Hi Xin Wei,
The fact that we can't say that the system is linear does not mean that there is no space for intentionally introducing non-linearities into the system that have a qualitative effect on the perceived output.
As you know, the emphasis on linearity in electrical engineering is there so that we can be precise about what kinds of non-linearities we are introducing. It's about minimizing unknowns, not eliminating them.
When I say system, I mean the metal stuff. The metal stuff has a linear response according to the datasheet. That means that even though g is non-linear, I know that any non-linearities in f are designed.
My argument is only that if we use analog components, there is no possibility to design the behavior of f. Instead, f is effectively linear, by which I mean, it's roughly a straight line and we just have to deal with it as it is (we are at the whim of the non-linearity of the measurement, or the lack of meaningful correspondence between the measurement and the salient 'parameter', which is whether the plants should be watered or not and how much).
Of course, when we consider the whole entangled system, the LED will not vary linearly with the salient 'parameter' even if the system is "linear". But since we're going into the design without any knowledge of any of the functions except for f, it's wise to make sure that f is manipulable so that we can adapt it to the situation. In this case it's actually less labor intensive and possibly less energy/material intensive to design the system such that f is manipulable, though less elegant.
Put another way, I think that we can introduce non-linearities that will improve the entrainment process for the same reason that one musical instrument can be called better than another.
See my response to Michal Re: ambiguity.
I'm not sure what you mean by "ceramic wicks", but as for DIY analog methods, sensors involving electrodes and some sort of filtration medium (such as plaster) can't guarantee monotonicity and are subject to hysteresis, hence our going with commercial calibrated sensors. If we wanted to avoid electricity, then fingers work fine ;)
Re: photocells, that's a great point. It's too bad we're not going with distributed, wireless acquisition nodes because there's going to be a lot of wiring involved (like 20 feet per sensor).
Morgan
"system is effectively linear"
This is vague. What's the "system"? Reread Maturana and Varela, or Arakawa and Gins, or Heisenberg, Planck, Bohr. Where does the organism begin and phenomenon end? Maturana would remind us that "system" in the sense that you might mean it, meaning the electronic sensor + wires + microprocessor + code + wire + LED + battery + wire is only an artificial "god's eye view" linguistic fiction. (Read also the (in)famous essay about the artifactuality of physical measurement in gyro guidance system for ICBMs, reprinted in the Science Studies Reader, ed. Mario Biagioli.)
The mapping from soil humidity to human felt experience of the humidity is a composition of mappings, only some of which are computable, most of which are non-linear, most of which are in fact unknown. Call these mappings f1, f2, ..., fn.
If f is linear, and g is non-linear, then f º g is non-linear.If f and g are both non-linear, you have to be pretty damn lucky to find f º g linear.
Psycho-perception, alone, aside from anthropology and phenomenology, tells us that linearity is bogus.
In fact, the situation is far worse. Since at least one of the factor mappings fi is indeterminate, even if all the other factor mappings were determined, the composition f1 º f2 º ... º fn would still be indeterminate.
In fact the point of asking for monotonicity and reproducibility is precisely because that they together yield an "unambiguous" measure upon which individual humans can train. "Reading" and "unambiguous" are vague and paper over vast areas of uncertainty. (Cf. the psych-perception literature; read Husserl.) I suggested monotonicity and reproducibility as two practical qualities sometimes attainable by experimentalists with modest technical means. At least if y'all achieve these two qualities in your sensor system, then we'll be able to muddle along.
To measure soil humidity, I bet there are ceramic wicks that'll do this totally analog, with no macro-current needed. A chief value for doing digital is not so much to display humidity to the human, but its side effect: to get the humidity data into our Ozone system. Speaking of which, I think we should run photo-cells into Ozone. Maybe Claire can make some up from the Memory+Place leftovers -- asking Zohar, Patricia ?
Regards, Xin Wei