheer metaphor. If it just points toward a fundamental problem at the basis of your ability to obtain singular knowledge, it may not have explanatory power.
|
|
|
Agent Tracking- a too simplistic account?
(2 réponses)
Nivedita Gangopadhyay, 24 mars 2005 9:34 UT
|
|
|
|
Nota: les flèches jaunes ( ) indiquent de nouveaux messages mis en ligne depuis votre dernière visite.
|
|
