Strange Native
Strange Native

Next Idea

07. Friendship Is Emergent

Amateur Thoughts On A Working Theory For Friendship

Sep 04, 2011

Amateur Thoughts On A Working Theory For Friendship

Today I starting thinking about the binary status of social networking.1 The idea of distinctly marking someone as a “friend” or “not a friend” has seemed too simplistic and dissonant to me (and to many others) when compared with our real-world experiences making, maintaining, and losing these friendships. This got me thinking about what friendship is and where it comes from. I ended up with the following framework, and I’m curious how it might affect the way we build social networking systems.

1. Why so serious? I was inspired to think scientifically about this after watching Geoffrey West’s TED Talk on The Surprising Math of Cities. I think with a serious theory of friendship, we could potentially create more fluid, responsive social networks models.

The Friendship Theory

Friendship is an emergent property of transactions between two people. Bio-costs (such as time, stress, thought and energy) are the currency of “friendly” transactions, and the frequency of these transactions determines the strength of a friendship. A close romantic relationship where transactions occur constantly between individuals would score very high, while two acquaintances with few, short interactions would score very low.

Transactional balance is a key component to building friendship. Without a balance of bio-costs, the transaction rate is likely to slow. For friendship to emerge, transactions must reach a certain level of frequency (a “friendship waterline” if you will) for a sustained period (likely there is a “minimum viable period”) and the balance of these transactions must finish near zero. For friendship to be sustained, transaction frequency must reach periodic peaks equal to or greater than the friendship waterline.2 2. There is likely a “gray area” around the friendship waterline where, depending on the mood or last interaction with an acquaintance, a subject might respond positively or negatively to the question of whether said acquaintance is a “friend”. It would follow that the “Friendship Score” is simply a running average of transaction rate over time.

It seems important to clarify: I’m defining “friendship” distinctly from “affinity” or “affection”. The former being the emergent property of a sustained transaction level over time, while the latter two are powerful emotional qualities which can be surmised, but not entirely understood or predicted by a Friendship Score. We may dearly love someone we haven’t spoken to in years. Yet, perhaps a sustained period of high transaction rates balancing to near-zero can sustain affection over a long transaction drought while still adhering to this theory of friendship.3 3. There are a number of people for which I have a high level of affection, but haven’t spoken to in years. However each one of these relationships began with a sustained period of intense transaction.

This model seems flawed in one particular case: that of the externally coerced, mutually disaffectionate relationship. Let’s say for example there are two business partners who share a disaffection for one another, yet are forced to interact on a daily, even hourly basis. Their transaction rate would remain high, while their balance sheet would remain near zero (remember, they each share equal amounts of stress in dealing with the other). Based on the above model for the Friendship Score, this might read like a strong friendship. Affinity seems to be the important third factor of our equation, but how to determine, quantify and include it?4 4. Any ideas?

While I’m speaking of friendship here, there may be other emergent social properties which can be described by distinct patterns in bio-cost transactions. I’d love to see any research done in this area.


  1. Again Russ, this is really original and sophisticated thinking. Your posts are pure quality!

  2. On the note of affinity — it seems like the best way to determine it would be to examine the content of interactions. Of course, there are massive privacy issues involved, but I can imagine something similar to how Google recommends advertisements based on the contents of emails without ever (purportedly) using the data elsewhere.
    If the content of interactions — from things like word frequencies/tones in text communication to … smiles (perhaps? a bit of a stretch) in photographs — could be extracted and analyzed in situ without using the data for anything else, it could be useful for determining affinity.

    • And as for other things that are emergent of bio-costs, one of the ideas I’ve been thinking about is work. The current system of paying people on a per-hour rate seems to be antiquated, especially for more intellectually oriented jobs (e.g. programming, designing, etc.), but for others as well. Looking at factors other than time, the more abstract things like “thought” and “energy” could play some sort of role.
      After all, productivity is *not* proportional to time spent, but at the same time it’s not easily defined by its constituents (read: it seems emergent of “bio-costs”). Not sure where that leads, though.