The Bayesian Trap

Interesting point about debaters not being convincible -- I think the goal of these public debates is never quite the convincing of the other, though (or at least, it never really seems to do much of that); rather, the goal seems to be to show to the audience what the best thinkers in their field think about their stance and how they defend it.

This is the best explanation I’ve seen so far. Not letting Bayes’ theorem become invalid is the most basic and also an important understanding.

Strangely Hume(an).

Whoa!, the practical Bayes Theorem is quite wild than the theoretical one.

Note: this was made mere six years ago. And look at how much damage to the scientific method can be done in such a short amount of time.

Cool video, Derek! The part about 'you probably not having the disease' makes sense. I like logical thinking like that.
I understand it's not the main point of your video, but I just want to note that we shouldn't fully confuse "what has happened in the past" with "doubt/confidence" even though they are correlated; they are not the same.
You said something towards the end about trying new things which is good. However, there are modern beliefs out there about needing to "search" for new things which is simply a belief ... which I don't hold. It is commonly known that the more someone knows, then the more that person understands they don't know ... so, if you are the type of person who is a bit uncomfortable with what you know and so you search for truth (i.e. you have doubts), that will make you realize there is a lot that you don't know and you may become more "unstable". Seeking and trying things doesn't necessarily lead to stability. That being said, as put in your video, 'if you are unhappy with the results and do the same thing over and over, try something new' - I agree. Absolutely! Getting out of something bad so you can get to something better is a great thing - it can "increase" confidence and stability too. Change is not bad ...just, you have been forewarned....

James 1:6b "...because the one who doubts is like a wave of the sea, blown and tossed by the wind." (NIV)

I watched this video again after 6 years. A lot of things have changed for me, you made science interesting and then there was no looking back for me. I love you, Dr. Derek Muller. Cheers to the day you decided to leave your full-time job as a professor and started to make these videos that can reach so many more people.

LOve youu Brooo

It explains so much of our current polarized world. And all that is needed is for people to have a little doubt about everything.

Pease keep doing your fantastic job. I know: You only explain what others found. But you do it in such a great way that this deserves credit from all of us.

is this filmed in cheddar gorge?

Did you change this video title? I watched this video several times to get Bayes Theorem intuition. I'm pretty sure the title wasn't "Trap"

amazing video

what the hell, why she? who is she?

To Peace ✌

Really nice video ! Thank you :)

Big fan of your videos. Incredible content. But here I think i spotted a misleading medical information that is worth mentioning. What is the definition of a 99% medical effective test? IRL it is never a test that gives you 1% of false positive. It may be 1% false negative... which is absolutely different. Just think about pregnancy test for example. 99% effective they say, but the rate of false positive is way lower. In your example, your patient should start treatment asap and not take a 2ond test.

Who is talking? I want to give him credit but it doesn't say.

Watching this for the zillionth time

It's 90% straight up if it's a self-selecting sample. In other words, only people who exhibit related symptoms would normally be tested. Otherwise how did the 90% figure arise in the first place.

Awesome

Where was this filmed, I was so distracted by the background haha

Is the prior probability 0.1 or 0.001%

2020/2021 would have been a fine time to apply bayes' theorem. When I did I got called names.

I learned in the navy to not believe anything until it happens then check the logbook.

What confuse me is that the Bayesian statistics is seen as entirely determined by the Bayesian theorem, which is a form of conditional probability. This theorem or formula can be derived from axiomatic probability theory using sample spaces and random variables. RA Fischer called Bayesian methods "The error of inverse probability" So what the hell is going on? .What fundamentally is the difference between Bayesian and Frequentist ? I have been trying to teach this stuff for the last 30 years and I am still confused. Does it come down to belief vs practical repetition of random experiments . I know myself I use both or maybe there are even more methods? Use what works?

Thanks for a very clear explanation of Bayes theorem, and for the insight that prior probability is such a significant factor in the result, and we often do not know what to use as the prior probability, as some of the comments show.

What if you were to run the formula for Bayes' theorem over and over except the only thing you change is you have the output from the past run as the prior probability for the next run? Does it converge on anything in particular?

This ESPECIALLY resonates with entrepreneurs. A whole show could be made about this; an entrepreneur is both the flashpoint and catalyst for the creation of a complex belief system (every component that makes up a company; employees, investors, clients, bystanders, etc) with NO PRIORS. I.e. do entrepreneurs have an intuitive ability to ignore Bayes’s theorem in the early founding days? Would love your thoughts @veritassium

- a fellow Queen’s physics grad

Hi Derek,
I'm a retired high school physics teacher, and a big fan of your videos. I used many of them during my teaching years to augment my lessons, so thank you for those. I was recently looking at a couple of videos on quantum tunneling and those prompted some questions (amazing huh?) so I am wondering if you might be so kind as to help me out with those. In those videos there was much reference to the probability wave front which brought these thoughts to mine:
1. How far does the probability wave front extend from the hydrogen electron in the ground state?
2. If that atom is excited does the probability wave front increase proportionally to the energy increase?
3. What is the extent of the probability wave front of electrons in an electron beam as it approaches the speed of light?

Kinda looks like the definition of the derivative without lim h ---» 0 ^^

Of course the cool thing about the sensitivity (identifies 99% of people with the disease) and specificity (wrongly identifies 1% of people without disease), is that you can alter these figures based on where you draw the line with the test result as to what you call a positive and what you call a negative (bear in mind that most tests are a measured concentration of an analyte and the pos/neg cut off is something we decide when setting up the test). If you wanted to increase specificity (less healthy people get a false positive) then you would at the same time decrease sensitivity (more diseased people would be missed by getting a false negative).

The test you described would be a good screening test - a good ‘rule out’ test.

Humility of uncertainty is what would be the feedstock for the theorem.
A seer is one who sees without any value judgement attached.

Phieieheiehiehie😂😂😂

Every now and then I return to this episode. It fills in the gaps. The story line, the content and the visuals turn this clip into a gem.

Exceptional

I always thought that a tests accuracy meant that it worked that amount of the time i.e whatever result it was 99% certain that it wa right. Bayes theorem has always felt unintuitive to me, probably because thats just the natural respinse to it

This is another gem!

A Bayesian failure: October 7th 2023. "It's the Hamass practicing".

The thumbnail reads PEWDIEPIE

I clicked on the video thinking that by 'The Bayesian Trap', he meant that the theorem is false but this turned out to be the perfect example of a clickbait.

this should be the gold standard for a motivational video!

so like where is he walking. the place?

What happens if our certainty vs uncertainty percentages apply to our belief in Baye's Theorem. If they are directly computable, can we apply Baye's Theorem to those data and get the overall certainty percentage reliably?

I believe Bayesian idea not exactly counter-intuitive, but our idea about percentages (a learnt quality ) makes us take the test probability at face value, as far as the disease test example go. Probably human brain has a better concept of Baysian theory at cognitive level, but learnt mathematics (through language) makes us fall for the easier route.

Ripped off without credit. I don;t like this.

This has got to be the way math should be explained. As a means to relate to it in real life. We often do not do math justice by not explaining it in a real-life applicable instance. Impressive job.