MaCro Philõsophy
04-February-2018

Representative Reasoning

ML4E6C5 flitted towards the outskirts of Earth's diffusion field. They weren't really looking forward to it - it was more a kind of due diligence than anything else. For every ten thousand or so surveys of a 99.9997 or a 99.9998, it was standard practice to revisit at least one Niner (the term given to lifeforms with a 99.9999999%+ chance of extinction before reaching the unbounded proliferation phase). These lifeforms would invariably either run out of resources or, more commonly, use their resources to wipe themselves out. Earth was, at least, a little interesting, in that it was not yet obvious which of these it would be.

ML4 had been assigned Earth because it presented one of the few cases with an unexplained mechanism for intelligence. Earth creatures used what looked like a standard multi-perspective aggregation approach, but something was wrong with the aggregation part. They would invariably reach the wrong conclusions, even in simple cases. It would be trivial for ML4 to find out the fault with only a minor infraction of the non-interference policy, but there was no need. It was nice for there to be some things left to solve.

In cases such as this, ML4 had taken to recording a single unremarkable event. The idea was that there was little benefit in adding yet another population defining events - the corpora already had enough of them. In this case, ML4 chose to record a non-event on a bus.

Report 1AA5DA9481DD1DDC: A non-event on a bus.

Method: Reasons report with streams clustered by origin location (.3) and frequency (.7).

  • v0: That person looks like they're coming towards my seat.
  • v1: They're probably not coming towards this seat. People normally aren't.
  • v2: Are these windows made of glass?
  • v3: The last person that got on the bus didn't come towards this seat.
  • v4: What if the interviewer asks me something I don't know?
  • v5: They're going to sit next to me.
  • v4: I should have prepared more about my previous job.
  • v6: They should sit in the empty seats near the front.
  • v0: Why didn't I prepare better for the interview?
  • v1: They probably didn't see the empty seats.
  • v7: They're not interested in us.
  • v0: Are these windows made of glass?
  • v2: There could be a picture on the side of the bus of a body and I am the head.
  • v3: They chose the seat next to me.
  • v0: They're sitting next to me to spite me.
  • v4: What if I use a word wrong in the interview but I think I'm using it right?
  • v3: I received a tap.
  • v5: Did they just hit me as they sat down?
  • v1: They probably didn't even notice.
  • v6: They should have said sorry.
  • v4: I should check the definitions of every word I know
  • v0: What does compunctious mean?
  • v6: I should say sorry to Rebecca.
  • v3: The tap was incredibly minor.
  • v5: I might have already missed the bus stop.
  • v6: Perhaps they have an interview too.
  • v7: They don't care that they hit me.
  • v4: Are they a spy for the interview?
  • v0: It was rude of them to hit me.
  • v6: It was just their bag that brushed me.
  • v1: They probably didn't notice it brushed me.

Timespan: 3s

Anomalous Result: Narrative about 'rude' behaviour continues through the day leading to reduced efficiency on future tasks.

Conclusion: NONE

Another garbled mess with only a probabilistic link from reasoning to future behaviour. ML4 submitted the report and turned to their next destination as the database began the automatic process of aligning it with every existing permutation of entries. The next system would be a good one. It contained a lifeform of vibrational energy that had just learnt how to expand by resonating in sync with its surroundings. It was a 99.9997. About as low as it got. A reward for cycling a Niner database perhaps.

About halfway between Earth and their new assignment, ML4 first considered the hypothesis that would later be assumed to explain human behaviour. What if they didn't just preference a single stream (v0 in this case), but actually allowed it almost complete control? It was easy to see how this hypothesis could have been missed. Who would have considered a possibility so obviously unworkable? After ML4 ran the hypothesis against a few test cases sampled from the Earth corpora, they were satisfied with their theory.

At an early stage of development humans would elect a single stream as representative. This was the kind of energy-saving, simplifying approach that could only be taken by apex predators with abundant resources, but that would normally quickly die out. During development, humans come to identify with the stream and it becomes a single voice, tasked with diligently aggregating the opinions of all the others. The voice would start out well, giving a useful overview for quick decision making. However, over time it would fall prey to a negative feedback loop. The voice would slowly develop its own idiosyncrasies, becoming more influenced by its previous utterances than all those streams surrounding it. The identification was so strong that the voice itself would barely hear the others. Human behaviour wasn't so much irrational, just incredibly poorly designed from the perspective of long-term survival of the species.

ML4 couldn't believe they were the first to notice this. A main tenet that pushed ML4's kin to become pure observers, forbidden from interaction, and especially creation, was:

Whatever we create, there is always a part of us in it, whether we realise it or not.
The humans had unwittingly modelled some of their political systems on the same principles that their minds functioned. They called it representative democracy. In their political system they wisely added time limits and other regulations to prevent one voice from taking over for long enough to become corrupted. However, even so, the cracks were beginning to show.

ML4 submitted a new report containing their hypothesis. As it sent they double checked to make sure their ID had printed correctly. There was no way it couldn't have, but they were quite proud of this one. The hypothesis was incorporated into the database and it was quickly established as the dominant explanation. The Earth system was reclassified. It was still a Niner - in fact, there were now even more nines - but now it was also marked as uninteresting, with all open questions resolved.