David,
the paper and an open lecture are here:
http://spacesyntax.tudelft.nl/media/Long%20papers%20I/alan%20penn.pdf
https://youtu.be/NkePRXxH9D4
Briefly, we need to be a bit careful about terminology, by which I mean that the word ‘agent’ in this case may be misleading, and so mustn’t be taken as a direct proxy for the philosophical concept. Although we didn't really understand this at the time, the basis of our EVAS ‘agent' based representation was in fact a deterministic analysis of spatial geometry (and topology for that matter) and as such extremely reductive. It happens to have the interesting property of incorporating anisotropy (a property of an embodied agent perhaps) in its geometric and topological representation. It is however entirely deterministic and thus an ‘analysis’ or the environment rather than any kind of ‘simulation’ of individuals with any kind of autonomous agency aside from randomness. In this sense it is just as holistic as any other space syntax representation.
EVAS stands for 'exosomatic visual agent simulation’ since the idea was that the representation in memory of the environment might lie outside the body in the environment itself (this builds on the space syntax notion that intelligibility might be a property of the environment itself). The thesis is that as society constructs cities through an essentially distributed process involving many different decision takers it builds into the geometry, topology and other aspects of the architecture - e.g. scale, land use, decoration etc. - patterns of relations between properties that make the world meaningful. This all sits ‘out there’ in the world and is readable by all the individual users of the environment in common. It is the distributed process of production of the built environment that leads to it being holistic and so amenable to syntax type analysis.
So one of the experiments we did was to place ‘merchandise’ in a series of environments. Different ‘flavours’ of merchandise were represented as a vectors on the surface of building surfaces in the environment. Agents were given a ’taste’ for a specific ‘flavour’ (the hunger was also represented as a vector). That way the agent could search the environment looking for merchandise that matched it’s taste. This allowed us to measure how long it took agents to find merchandise that best matched their taste under different spatial conditions. This is about as far as we have got so far on the specific ‘shopping’ activity work, but is obviously one reason that we developed the so called ‘agent’ approach in that it allows us to experiment with the twin notions of configuration and attraction as drivers of human behaviour.
A paper on this is here:
http://dx.doi.org/10.1057/palgrave.udi.9000120
I am not sure whether this would count as a ‘manifesto’, but certainly these are some of the motivations behind developing this methodology. What it is not, though, is something that would pass muster amongst social simulators as any kind of simulation in which the entities had any kind of ‘agency’. For that they would need beliefs, desires and intentions as well as memory and some ability to compute and so adapt behaviours depending upon their perceptions. In this sense it is really an extension of space syntax representation based analysis, but extending the range of kinds of things that we incorporate in the representation (eg. forward face vision, attractors in they environment).
Now, Bin often seems to come back with a criticism of these kind of developments in methodology. I find that hard to understand since what really matters is the empirical utility of a methodology - does it cast light on the phenomena we observe in the world? Angular segmental representations clearly do since they account empirically for observed human movement behaviours better than axial or ‘named streets’ although these seem not to have been subjected to the same level of testing against empirical movement data. I believe that the agent simulations also do. This is not to detract from the simplicity and elegance of the axial representation.
All the best,
Alan
> On 10 Dec 2016, at 19:18, David Seamon <[log in to unmask]> wrote:
>
> Alan,
>
> You mention your paper on the IKEA research: might you provide a reference?
>
> I understand your point about embodiment, though there is the phenomenological question of reductive activation in that the "agent" is limited in what its environmental and spatial "experiences" are. But I appreciate the direction you're attempting to move.
>
> One thing regarding the IKEA situation is that I don't see how the agent-based account can incorporate the merchandise displays along the sales route. The IKEA experience is more like a "stroll" than a trip with intention. I understand why some shoppers feel uncomfortable in an IKEA showroom space because one is quickly disoriented, and if one starts to try to figure out where he or she is, it becomes annoying. But my sense is (and I've only been to an IKEA store twice so keep that in mind) that most shoppers aren't too bothered because there is so much to see (and buy). So we move into a phenomenology of different movement and pathway experiences (stroll vs. constitutional vs. walk to work vs. walk through park etc.). In other words, not all movements are the same experientially.
>
> My larger confusion is what the "agent approach" has to do with space syntax, which to me is singular because it is topological and thereby holistic. The "agent perspective" is somewhat "global" in that the agent makes use of visual permeability. But it does seem to me that the emphasis remains more partial in that the amount of movement on a particular pathway, grounded in degree of integration in the pathway system, is somehow lost.
>
> Do you see the agent-based work as integral to space syntax theory, or should it be considered something different? If folks there are beginning to integrate it into space syntax theory, then it would be helpful to have a "manifesto" as to why. I'm not aware of any such discussion and would like to know of it, if has begun.
>
> David Seamon
>
> p.s. I apologize for starting new threads each time I post. For some reason, my "return email" for this list serve won't work. Reem is always kind to tell me "didn't go through" so I then post a new email. Thanks, Reem!
|