You must observe before doing experiments. Otherwise you would not know if you are asking the right questions. It takes a lot of observation before you can gather enough data to ask intelligent questions. you can ask questions before gathering 'enough' data... and those questions are useful... but you won't know which questions will lead you down a wrong path (alchemy) until you observe why those questions don't make sense (chemistry).
tl;dr: you both are right, and are talking past each other.
> tl;dr: you both are right, and are talking past each other.
Then you misunderstand my pushback. I fully agree, you must do both, doing both was explicitly my point.
parpfish said
>>>> there needs to be a long period [...] where people aren't doing experiments and are just observing and describing.
This is what I pushed back against because I agree with you, you need both. I understand my first comment was not as clear, but the one you replied to I think I made this apparent.
Observation and experimentation are tightly coupled. They drive each other. You need not start with one and you can also perform them in parallel. I think observation driving experimentation is clearer, but the reverse happens too (coupled). But you need to intervene, because experimentation is much like bug hunting in software. You need to find out where things differ, and those are always in the rare and unexpected events. If you just observe, you wouldn't know what to observe. You'd have oodles of nominal data, and very little about the edges. This is why it will lead to wrong conclusions. Interventions is about making the black swans appear in a more timely manner. You say "okay, it works with this case, but what happens when I change this one variable?"
Every scientist is intimately familiar with this because when you begin to experiment, you find that you need to start observing different things than what you initially expected. We can draw parallels to coding, where you have likely experienced that you cannot formulate what the actual code will be until you start to get your hands dirty. Sometimes it is similar to what you expected and you only need minor changes, but many times you discover you had naive assumptions and things need to change drastically. It is easy to believe only minor changes happen because the drastic changes are composed of many minor ones. If you don't believe me, start documenting or go look at commit histories.
I'll give an example of a real life event: infrared radiation! Herschel had the belief that different colors were associated with different temperatures (which is true) and the experiment he devised was one thermometer to measure the colors and two for controls. The story goes that he decided to place one thermometer just beyond red and discovered that it was the highest of them all, but truth is he probably just put one of his controls there because it is a natural place to place them saw a higher temperature and said "what the fuck?" (doesn't really matter which is true). This, accident or intentional, was intervention. *It was experimentation that led to observation.* This led him to devise more experiments to determine what was actually going on. You would never get here from observation because you can't observe infrared without specialized equipment (e.g. thermometers). You also won't get the right conclusion if your observations are just that thermometers are warmer around dark objects (a known fact Herschel used!) nor by observing thermometers reading higher temperatures near glass (lenses explain this). What are you going to observe? It is invisible!
I should also mention that at least a decade earlier Pictet was experimenting with mirrors and was even able to demonstrate the transmission of heat without the transmission of light! Hershel was probably well aware of Pictet's work. But what Herschel's work did was narrow down what was actually happening in Pictet's experiment. Because he logically understood that the prism was separating the colors in light and that this is done so in a continuous manner, so is doesn't seem unreasonable to think there could be "unseen light" beyond the red (that does fade in intensity to the eye). It may also be unsurprising that a year after Herschel's work that Ritter discovered ultraviolet light, but for this to be confirmed a thermometer wasn't really sensitive enough, but he could use silver chloride, since it was known that violet light caused a faster reaction and after knowing about invisible light beyond red, why not search for invisible light beyond violet?
It may again be easy to frame these as "observation first" but that's not accurate, they are coupled. Most certainly Ritter went hunting, and hunting for something that had not been observed. Because frankly, the observation was the proof. You'd never get there "from observation alone" (or rather not until you have a huge population and there are just enough random events and you have enough time. But even then, many things probably would not happen within the lifetime of the sun).
tl;dr: you both are right, and are talking past each other.