Writing from Within a Network

Author: Burcu Eke Rubini, Ph.D.

As a statistician working on network formation, I inevitably see life through a network lens. To me, everything is a network: the people I follow on LinkedIn, my WhatsApp conversations, international flows of exports and imports, the books I read, the laundry hanging from balconies. Because all of them, people, institutions, objects, can, at some level, be represented as nodes. And the relationships between them are edges. In proper graph theoretic terms: every entity is a vertex. Every relationship between them is an edge.

My interest in network science began during my doctoral years. But after some time, I realized that I always had been seeing the world this way. I think it is because I grew up in Istanbul. In a city that is half Europe, half Asia. In a city where you can cross between two continents with a fifteen-minute ferry ride. I am a child of the Anatolian side, “our side” that is. The place I called “the other side” is, naturally, the European side.

But “the other side” is always relative. While European side was the other side for me, Anatolian side is the other for those who grew up there. This suggests that the ties in this network are not symmetric; like all perceptual relationships, they are directed and must be modeled as such.

If we consider Istanbul’s neighborhoods, say, Caddebostan, Yıldız, and Beyoğlu, as nodes, the first reflex is to define edges as physical transportation: bridges, metro lines, ferry routes. But this simplifies Istanbul into a spatial network. Istanbul is not a single-layer network. Physical transportation is only one layer of the system, perhaps even the one with the lowest weight. Alongside it, there exist equally powerful social ties, economic ties, historical ties, and ties of the heart. A ferry ride and a neighborhood bond do not carry the same weight. Istanbul is a multilayer network. The same nodes are connected through multiple types of relations across different edge sets.

Perhaps the Bosphorus is not merely a geographic divide. If one were to apply a community detection method to Istanbul’s social tie structure, it would not be surprising to observe two dominant subgraphs, divided by the Bosphorus, characterized by high internal connectivity and relatively lower external connectivity. The issue is not geography itself, but the distribution of ties.

Istanbul’s degree distribution is not homogeneous either. It may not be fully scale-free, but it is skewed. A small number of nodes have high degree; like universities, financial centers, media institutions. Influence is not evenly distributed.

This is where eigenvector centrality becomes relevant. It is not enough to be well-connected; it also matters to whom one is connected.

Yet all these measures rely on observable edges. And some of Istanbul’s strongest ties are not observable. They are latent. They come from the heart; their presence settles within you.

This makes Istanbul’s latent space highly multidimensional. A two-dimensional Euclidean embedding is often insufficient. Social distance differs from physical distance; cultural proximity does not necessarily align with geographic proximity.

To write from within a network is to feel its parameters from the inside.

Istanbul has a high clustering coefficient. At the same time, it contains structural holes. Homophily is strong; the assortativity coefficient is positive.

And perhaps the real issue is not density.

It is possible to model Istanbul.

But not with a single measure.

Because some edges cannot be drawn.
Some are memory.
Some are born of love.
Some are simply the act of sharing the same city’s sound.

Perhaps not everything is a network.
Perhaps I only want to see everything that way.