[News]

News: Welcome to The Oktave Forum.

[sri]

sri

sri

Show sri's last posts.

Show general stats for sri.

• 
• Administrator
• Sr. Member
• Karma: 7
• Posts: 264

(cross posted on the algorithmic worldview..)

(posted here as multi-agent systems are intrinsically about optimization)

One of the common patterns we see year after year is that whenever a new group of students arrive, they quickly form groups among themselves. No, the groups are not based on caste (necessary clarification for readers from the West reading about Indian society :-); typically these groups are based on language. There are Telugu groups, Kannada groups, Hindi groups, Tamil groups, etc.

So, is this an evidence of intolerance in our society? Are we averse to diversity and seek to mingle more with our own kind?

On the contrary, formation of cultural groups may actually indicate more tolerance than less! And we can prove* illustrate this without going into socio-psychological studies.

The party game sample from NetLogo illustrates this further. Consider that we have a party with people belonging to two groups A and B. Each individual has a certain threshold called "tolerance" to members of the other group.

Initially clusters are formed randomly. Now each member decides to stay in the cluster or move to another cluster depending on the number of people from the other group in his/her cluster and his/her tolerance level. When a member changes clusters, the demographics change and creates a chain reaction.

Such a simple model can elucidate very surprising results. When tolerance levels are high, the system quickly settles down to a set of clusters, where each cluster w.h.p (with high probability) will be homogeneous! That is, comprising people from the same group.

When tolerance levels are low, the system never stabilizes and people are continuously changing clusters. But at any instance of time, the demographic distribution of any given cluster is equally likely to contain members from either group!

We are needlessly worried about groupism! All the cynics can find something else to moan about. There is nothing fundamentally wrong with the current generation.. :-)

* Somebody should have caught me on this "proof" thing

[sanket]

sanket

sanket

Show sanket's last posts.

Show general stats for sanket.

• 
• Global Moderator
• Full Member
• Karma: 4
• Posts: 104

NetLogo looks like a nice tool. I also agree with the ideas in this post. There is one point that I want to make. These are rough ideas; not systematically thought out ones.

Although, systems seem to reach a stability with a set of clusters formed, the stability is not a strong one. To continue with the example of the party game (or even that of students forming language based clusters), we can observe preferential attachment (assortativity) and detachment (disassortativity) in work periodically. In a party, let's say someone is dishing out interesting gossip, the size of the group increases, and the size of the group attracts more people, who break away from other clusters to join this one. However, after a while people get disenchanted and return to their former clusters or form new ones. Similarly, in case of the language based groups, there is a threshold up to which a group grows. Once the threshold is crosses, people start defecting and form new groups, even though they may be speaking the same language.

[sri]

sri

sri

Show sri's last posts.

Show general stats for sri.

• 
• Administrator
• Sr. Member
• Karma: 7
• Posts: 264

Yes of course, real human systems are much more complicated than the party game. But capturing some intrinsic variables helps understand larger patterns easily.

The party game models tolerance, but to model the kind of fluidity that you describe, we should perhaps have another variable called (say) boredom, where boredom increases with too much homogeneity or something like that. It would be interesting to see what patterns emerge.

The dynamics of multi-agent systems are so counter intuitive; I think there is a lot of uncharted territory in this field.

[sri]

sri

sri

Show sri's last posts.

Show general stats for sri.

• 
• Administrator
• Sr. Member
• Karma: 7
• Posts: 264

Er.. one clarification is necessary.

In a system where tolerance is less, the characteristic is flux -- constantly changing groups. If tolerance levels are very high (say above 95%), the system hardly changes from its initial random configuration comprising of heterogeneous groups.

Hence homogeneous, stable groups are characteristic of systems where tolerance is high, but not very high. So if you were in a stable group of people from heterogeneous backgrounds, don't fret. It does not indicate intolerance; rather very high tolerance.