I've recently been learning more about machine learning, and I had a realization late at night that I thought was worth sharing. It's an example of how studying computation can be useful for developing ways of thinking outside of computation.

The seed for this realization was planted when I watched a talk about Tesla's machine learning stack on Youtube. The presenter said that there were three important qualities for their training data: 1) There needs to be a lot of it 2) There needs to be a lot of variety in it 3) It needs to be real, not simulated.

On a run, I was thinking about how after a few years of working with Joe Shouldice at littleBits, I felt like I had a little mini Joe who lived in my head, who would preemptively chime in on projects before I brought materials for review. It struck me that the three criteria for Tesla's self-driving car data also applied to my experience working with Joe. I had a lot of data; we worked together for a few years. We had a lot of variety; I had worked with Joe on fun projects, and not-so-fun projects - tight deadlines and relaxed deadlines; graphic design projects, video projects, socks, signs, etc. Lastly, the experiences were "real".

This gave me a new view on interactions I've had with others. For one, social events at work suddenly made a bit more sense, they are a great way to provide varied and real experiences for colleagues to interact. However, once that interaction feels forced, it's no longer "real" so it's no longer useful as a way to get to know someone.

I think that when I'm having conflict in a relationship, because I don't understand where someone is coming from, I can use this as a lens to debug my mental model of the other person. If I haven't known them for a long time, I can just remind myself that a lot of data is necessary, which makes it easy for me to be patient with them. If I've only interacted with them in one context, I can remind myself I need varied data, and try to get to know them better as a person. Lastly, I can remind myself that only real data is helpful. Imagining hypothetical interactions with someone, or telling myself stories about another person's motives is not actual helpful to creating a representative mental model of who the other person is.

This can sound like a really lame and robotic way to think about others. But in the end, it's a way to help myself stay patient, curious about others, and empathetic. These are ways of being that I already want to be, but this is a helpful reminder sometimes. In general, I lack many frameworks for troubleshooting interpersonal conflict. I don't think this is something to live by. More just an interesting way to reasoning about relationships and mental models of other people. I was pleased that all three pillars of the model mapped so clearly to my experiences. There is probably some "it makes sense to me easily; therefore it must be true" cognitive bias at play.