godelski • yesterday at 9:48 PM

  > You either have something documented and quantified and measured and objective criteria tickboxes and deal with this style of failure mode, or you rely on subjective judgment and assessment and accept the failure mode of bias, nepotism, old boy's clubs etc

My argument is that our current pursuit of the former only reinforces the existence of the latter.

You have a fundamental flaw in your argument, one that illustrates a common, yet fundamental, misunderstanding of science. There is no "objective" thing to measure, there are only proxies. I actually recently stumbled on a short by Adam Savage that I think captures this[0], although I think he's a bit wrong too. Regardless of precision we are always using a proxy. A tape measure does not define a meter, it only serves as a reference to compare with. A reference where not only the human makes error when reading, but that the reference itself has error[1]. So there are no direct measurements, there are only measurements by proxy.

You may have heard someone say "science doesn't prove things, it disproves them", and that's in part a consequence to this. Our measurements are meaningless without an understanding of their uncertainty (both quantifiable and unquantifiable!) as well as the assumptions they are made under.

I'm not trying to be pedantic here, I think this precision in understanding matters to the conversation. My argument is that by discounting those errors that they accumulate. We've had a pretty good run. This current system has only really started to be practiced in the 60s and 70's. So 50 years is a lot of time for error to accumulate. 50 years is a lot of time for small, seemingly insignificant, and easy to dismiss errors to accumulate into large, intangible, and complex problems.

There's something that I guess is more subtle in my argument: science is self-correcting. I don't mean "science" as the category of pursuits that seek truths about the world around us, but I mean "science" as a systematic approach to obtaining knowledge. A key reason this self-correction happens is due to replication. But in reality that is a consequence of how we pin down truth itself. We seek causal structures. More specifically, we seek counterfactual models. Assuming honest practitioners, failures of reproduction happen for primarily for one of two reasons: 1) ambiguity of communication between the original experimenters and those replicating or 2) a variation in conditions. 2) is actually quite common and tells us something new about that causal structure. In practice it is extremely difficult, if not impossible, to exactly replicate the conditions of the original experiment, so even with successful replication we gain information about the robustness of the results.

But why am I talking about all this? Because without the explicit acknowledgement of these limitations we seem to easily forget them. We are often treating substantially more subjective measures (such as impact or novelty) as far more objective than we would treat even physical measurements. It should be absolutely no surprise that things like impact are at best extremely difficult to measure. Even with a time machine we may not accurately measure the impact of a work for decades, or more. Ironically, a major reason for a work's impact to be found only after decades (or centuries) is the belief that at its time it had no impact, and was a dead end. You'd be amazed at how common this actually is. It's where jokes similar to how everything is named after the second person to discover something, the first being Euler[2]. But science is self-correcting. Even if a discovery of Euler's was lost, it is only a matter of time before someone (independently) rediscovers it.

I'm talking about this because there is no perfect system. Because a measurement without the acknowledge of its uncertainty is far less accurate than a measurement with. I'm talking about this because we will always have errors and the existence of them is not a reason to dismiss things. Instead we have to compare and contrast both the benefits and limits of competing ideas. We are only doing ourselves a disservice by pretending the limits don't exist. And if we mindlessly pursue objective measurements we'll only end up finding we've metric hacked our way into reading tea leaves. As we advance in any subject the minutia always ends up being the critical element (see [0]) and so the problem is it doesn't matter if we're 90% "objective" and 10% reading the tea leaves. Not when the decisions are made differentiating the 10%. In reality we're not even good at measuring that 90% when it comes to determining how productive academics are[3-5]

[0] https://www.youtube.com/shorts/JGa_X4QfE-0

[1] https://www.youtube.com/watch?v=EstiCb1gA3U

[2] https://en.wikipedia.org/wiki/List_of_topics_named_after_Leo...

[3] https://briankeating.substack.com/p/peter-higgs-wouldnt-get-...

[4] https://yoshuabengio.org/2020/02/26/time-to-rethink-the-publ...

[5] See the two links in this comment as further evidence. They are about relatively recent Nobel works that faced frequent rejections https://news.ycombinator.com/item?id=47340733