This is becoming too broad a discussion, from my point of view. My view in a nutshell is

• cognitive agents are necessary but insufficient to account for societal phenoemna: what else is needed? Theories of social dynamics, for which cognitive architecture is irrelevant.
• a script-like behavior like going to a restaurant is a convention-based rather than norm-ruled behavior ,
• conventions are behavioral regularities, more or less as Lewis defined them: i.e. preferences which increase in presence of others’ preferences for a given action. Norms are instead deontic commands, either emanated by personal authorities (Von Wright), or emerging spontaneously and gradually
• a norm emanated by a personal authority is a legal norm; a norm emerging spontaneously is a social norm
• communities of practice usually generate conventions, and sometimes social norms
• agents subjects to a norm do not emanate them, but they simply allow them to emerge
• goals are essential to model autonomous agents, including going to restaurant, but they are not intentions (in Bratman’s sense)
• theories like Newell and Simon’s, or Miller et al.’’s, are not theories of human cognition, but address that zone of the scientific noosphere which is populated by abstract intelligent systems, and this is exactly what those theories meant to do. Whether they apply to human cognition is not clear at all.

Sorry for the schematic answer :-)
ross

Il 19-01-2007 7:57, "Maarten Sierhuis" <[log in to unmask]> ha scritto:

I must say, this is turning into an interesting debate. Let me at least state that my in intent is to come to common understanding and not to attack those who might have a different view on things.


Bellow, I am trying to make some comments ... again ;-)


Doei ... MXS

On Jan 18, 2007, at 1:09 AM, Rosaria Conte wrote:

 

 But, then, what does the statement mean? Can you give some examples of the use of cognitive agents that are not based on a cognitive architecture?
 
 

There is a misunderstanding here:  I said that a theory of norm emergence based on cognitive agents does not imply that a cognitive model (and the underlying architecture) is a model of large scale societal behavior: in other words, although cognitive agents contribute to societal processes, the latter should not be modeled as cognitively designed (a great deal of social dynamics is extra-mental).

Ok, I think I get it. Just to clarify ... so you are talking about wanting to model large scale societal behavior and not individual problem solving. And, to model this large scale societal behavior you want to model the society as individual cognitive agents, using a cognitive architecture that is based on goal-driven planning agents. From this simulation you now want to observe the large scale behavior and, probably, categorize this behavior into norms. This will then lead to a theory of norm emergence in large societies.

You say that. "it does not imply that the cognitive model [of the individual agents?] is a model of large scale societal behavior:" Here is where I get stuck. Am I correct in restating your claim as; Doing this does not mean that your theory is based on any cognitive theory. Or is this not what you claim?  Let me argue against this claim (assuming that is your claim for a moment): If the large scale behavior in your model is generated based on the behavior of individuals in the model, than the large scale behavior IS created by the individual agents. Then, if the behavior of the individuals is generated by a cognitive model for each agent, is it not true that the large scale behavior is also generated by the cognitive models? Ergo, the theory you create based on this behavior is also based on the cognitive theory that underlies the cognitive models of the individual agents. Thus, if you use a goal-driven planning theory (such as the BDI theory of Bratman), than, your norm emergence theory is based on Bratman's BDI theory. Not?

 

My personal view is that norms and practice are closely related.
 
 

This is a conventionalist view of norms. In any attempt to distinguish them from conventions, norms are accounted as something rather different from practices.

I have not done any research on norms or norm emergence, so forgive me if I am way off. However, I have read some of the work of Geert and Gert Jan Hofstede on Cultures and Organizations. I don't recall them talking about norms specifically, but they do talk about values and practices. Here is an interesting quote from them, that I was implicitly using in my debate:

"Symbols, heroes, and rituals have been subsumed under the term practices. As such, they are visible to an outside observer; their cultural meaning, however, is invisible and lies precisely and only in the way these practices are interpreted by the insider ... The core of culture ... is formed by values."

One of my questions is if Hofstede's rituals are practice activities. I think so, but that's beside the point.

Question: How are norms the same or different from values?

If they are similar, or at least related, then I would still claim that norms are related to practices.

Here is an example with which I can try to explain this relationship, imho:

- Practices are like norm-based scripts that everyone within the community of practice knows and uses to behave in a similar way.

- Take for example, the script of "Eating at a resaurant."

- There is a norm-based script (practice) of how to behave; For the customer, the waiter, the cook, the owner of the restaurant, etc.

- This script is different depending on where you are from, i.e. your community of practice and thus ... your norms.

- For example, In Holland, when you go to a restaurant, you walk in and sit down at any empty table you might like to sit at. Then, the waiter will come over and asks you what you would like. In the US, this is not how you do things normally, although in some restaurants you may, which confuses the hell out of most Americans. The norm is that you wait to be seated by a hostess. To emphasize this norm, most restaurants have a symbol, namely a sign that says "wait to be seated."

This is how I relate norms to practices and societies (i.e. communities of practice), and I would say that the norms in a community are generated by the practices of that community. The norm *is* implemented by the script, or what I like to call, the activities of the individuals in the community.

Note that scripts are what activities are about. There is NO need for goals or goal-driven planning; Possibly only when we talk about a problem solving activity, but there is no problem to be solved to go eat at a restaurant, and thus there is no need for goals, and in fact, the concept of a goal is simply confusing and not a good representation of how we go eat at a restaurant. In that sense, I agree with you that the norm of eating at a restaurant has no relation to any goal-based cognitive theory. But, that also means, don't use such a theory to implement the activity of having an agent eat at a restaurant. Use an activity-based theory, that talks about, what you call, "extra mental" behavior. Activity Theory by Vygotsky, and Situated Action by Suchman, as well as Cognition in Practice by Lave are more suited in this case, then Bratman, Newell or Anderson's theory of cognition.

 

How can it be that entities reason upon representations of norms, but don't issue or understand these norms?
 
 

Very simple. Lets start from legal norms:

What are legal norms? This suggests there is a classification of norms.

agents need to form a mental representation of the  norms impinging on them in order to autonomously decide whether to comply with them or not.

Ok ... so that's what I would call that they need to "understand" norms. In terms of my example of the "eating at a restaurant" activity, agents will understand this by being able to perform the activity in the appropriate situation, i.e. when going to a restaurant. Belonging to a group (a community of practice) makes you "inherit" the activity and thus the norm.

However, agents usually emanate no norm:

You mean, agents don't create norms? How about fads? They are started by someone. I wouldn't call Malcolm Gladwell's The Tipping Point a scientific treatise, but he does tell a pretty good story about certain fads that are created by individuals as tipping points that create a fad.

often, they don’t even have the faintest idea how norms are issued, by whom and through which processes. The same a fortiori is true for social norms.

Who is the "they?" The agents that don't emanate norms? I guess this is true in some circumstances, but not true in others. For example, I think most of the teenage girls in California can tell you exactly who created and how the fad of wearing expensive purses and getting little dogs came about; Paris Hilton, via television and smut magazines :-(

 

 I am not aware of any cognitive agents that are not, in some way shape or form, based on a theory of human cognition.
 
 

GPS is no theory of human cognition but it is meant to be a General  theory of intelligent, planning systems.

Wait a minute. Newel's last book was even called Unified Theories of Cognition.

Miller, Galanter and Pribram’s work (from the early 60s) is a general theory of intelligent, goal-driven systems which has poor relation with any pre-existing theory of human cognition.

Not according to Simon and Newell. Moreover, the tools that are being used for modeling and simulating "intelligent" agents are almost all based on goal-driven systems (Brahms is one of the, if not the, exception).

 

It is probably my lack of knowledge about those who define a theory of cognition for animals, other than human, that I can't say much about this. But, I would say that the theory of artificial minds is very much related to that of the theory of human cognition.
 
 

Perhaps  unfortunately,, this is not the case. The hegemonial, although no more so recent, trend in cognitive psychology is the modular view of the mind, that so far to my knowledge no-one took as a reference for computational modeling and agent architecture. 
As to symbolic theories of animal intelligence, in cognitive primatology, lots of people (Tommasello, Visalberghi etc.) are working on primates’ and children’s social imitation and cooperation with experimental and observational approaches, trying to understand related cognitive determinants of these behaviors.
 
 Cheers
 ross
 
 
 

 
 
 

 

  (...)
  More so, many scientist (e.g. neuroscience, anthropology, cognitive science) have in recent years developed counter theories to the theory of the human mind as a "symbolic copy machine."
  
  
 

Although it is not entirely clear to me what a symbolic copy machine is, I do believe instead that cognitive science in general has no much to say against the theory of human mind as a symbolic system.
 

 To claim that cognition is based on symbolic processing, it means that that there is a copy function within the process, and symbolic structures are copied from one place to another in order to store and recall the symbolic structures.
 
 

However, this by no means implies a particular commitment to a view of agents as necessarily conscious, ratiomorphic, and deliberative.
 

 Yes it does, at least deliberative, which I would posit needs consciousness. I am not sure what ratiomorphic is.
 
 

A cognitive (based upon symbolic representations) view of the mind should not be equalized with a strictly deliberative view of agenthood.
 

 Maybe not in the field where you operate, but I would claim that in philosophy and cognitive psychology it does. Maybe you can give some examples that make your claim explicit.
 
 

 
  
 

... (but, alas, not every human activity is goal-driven).
  
  
 

Of course. However, a cognitive theory of goals defines them as symbolic internal representations triggering and guiding actions; by no means, again, this implies that goals are also attributed the property of being rational, consistent, conscious and necessarily chosen for action (and therefore planned).
 

 But that is not what the goal-based theories say. More importantly, if one uses a BDI agent architecture (or an expert system based architecture, such as Jess) to model reasoning in your agents, then you are either implicitly or explicitly claiming that "goals are also attributed the property of being rational ..." Simply because these architectures are based on the theory that rational, consistent, conscious choosing of actions is planned and goal-based. In other words, imho, you cannot use these architectures to implement your agent system and then claim that your model does not rely on these theories. That is why we developed our own BDI-like architecture that is not based on these theories, but on theories of situated action and activity theory, which do not use the concept of a goal to model reasoning, and does not use goal-based planning to simulate perception-action and deliberation.
 
 

  Cheers
  ross