Hacker News new | past | comments | ask | show | jobs | submit login
We Don't Really Understand Randomness (adolos.substack.com)
7 points by adolos on July 22, 2020 | hide | past | favorite | 2 comments



> How do we break out of this feeling of certainty and discover the truth? When looking for cause, try to disprove rather than prove the correlation. Ask what could go against the pattern of the event you seek — where you have randomness.

The more general approach is called Abduction: https://en.wikipedia.org/wiki/Abductive_reasoning

Abduction is the process used in a legal trial--the crime or claim is a consequence, the prosecution proffers the hypothesis that the defendant is the causative agent and presents their evidence, while the defense attempts to disprove the hypothesis while also proffering their own alternative hypotheses. The entire trial and its procedural rules play out as an abductive process. (The book "Analysis of Evidence" describes this more formally, including exploring the relationships to Bayesian reasoning and other systems of logic.)

A doctor uses abduction during a diagnosis--take all the symptoms and patient context (prior illness, family history, etc), generate a set of potential hypotheses about the cause, and compare-and-contrast them to see which is the more likely. This is very often a more iterative, interactive process as you can intervene (ask questions, run tests, etc) to bolster or filter various hypotheses.

In science an informal abductive process is how you generate a hypothesis for rigorous experimentation. Abduction is how you pinpoint a potential cause from a consequence. The scientific method then starts from a cause (the hypothesized starting condition) and attempts to empirically reproduce the consequence, although formally you're attempting to disprove the consequence.

Abduction is a process of generating, comparing, and winnowing hypotheses, not making a single guess and looking for proof. And at the end, your chosen hypothesis, if you have one, is always contingent and probabilistic.

Abduction is often conflated with induction, but it's not really the same thing. For one, it's better characterized as an indefinitely iterative process rather than a discrete mechanism. (You can't string together a set of discretely abduced conclusions to infer something new; the new inference should begin anew by holistically assessing as many of the previous inputs as possible, as the broader context might reveal something new.) Recognizing the distinction permits far more rigorous and consistently analytical thinking, both professionally and in day-to-day life.


“For example, Sam L. Savage was convinced he could use statistics to accurately predict fruit prices, which was undoubtedly unwise. He hired a commodities trading company, but lost $300 million in a matter of weeks.”

I’m not familiar with that story. Is it the same person who wrote this book?

https://www.amazon.com/dp/B0096CT4VY/ref=dp-kindle-redirect?...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: