---
tags: ggg, ggg2020, ggg298
---
# Notes: GGG298 Discussion - Week 5

Reading: [Upgrade your cargo cult](https://meaningness.com/metablog/upgrade-your-cargo-cult)

## Notes

"The only people who shouldn't do science are those who never doubt the process."

"turns out no one really knows what science is or how to do it", kind of like money being a consensual hallucination. (see: bitcoin)

Navel gazing challenge: making progress on specific tasks can't be done easily while wondering about the whole basis for your task. Challenging...

Why does this matter for data science? Search for understanding is in common.

---

when do you stop doing cargo cult science?

* never, imitation is always a part of it
* when you start asking why and not how
* when you care about the actual answer for its own reason
* no one pivotal point, gradual transition
* it's up to you! (but you have to recognize it)

is there an inextricable connection between applying/developing _new_ methods and doing "real" science

Who decides if you are doing science?

---

> "Honesty comes out of curiosity, mostly, I think. If you really do want to know, there’s much less motivation to promote a wrong answer—arrived at either through deliberate fraud or sloppy, inadequately-controlled experimentation."

who decides if you are doing science?

---

> "Tacit knowledge often contradicts explicit standards—and therefore could not, even in principle, be learned from manuals. In every workplace, there are the official rules, and then there is “the way we do things,” which involves extensive implicit exceptions."

how does this connect with reproducibility?

----

(connect with JJ's question about why can't we just all collaborate?)

> "Every social group has two inseparable aspects: it is an invaluable and inescapable resource, but also a zone of socially-enforced conformity, thought-taboos, and dysfunctional practices and attitudes"

----

why does any of this matter?

is there a better way?

---

what does this mean?

> Recognizing the limitations of rationalist rituals does not mean abandoning them. You have to use methods, and you also have to relativize them. You need meta-rational competence to recognize when a method is appropriate, and when it is not.21 There is no explicit method for that—but, like riding a bicycle, it can be cultivated as tacit know-how. “Reflection-in-action” describes that meta-level learning process.

----

> Communities (including, but not only, institutions) can take a meta-systematic view of themselves. They can reflect on their own goals, structure, dynamics, and norms. Such reflection may afford much greater leverage than incremental process optimization.

How does this all apply to data science ?

### Some supplementary reading/viewing for those who are interested

These are resources related to the discussion of the "Upgrade your cargo cult for the win" article.

[Kary Mullis talking about the birth of the Royal Society and modern science](https://www.ted.com/talks/kary_mullis_play_experiment_discover/transcript?language=en) - includes transcript. Fascinating stuff! The book [Leviathan and the Air Pump](https://www.amazon.com/Leviathan-Air-Pump-Experimental-Princeton-Classics/dp/0691150206) is mentioned in this and is a good (though challenging) read.

[Darwin's Ghosts](https://www.amazon.com/Darwins-Ghosts-Secret-History-Evolution/dp/0812981707) is a strong read for those interested in how eugenics permeates biology. Highly recommended to me by several STS faculty. Note, here "eugenics" refers to the larger issue of the role and value of diversity in populations. tl;dr many of the early figures in biology and population genetics had strong positions on this.