date  thursday june 25th

place meetiing room, 5th floor, DNC - house, nørrebrogade 44, build. 10G

to register please email steenberg@cfin.dk. participation is free of charge

programme 

13:00: Don Braxton: Infinite Possibilities in Multi-agent Simulation - Attending to Complexity but Seeking Simplicity

13:40: Ivan Tchalakov: Simulating Network Dynamics: A Stacked Networks Approach

Coffee

14:40: Kristoffer Nielbo: Can dogs make a stone breathe? - modeling transfers in counter intuitive concept acquisition

15:20: General Discussion

15:40: Small reception

Abstracts:

Don Braxton: “Infinite Possibilities in Multi-agent Simulation  - Attending to Complexity But Seeking Simplicity”

Many strategies for simulating cultural dynamics are possible using multi-agent software.  I will introduce two types of multi-agent simulations where basic choices in modeling strategy are illustrated:  an epidemiological model of the spread of cultural variants; and a sociological model where relations constrain cultural dynamics.  Each approach makes certain kinds of investigations possible but also blinds the model to other salient features.  This illustrates a crucial insight: Models are always driven by what questions one seeks to ask.  In general, cultural dynamics are always more complex than any given model, but simplicity is a virtue in identifying and understanding exact operations of specific mechanisms.  No apologies are really necessary so long as the modeler understands what a simulation can and cannot do.

Modeling_Ritual_Distributions.pdf

Ivan Tchalakov & Petar Kopanov: “Simulating Network Dynamics: A Stacked Networks Approach”

We are presenting an approach in modeling the (social) networks dynamics that incorporates the basic assumptions of actor-network theory. By critically juxtaposing it with the views on social actors as ‘governed by particular rules’, the authors move away from the notion of ‘guidance’ in favour of the idea of opportunistic evolution of a given system / form of life. We adopt ‘ex post’ stand to the processes going on in a specific form of life considering the properties of the studied entities (agencies, mediators, etc.) as an outcome of the evolution and not as pre-given attributes. Similarly, notions such as the ‘reflection’ or ‘rationality’ are approached as actors’ (including researchers’) mode of accounting of the processes they are modelling, hence the desired mathematical models is be an outcome of specially designed and collected sociological (using both qualitative and quantitative methods) data in concrete areas – the models are to be build built on these studies and in a way are verifying them.

To analyze the network dynamics we depart from the idea of so-called ‘stacked’ or ‘layered’ networks. The stacked networks are formally independent / autonomous networks sharing some common entities, which are considered as actors and/or mediators, acting in each of these networks under different ‘identities’. Hence the common entities also form actor-networks, if in a different way. We assume stacked networks differ in their stability and that there exist patterns of ‘tuning in’ between them that channel the way they influence each other. These patterns of ‘tuning in’ in turn bear on the ways different identities of the same entities, engaged in the stacked actor-networks, go together.  

The presentation outlines our current efforts to elaborate a simplified mathematical model on stacked networks dynamics as first step in designing software that will enables us to simulate it. We provide the basic axioms of the model and our approach in solving some of the key problems, such as structural and functional complexity of the modelled phenomena, and mathematical approaches in their evaluation that will allow us to represent the interactions between studied entities as such between non-determined stochastic automata. 

What do you think a simulation is.pdf

Language and Perception Tchalakov Graz 2004 Yearbook.pdf

Kristoffer Nielbo: “Can dogs made of stone breathe? – modeling transfers in counter intuitive concept acquisition”

In cognitive approaches to religious phenomena the theory concerning counter intuitive concepts, as proposed by Pascal Boyer, has been very influential. Religious concepts are counter intuitive because they violate our intuitive expectations associated with more or less innate domain-specific knowledge, e.g. a living dog that is made of stone. The central claim is that counter intuitive concepts have a transmission and memory advantage in cultural selection because the violation makes the concepts salient, while it is still possible to sustain predictions from their respective domain - the stone dog still move and breathe like a dog. Some concepts can be too counter intuitive though which makes them impossible to remember, because the cognitive load is simply too high, e.g. a dog that is made of stone on Mondays, speaks every second week and sometimes is invisible. In other words there seems to be some constraints on counter-intuitiveness, which when breached result in a conflict between conceptual categories and an eradication of memory. With the aid of a simple feedforward neural network modeling concept acquisition, originally proposed by David E. Rumelhart, I will try to illustrate how transfers between concepts, while creating counter intuitive concepts, can result in category conflicts and concept eradication. Furthermore I will stress that concepts have to be constrained by a context. And finally being a bit wary about certain so-called theory theory and nativist assumptions inherent in Pascal Boyer’s cognitive framework, I will moderate such assumptions and point to the relevance of learning through patterns of covariation.

A topological approach to cultural dynamics