4 ms·

It’s a good recipe for making a lot of estimation errors. Worse, it will leave you feeling more confident than you should be. Then, when you least expect it, it

It’s a good recipe for making a lot of estimation errors. Worse, it will leave you feeling more confident than you should be. Then, when you least expect it, it’ll blow up in your face.

mindcrime 7y ago
There is merit to what you say, but that's the reason you have to have actual domain experts working on building the model, and why the calibration training matters. That said: "All models are wrong, but some are useful"
yen223 7y ago
One of the important points made in the book was that one should always be thinking of a range of values, not a single value. This ensures whatever model you use will capture the uncertainty that is inherent in any measurement.
hutzlibu 7y ago
It also sounds a bit like real life parody to me. How to find out employee morale? Well, you need to hire some statistical experts, who model everything and than give you an estimate and if they got the callibration right, the result might be even close to reality. Definitely easier than the other solution. Like talk to your employees once in a while and have a culture where they are not afraid to speak up if something is bothering them. Ah no, too complicated. Back to math. That works better when dealing with those irrational emotional humans.
- eyegor 7y ago
  Unfortunately this only works with a small set of workers, in a physical location. If you have 500+ employees, some remote, spanning different countries/cultures, it's not easy to tell how "company morale" is. You can try to send out anonymous surveys and such, but then you run into reporting bias, etc. It's a complicated problem at scale.
- mindcrime 7y ago
  The morale thing was just an example to help illustrate the principle...
  - hutzlibu 7y ago
    I know. And my point was, that you can take the statistic too far, especially when dealing with humans. And this is happening in my opinion since a while. Even Stalin said: "the death of one man is a tragedy, but the death of millions just statistic." And by his actions, they were just statistic to him. Humans are more than a number ... but when you are at a desk remote and only have numbers, than you are disconnected and it is much easier to make inhuman decisions as you are dealing with numbers and not humans anymore. So that being said, yes, having the numbers additional that signal a decline in morale for example, is probably not a bad thing to have. If this does not replace actual human contact, because human moral also drops, when they not feel treated as humans anymore and just as numbers.
    - mindcrime 7y ago
      So that being said, yes, having the numbers additional that signal a decline in morale for example, is probably not a bad thing to have. If this does not replace actual human contact, because human moral also drops, when they not feel treated as humans anymore and just as numbers. Agreed. And this is true of every tool... it has a place, and when used properly it can be of benefit. But it can also be misused in a way that is harmful or destructive.
- rramadass 7y ago
  True somewhat. That is what bothered me with this book. While the importance of Statistical Models is undeniable, i am not sure whether they are applicable everywhere. Nature has endowed us with certain heuristic decision making models (see the works of Gary Klein on RPDM) and many a time one can decide based on "gut feelings" (see the works of Gerd Gigerenzer).
- acomjean 7y ago
  I worked at a clueless large company, but even they knew morale was bad, thus me being placed on the morale comititee. Our one meeting we decided it was difficult to improve morale, but after some discussion we did end up with an ice cream social..
- grumpy8 7y ago
  What about both though? Or something similar but different, what is better between talking to users or looking at data points? What about both?
memetomancer 7y ago
based on your lack of suggested alternative, I'm gonna guess you are just cutting shit down out of hand. i mean, what are you saying? _don't_ try to quantify stuff? rely on intuition? weak comment pal.
- eyegor 7y ago
  Please edit this to be less disparaging. I agree with your complaint, it's just presented in a mean spirited way. From the guidelines, > Be kind. Don't be snarky.
  - memetomancer 7y ago
    fair enough, I was being more flip than the parent comment warranted. embarrasingly, I can't seem to figure out how to edit a comment so leaving this here: OP: I am sorry for harshing ya. I shouldn't be so dang cranky. cheers. * Edit: I guess I waited too long and the edit button is gone now.
rramadass 7y ago
Exactly! This book has to be read with Nassim Taleb's books for a healthy dose of skepticism.
- mindcrime 7y ago
  Even though I am a fan of Hubbard's work, I would agree with this to a point. I am also a big fan of Taleb's work, and I do believe that a healthy measure of skepticism in all things is a Good Idea™.

It’s a good recipe for making a lot of estimation errors. Worse, it will leave you feeling more confident than you should be. Then, when you least expect it, it’ll blow up in your face.

mindcrime 7y ago

There is merit to what you say, but that's the reason you have to have actual domain experts working on building the model, and why the calibration training matters. That said: "All models are wrong, but some are useful"

yen223 7y ago

One of the important points made in the book was that one should always be thinking of a range of values, not a single value. This ensures whatever model you use will capture the uncertainty that is inherent in any measurement.

hutzlibu 7y ago

It also sounds a bit like real life parody to me. How to find out employee morale? Well, you need to hire some statistical experts, who model everything and than give you an estimate and if they got the callibration right, the result might be even close to reality. Definitely easier than the other solution. Like talk to your employees once in a while and have a culture where they are not afraid to speak up if something is bothering them. Ah no, too complicated. Back to math. That works better when dealing with those irrational emotional humans.

eyegor 7y ago
Unfortunately this only works with a small set of workers, in a physical location. If you have 500+ employees, some remote, spanning different countries/cultures, it's not easy to tell how "company morale" is. You can try to send out anonymous surveys and such, but then you run into reporting bias, etc. It's a complicated problem at scale.
mindcrime 7y ago
The morale thing was just an example to help illustrate the principle...
- hutzlibu 7y ago
  I know. And my point was, that you can take the statistic too far, especially when dealing with humans. And this is happening in my opinion since a while. Even Stalin said: "the death of one man is a tragedy, but the death of millions just statistic." And by his actions, they were just statistic to him. Humans are more than a number ... but when you are at a desk remote and only have numbers, than you are disconnected and it is much easier to make inhuman decisions as you are dealing with numbers and not humans anymore. So that being said, yes, having the numbers additional that signal a decline in morale for example, is probably not a bad thing to have. If this does not replace actual human contact, because human moral also drops, when they not feel treated as humans anymore and just as numbers.
  - mindcrime 7y ago
    So that being said, yes, having the numbers additional that signal a decline in morale for example, is probably not a bad thing to have. If this does not replace actual human contact, because human moral also drops, when they not feel treated as humans anymore and just as numbers. Agreed. And this is true of every tool... it has a place, and when used properly it can be of benefit. But it can also be misused in a way that is harmful or destructive.
rramadass 7y ago
True somewhat. That is what bothered me with this book. While the importance of Statistical Models is undeniable, i am not sure whether they are applicable everywhere. Nature has endowed us with certain heuristic decision making models (see the works of Gary Klein on RPDM) and many a time one can decide based on "gut feelings" (see the works of Gerd Gigerenzer).
acomjean 7y ago
I worked at a clueless large company, but even they knew morale was bad, thus me being placed on the morale comititee. Our one meeting we decided it was difficult to improve morale, but after some discussion we did end up with an ice cream social..
grumpy8 7y ago
What about both though? Or something similar but different, what is better between talking to users or looking at data points? What about both?

memetomancer 7y ago

based on your lack of suggested alternative, I'm gonna guess you are just cutting shit down out of hand. i mean, what are you saying? _don't_ try to quantify stuff? rely on intuition? weak comment pal.

eyegor 7y ago
Please edit this to be less disparaging. I agree with your complaint, it's just presented in a mean spirited way. From the guidelines, > Be kind. Don't be snarky.
- memetomancer 7y ago
  fair enough, I was being more flip than the parent comment warranted. embarrasingly, I can't seem to figure out how to edit a comment so leaving this here: OP: I am sorry for harshing ya. I shouldn't be so dang cranky. cheers. * Edit: I guess I waited too long and the edit button is gone now.

rramadass 7y ago

Exactly! This book has to be read with Nassim Taleb's books for a healthy dose of skepticism.

mindcrime 7y ago
Even though I am a fan of Hubbard's work, I would agree with this to a point. I am also a big fan of Taleb's work, and I do believe that a healthy measure of skepticism in all things is a Good Idea™.