|||

Instructions for Humans - Day 4

Saturday, the forth day of the exhibition, happened. People visited. I talked to them. I was very tired so not much was retained. This was a problem. But it was otherwise a good day.

The main thread that seems to be emerging is confirmation bias and reality tunnels. Specifically how our perception of (what we assume is an) objective reality is filtered by our inability to comprehend the immensity of objective reality. Selective sampling of the world is not necessarily a problem (photography is entirely about selecting what to capture so we might say something profound about the world) we are surprisingly unaware of how subjective our opinions are.

This ties nicely into issues of bias in machine learning AI systems, and the belief that by adding more raw data will mitigate bias problems. We know reality can be broken down into data points, so the goal is simply to gather enough of them, rather like Bruce Sterling's camera-of-the-future, brought to my attention by James George. He explains it here:

"It would record all the photons in a room, then turn cinematography into a computational problem of selecting scenes and camera angles from an infinite data set."

Obviously it's a thought experiment as recording all the photons in a room is impossible, so like all good impossible things it tells us something about what is possible, that all our samplings, be they psychological or technological, are partial and therefore biased. The lesson seems to be, embrace the bias.

In the first week I was explaining my ideas to people. For the second week, to give my sore throat a rest, I'll be asking people to explain what they see surveillance and data profiling as, to get me out of my bubble.