>>If the human is part of the system, then it would use its observe() method. :)
>...
>
>My question is, if we construct the human not out of an observer class, not as a single particle, but as a bunch of particles not bound together by anything besides the physical forces, then there is no object reference to the human observer anywhere.
>
>There's not even a class definition for it.
>...
>
>... Somehow, the observer needs to produce a measurement.
>...
>So the photon particle is how our observer receives information.

I'd say, let's ignore the notion of human for now. We agree that the carrier of information is a (photon) particle, by means of which the observer receives information. But, obviously, the information must be recorded first. Thus a particle's state is recorded as values of its own object properties, which means that, in an absolute sense, each particle is a carrier of information.

Back to the observer: As you said, we don't have a class definition for the observer, but we know we'll need one at some point.
An observer records events, which are basically moments in time. So, our observer class would need a table of relevant events. But, some events may be relevant to a process, other events to another process and so on, and that makes it impossible for us to define the absolute observer because we can't predict all relevant events for all processes.
So, what I am saying here is that the notion of observer is directly connected with the notion of event.

Therefore, an event may be recorded only by the entities that are "present" at the event:). As of now, the only entities capable of recording the event are the particles. Thus, during every event, each particle would record the event ID (name, time, etc) in its own table of events, and its state in its own history table.

Later on, when and if, we need to analize a series of events, we'd just have to query the history of each particle for the events that are relevant to our analized process.
I hope it makes sense.