Dynamics of Information Osmosis and Place Perception


Let us imagine a number individuals who hold a socially conditioned opinion on a particular topic. Each individual's stance can vary in one dimension, as in the common agree > neutral > disagree scale. When assessing influences in this conditioning we consider two opposing quantities, which are transmitted through social contact. These quantities represent the degree to which he subscribes to opinions A and B on the topic in question. The difference between the quantities indicates the strength of the individual's opinion in the dominant direction. For instance, an individual firmly convinced that A is correct might have A=50, B=10.

When an individual communicates with another, his quantities A and B affect the corresponding quantities in the other individual. In the above example, his enthusiasm for A increases the other's enthusiasm for A, and his relative skepticism of B decreases B's appeal to the other individual.

If a fair number of such agents move about in a shared space and interact with each other upon meeting, we can expect strongly convinced agents to infect others with their opinion, and for opinions to be strenghened by interactions between like-minded agents and moderated by interactions between agents of opposing opinion. Such effects are intuitive from real life.

Population: 400, Area: 30x30

10 cycles

50 cycles

100 cycles

400 cycles

700 cycles

In the above example, two agents with opposing viewpoints (red and blue standing for dominant opinion A and B, respectively) are placed in a neutral population (grey, no dominant opinion). They quickly influence neutral agents around them, each interaction strengthening their conviction. This results in the division of the previously neutral population between the two camps in a spatially deliniated manner after 100 cycles. After this point, the dominant process is that of homogenization as one opinion gradually wears the other down to extinction (in this case, A (red) winning out over B (blue)).

Due to the self-reinforcing nature of this process, any initial advantage is likely to get amplified over time in a sort of constructive feedback loop. Complete homogenization can be avoided by altering the mechanism to undermine this feedback, for instance by giving more strength to novel opinion than to the confirmation of an already firmly held belief, which may also correspond more accurately to human mass behaviour.

In the next phase of our study, the space in which the agents interact will assume a more important role. In addition to being the stage for the unfolding of our simulations, it will also become the subject of the agents' communication, replacing the simple one-dimensional quantity we used previously. However, the competitive aspect of multiple opposing quantities will be removed. In this phase as in the first, agents move about in space. Through their journey and their interaction with others, they build knowledge of the space in which they exist. Each location carries a unique piece of site-specific information, which is acquired by passing through that location or by receiving the information from another agent through interaction. This is analogous to our own learning about the world both by visiting places and by hearing about them from others. During this process, each agent develops a unique psychogeographical map, divided into known and unknown areas. As each agent's knowledge grows, the variety of place information it can share with others increases, resulting in ever-faster expanding knowledge in each individual case as time passes. To illustrate more concretely, an interaction between agents early on can result in the exchange of place-information only from the immediate vicinity, as neither has had enough time to travel very far. As time passes and agent journey lengths approach the breadth of the entire space, information exchanged is more likely to be from almost anywhere.

Population: 700, Area: 50x50

200 cycles

600 cycles

1000 cycles

These effects are apparent in this graphical representation of a random agent's accumulation of place-information over time. Black stands for locations whose information the agent possesses, and white (or greyish) represents unknown areas. In the early stages (first image), its knowledge is confined almost exclusively to areas visited, with a spattering of close-by areas, visible as discontinuous dots around the main area of familiarity (the black blob). As time passes and the agent moves around, the blob grows linearly, but the the number of discontinuous dots grows at a faster rate under the previously discussed dynamic. For instance, if we examine the top-left quadrant, the area farthest from this agent's initial location and one it never visits, we see that the number of dots (places this agent has 'heard of') grows from 0 to 5 to 19 in these equal intervals, indicating exponential increase.

After outlining the basic dynamics of information gathering and osmosis in the context of a closed geography, we proceed to a synthesis of these concepts. The next demonstration has as its subject the spread and consolidation of place-opinion among a population of mobile agents.

As in the previous demonstration, each agent holds a unique psychogeographical picture of the 'world', which it develops as it travels through it and exchanges experiences with others. However, in addition to simply 'knowing' about a place, the agent also has an opinion about each place it is aware of. This opinion is based on its 'mood' (more about this later) during the time spent at the location or, in the case of places it has only 'heard of' from other agents and never visited, on those agents' opinions of it.

The wandering agents also maintain a state (akin to a 'mood') which is influenced by their subjective experience of places they move through. Their opinion of their current location depends on their preconception of it (if they had one, based on communication with others), their 'mood' going in to the location and the dominant opinion of the place by those who passed through it. This last quantity is maintained by each location. It is an agglomeration of the moods of all the passers-through, and can be thought of as analogous to the state of good repair of a place in the real world depending on people's care and affection. It is important to note that an agent's mood going into a place is a factor in the computation of its opinion of the place and its mood going out. This recursive process makes each place-opinion highly complex and effectively dependant on the entire life-path and accompanying opinions of the agent leading up to its entry into the location.

So, in this demonstration, the agents develop subjective psychogeographic conceptions of the closed space in which they move, based on their travel and interactions, and they affect places and are in turn affected by them, as they move through them.

The following snapshots track the development of two randomly selected agents in the population over time. The first column represents the changing map of place mood-agglomerations.

Population: 400, Area: 40x40

Dominant Impressions

50 cycles

200 cycles

600 cycles

1000 cycles
Agent 1

Agent 2

In the above images, the first column presents maps of dominant place-opinion agglomerations. Note that the distribution of this quantity becomes increasingly ordered over time, quickly consolidating into sharply deliniated regions of positive and negative opinion. As in our first demonstration, unless checks are built into the system (giving more strenth to novel opinion, putting a cap on how far in one direction an agent's mood can escalate, etc.) homogenization proceeds until the entire map becomes uniform. The key conclusion here is that the effects of place-opinion in the kind of system we've outlined compound over time, creating distinct 'good' and 'bad' areas.

Tracking the progress of the two agents we selected, we note their developing psychogeographical knowledge and its rough correspondence to the agglomeration map. We can also see that this is clearly not the determining factor and that residual moods from travel through different parts, as well as historical memory of places not recently visited, play a big role in place-perception. These deviations are spread wider by the process of agent communication, though we can see that the scattered opinion-points in areas of discontinuous knowledge (where the particular agent's opinion is based purely on 'heresay') describe fairly accurately the 'objective' (ie. agglomerated) situations in those regions.

So far, all the agents' movements have been chaotic and random. What happens when structure is forced on them? Does the system still manifest the phenomena we have noted? What effect does the system's structure have on the psychogeographical conceptions of individual agents?

In the following example, we slightly modify our previous system to include a kind of gravitational pull toward the centre of the map - agents' movements away from the centre are 'harder' (less likely to occur) the farther they are from it, creating a gradient of decreasing freedom of movement around the centre. This will artificially create a dense core of population in the centre, surrounded by a more sporadically populated periphery, not unlike in a city.

Population: 250, Area: 40x40

Population map

50 cycles

200 cycles

500 cycles

Agent 1

Agent 2

The previous images again represent the evolving world-conception of two randomly chosen agents among a sizeable population. The first column consists of population maps, the coloured dots representing agents, red or blue depending on their 'mood' at the moment in time. From the first column we can see that the same homogenization process quickly takes hold as before, probably accelerated by the densification of population in the centre.

From the agents' individual psycho-maps, we see that the newly introduced density gradient has had a profound effect on their world-conceptions. The structure of their knowledge is directly affected by the force we have simulated, as knowledge (and conformity of opinion) increases exponentially as we approach the centre. It is interesting to note that liberating, say, Agent 1, from this force, would do almost nothing to change the structure of its psychogeographical picture as most of the information is acquired through exchange with others. The overall forces guiding movement through space shape in an analogous way the 'minds' of the movers with respect to their environment.

Why is any of this relevant?

Firstly, examining the dynamics of these mechanisms can give us insight into many real-world phenomena, especially with regard to urban culture.

Secondly, and more importantly, the limited scenarios used here are entirely generalizable beyond the context of spatial information exchange. Information accumulated and spread by agents could be of any nature, and the dependence of their mental conceptions on outside forces acting on their 'movements through knowledge' would still hold in any environment where most information is received and not independently arrived at.