When one is working, I think the confusion period of understanding a concept or problem is the most rewarding part of the learning process as it forces you to really understand what is going on. I am lucky in my current job to be given the time and space to really get to grips with the problem I am assigned to solve and to get to grips with the nuts and bolts through many periods of thinking 'what on earth am I doing?'
As a student, facing that same confusion is unsettling at best and terrifying at worst as you feel that understanding one particular concept is crucial as to not fall behind. One can easily fall into a cascading panic as week upon week new topics are introduced to you. As a student, dwelling in the uncertainty/confusion period can be terrifying as you know there could be so much riding on you achieving a certain level of performance.
Thanks for the comment Eoghan. I totally agree that one can learn to feel the reward, or sense the reward, in the center of the experience of confusion.
I loved how you expressed this: "As a student, dwelling in the uncertainty/confusion period can be terrifying as you know there could be so much riding on you achieving a certain level of performance."
That's so raw and real -- and I think that's true for any real period of learning and growth as well, in life, but with being a student there's a timetable on the learning, as it is supposed to be achieved by a certain date, and that really intensifies the terror.
So a good lesson for us in life is to give ourselves as generous a timetable as possible when our tasks involve much learning and growing, I suppose.
I also enjoyed this post. The fundamental difficulty with Bayesian thinking is that it is irreducibly probabilistic and subjective. The idea of the classical hypothesis testing is that, with enough data points, we can reach objective statistical certainty, at least in the sense of rejecting the null hypothesis. But with Bayesian thinking, the most you can hope for is that posterior distribution will be tighter than the prior, which is inevitably subjective. When statisticians try to incorporate "Bayesian" methods into a classical mode of thinking, grief ensues.
I promised myself I would not debate you but i couldn't help it. :)
Bayesian methods also have consistency and asymptotic gaussianity results just like frequentist methods! Bernstein-Von Mises being the most important, but of course there was the extention by Le Cam, Diaconis and co.
But also, in everyday practice, frequentist inference has major subjective components. The choice of test size is the most obviously arbitrary one. But the choice of model specification is also pretty subjective, which we try to get at with our also-subjective robustness checks, which overall as a practice is not working too badly I suspect.... or the subjectivity there is not the problem in my view. But of course I would say that. :)
TBC, I'm a Bayesian (or maybe post-Bayesian). I was trying to give my explanation of why Bayesian thinking hasn't prevailed over frequentism/classical statistics, and why even when "Bayesian" methods are adopted, they aren't used in a genuinely Bayesian way. Rereading what I wrote, this doesn't come across as well as I hoped.
Ah, I didn't realise indeed, thanks for clarifying! I agree that the perception among frequentists that the Bayesian method is more subjective -- while in fact it is simply more obvious and honest about its subjective aspects -- is the major stumbling block to adoption.
To spell it out, subjectivity is inevitable. The social conventions adopted by frequentists/classicists (significance testing and so on) have obscured this. That's beneficial in some ways (communication of results in an apparently objective form), but has contributed to very bad outcomes like p-hacking, which might have been avoided with Bayesian methods.
I liked this a lot. I have been co-teaching impact evaluation (so causal inference and identification) this semester and the main take-away I have is how hard all of this is. That makes me feel like I don't really fit in or get it, but I think it's right that is isn't supposed to be easy. So, I appreciated that post a lot. Yay to learning!
I honestly think that one of the cruellest things we do to ourselves and others is to say that things are supposed to be easy. I am so happy this resonated for you -- thanks so much for commenting and sharing the post.
Thanks for writing this essay. It’s better than a tweet, because it displays the complexity of thought that a tweet never could. Some thoughts, more to counter, and not needing a response to me:
1. Most important news will filter to you without Twitter even in the “Golden Age” of Twitter. Do you really need Twitter? Let the Nazi bar go bankrupt.
2. With me using the Bayesian term to mean “constantly updating my beliefs with lots of new data”: if the goal is to think more or better or whatever, should you consider your sources more? Twitter is beyond the case of “noise crowding out signal.” Even for an academic, isn’t there a spectrum of colleagues? I assume some are pinnacle, some are trash, and most are going for easy enough results to get the grant money. For the thoughts that you’d like to know, who are the best? Masters train masters. Who are the new surprising innovators? Trends begin with fresh thinkers. If you must remain on Twitter, use is more like a search engine: who do the best follow, and who do they follow? Where is the end of their knowledge graph?
3. If there’s a happy ending, it’s growth. Thinking is hard like weightlifting is hard or growing a tomato plant is hard. It’s the point. The insight, the gainzZz, the tomato is the fruit of the labor. Neuroscientifically, if you don’t stress the brain it won’t form novel connections and most teachers stop at conveying the information. The actual teaching is more like modeling or coaching: as the student attempts, the teacher is there to correct form, to shortcut inefficiencies, to correct nonsense (or to shout “good job!” when the child puts the square peg in the square hole). As opposed to one-to-one sessions, it’s hard to scale a room full of individual efforts by students and that’s worthily fretful.
I hope none of this comes off as oppositional—you’ve given me good thoughts for the day and hopefully this is likewise worthy the time.
Let me pose a corollary, or perhaps an important caveat. As you say, it is sometimes very hard to know what is the "right" answer. This is a common feeling as an academic (especially as someone in methods -- so many use cases do not fall into the "solved" categories).
But, you don't get to make decisions (practical or ethical) with certainty! Just saying you are confused or unsure or don't want to commit to something is making a choice in the end; either absconding from a position of influence, or endorsing the status quo.
This matters with banal things (hiring people, referee reports) and ethical things (supporting new ideas or positions). I think we often forget that.
Thanks Paul <3 Yeah, that's an important corollary. I go back and forth on it because I think I'm quite prone to the position that there is no such thing as abstaining or not making a decision, but now I wonder if that's true. One can postpone a decision for a while, and sometimes that feels important and useful. Maybe one way to put it is that there is no ultimate abstention, but sometimes it is possible to wait and see, gather more data, fortify yourself, etc.
Thanks again for commenting and glad it resonated with you :)
Really enjoyed this piece!
When one is working, I think the confusion period of understanding a concept or problem is the most rewarding part of the learning process as it forces you to really understand what is going on. I am lucky in my current job to be given the time and space to really get to grips with the problem I am assigned to solve and to get to grips with the nuts and bolts through many periods of thinking 'what on earth am I doing?'
As a student, facing that same confusion is unsettling at best and terrifying at worst as you feel that understanding one particular concept is crucial as to not fall behind. One can easily fall into a cascading panic as week upon week new topics are introduced to you. As a student, dwelling in the uncertainty/confusion period can be terrifying as you know there could be so much riding on you achieving a certain level of performance.
Thanks for the comment Eoghan. I totally agree that one can learn to feel the reward, or sense the reward, in the center of the experience of confusion.
I loved how you expressed this: "As a student, dwelling in the uncertainty/confusion period can be terrifying as you know there could be so much riding on you achieving a certain level of performance."
That's so raw and real -- and I think that's true for any real period of learning and growth as well, in life, but with being a student there's a timetable on the learning, as it is supposed to be achieved by a certain date, and that really intensifies the terror.
So a good lesson for us in life is to give ourselves as generous a timetable as possible when our tasks involve much learning and growing, I suppose.
I also enjoyed this post. The fundamental difficulty with Bayesian thinking is that it is irreducibly probabilistic and subjective. The idea of the classical hypothesis testing is that, with enough data points, we can reach objective statistical certainty, at least in the sense of rejecting the null hypothesis. But with Bayesian thinking, the most you can hope for is that posterior distribution will be tighter than the prior, which is inevitably subjective. When statisticians try to incorporate "Bayesian" methods into a classical mode of thinking, grief ensues.
I promised myself I would not debate you but i couldn't help it. :)
Bayesian methods also have consistency and asymptotic gaussianity results just like frequentist methods! Bernstein-Von Mises being the most important, but of course there was the extention by Le Cam, Diaconis and co.
But also, in everyday practice, frequentist inference has major subjective components. The choice of test size is the most obviously arbitrary one. But the choice of model specification is also pretty subjective, which we try to get at with our also-subjective robustness checks, which overall as a practice is not working too badly I suspect.... or the subjectivity there is not the problem in my view. But of course I would say that. :)
TBC, I'm a Bayesian (or maybe post-Bayesian). I was trying to give my explanation of why Bayesian thinking hasn't prevailed over frequentism/classical statistics, and why even when "Bayesian" methods are adopted, they aren't used in a genuinely Bayesian way. Rereading what I wrote, this doesn't come across as well as I hoped.
Ah, I didn't realise indeed, thanks for clarifying! I agree that the perception among frequentists that the Bayesian method is more subjective -- while in fact it is simply more obvious and honest about its subjective aspects -- is the major stumbling block to adoption.
To spell it out, subjectivity is inevitable. The social conventions adopted by frequentists/classicists (significance testing and so on) have obscured this. That's beneficial in some ways (communication of results in an apparently objective form), but has contributed to very bad outcomes like p-hacking, which might have been avoided with Bayesian methods.
I liked this a lot. I have been co-teaching impact evaluation (so causal inference and identification) this semester and the main take-away I have is how hard all of this is. That makes me feel like I don't really fit in or get it, but I think it's right that is isn't supposed to be easy. So, I appreciated that post a lot. Yay to learning!
I honestly think that one of the cruellest things we do to ourselves and others is to say that things are supposed to be easy. I am so happy this resonated for you -- thanks so much for commenting and sharing the post.
Thanks for writing this essay. It’s better than a tweet, because it displays the complexity of thought that a tweet never could. Some thoughts, more to counter, and not needing a response to me:
1. Most important news will filter to you without Twitter even in the “Golden Age” of Twitter. Do you really need Twitter? Let the Nazi bar go bankrupt.
2. With me using the Bayesian term to mean “constantly updating my beliefs with lots of new data”: if the goal is to think more or better or whatever, should you consider your sources more? Twitter is beyond the case of “noise crowding out signal.” Even for an academic, isn’t there a spectrum of colleagues? I assume some are pinnacle, some are trash, and most are going for easy enough results to get the grant money. For the thoughts that you’d like to know, who are the best? Masters train masters. Who are the new surprising innovators? Trends begin with fresh thinkers. If you must remain on Twitter, use is more like a search engine: who do the best follow, and who do they follow? Where is the end of their knowledge graph?
3. If there’s a happy ending, it’s growth. Thinking is hard like weightlifting is hard or growing a tomato plant is hard. It’s the point. The insight, the gainzZz, the tomato is the fruit of the labor. Neuroscientifically, if you don’t stress the brain it won’t form novel connections and most teachers stop at conveying the information. The actual teaching is more like modeling or coaching: as the student attempts, the teacher is there to correct form, to shortcut inefficiencies, to correct nonsense (or to shout “good job!” when the child puts the square peg in the square hole). As opposed to one-to-one sessions, it’s hard to scale a room full of individual efforts by students and that’s worthily fretful.
I hope none of this comes off as oppositional—you’ve given me good thoughts for the day and hopefully this is likewise worthy the time.
Very much appreciated these thoughts and also your careful framing to ensure I didn't read it as oppositional.
Re: (1) I am thinking of how to get off twitter... I have my excuses/reasons to stay on for the moment but I am hoping to pivot.
Your point (2) is very good and well taken. And (3) as well in fact.
Thanks for the comment!
Such a lovely post Rachael!
Let me pose a corollary, or perhaps an important caveat. As you say, it is sometimes very hard to know what is the "right" answer. This is a common feeling as an academic (especially as someone in methods -- so many use cases do not fall into the "solved" categories).
But, you don't get to make decisions (practical or ethical) with certainty! Just saying you are confused or unsure or don't want to commit to something is making a choice in the end; either absconding from a position of influence, or endorsing the status quo.
This matters with banal things (hiring people, referee reports) and ethical things (supporting new ideas or positions). I think we often forget that.
Thanks Paul <3 Yeah, that's an important corollary. I go back and forth on it because I think I'm quite prone to the position that there is no such thing as abstaining or not making a decision, but now I wonder if that's true. One can postpone a decision for a while, and sometimes that feels important and useful. Maybe one way to put it is that there is no ultimate abstention, but sometimes it is possible to wait and see, gather more data, fortify yourself, etc.
Thanks again for commenting and glad it resonated with you :)