Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

What Data Can’t Do

newyorker.com

Post reply on HN
Print view View original

21–30 of 106 posts

Re: What Data Can’t Do

#21

Post by mjburgess »

lmm wrote: post #17

Does the use of statistics actually amplify misunderstanding, or merely reveal misunderstandings that were already there? In any of these examples given - predicting rearrests, infant mortality, or so on - it's hard to imagine that someone not using numbers would have reached a conclusion that was any closer to the truth. Data has its limits, but the solution is usually - maybe even always - more data, not less.

It's pretty trivial to predict things without "data". Data just means using some measurement system to obtain measurements of some target phenomenon. Many targets cannot be measured, or have not occurred to be measured.

Reasoning counter-factually is trivial: What would happen if I dropped this object in this place in which an object, of this kind, has never been dropped before?

Well apply relevant models, etc. and "the object falls, rolls, pivots, etc.".

This is reasoning-forward from models, rather than backwards from data. And it's the heart of anything that makes any sense.

Data is not a model and provides no model. The "statistics of mere measurement" is a dangerously utopian misunderstanding of what data is. The world does not tell you, via measurement, what it is like.

Re: What Data Can’t Do

#22

Post by erichahn »

carlosf wrote: post #4

I am increasingly worried with people applying ML in everything without any rigour. Statical inference generally only works well in very specific conditions: 1 - You know the distribution of the phenomenon under study (or make an explicit assumption and assume the risk of being wrong) 2 - Using (1), you calculate how much data you need so you get an estimation error below x% Even though most ML models are essentially…

Isn't the point of ML exactly that you don't know the underlying distribution? How is this ever assumed in any way? ML is not parametric statistics.

Re: What Data Can’t Do

#23

Post by kazinator »

> doctors would be given a financial incentive to see patients within forty-eight hours.

Not measuring that from the first contact that the patient made is simply dishonest.

"Call back in three days to make the appointment, so I can claim you were seen within 48 hours, and therefore collect a bonus" amounts to fraud because the transaction for obtaining that appointment has already been initiated.

I mean, they could as well just give the person the appointment in a secret, private appointment registry, and then copy the appointments from that registry into the public one in such a way that it appears most of the appointments are being made within the 48 hour window. Nothing changes, other than that bonuses are being fraudulently collected, but at least the doctor's office isn't being a dick to the patients.

Re: What Data Can’t Do

#24

Post by _rpd »

visarga wrote: post #18

Data is very limited, indeed. We can't predict outside the distribution, or unrelated events (without a causal link), or random events in the future. We should be humble about the limits of data.

Sure, but coding a human-equivalent response to such events is trivial (because no one and nothing responds well to such events).

Re: What Data Can’t Do

#25

Post by analog31 »

carlmr wrote: post #19
analog31 wrote: post #14

Earlier quoted context omitted.

"Scientist free science," -- being able to optimize systems without understanding them, has been a dream of the business world since the dawn of time. There's always been a market for cookbook recipes that automate the collection of data, and interpretation of results. Before ML, there were "design of experiments," and "statistical quality control."

>Before ML, there were "design of experiments," and "statistical quality control." Statistical quality control, at least the way I know it, is very useful in finding problems in your process. I'm also not sure how this fits with your premise. It's about optimizing systems by first finding out where to look, and then looking there in detail with expert knowledge, i.e. deep understanding of your system.

I'm definitely with you there, but I've also seen the side of it where it turns into a cargo cult and runs headlong into the replication crisis.

Perhaps the good thing is that as the new things gain popular attention, the old techniques such as SPC are under less pressure to support success theater, and revert to being actual useful, solid tools.

Re: What Data Can’t Do

#27

Post by kazinator »

mdip wrote: post #13

Kind of love the initial story in the article about 48-hour wait times. I had a stint writing conferencing software for quite some time, and every once in a while we'd come across a customer requirement that had capabilities which were obvious to us developers "would be misused". As a result, we did the "Thinking, Fast and Slow" pre-mortem to help surface other ways that the system could be attacked (along with what…

When the doctor's office is inundated with patient visits, you cannot fix scheduling back-logs by fiddling with the scheduling algorithm, no matter how much data you have.

Say the calendar is initially empty and 1000 people want to see the doc, right now. You can fill them all into the calendar, or you can play games that solve nothing, like only filling tomorrow's schedule with 10 people, asking 990 of them to call back. That doesn't change the fact that it takes 100 days to see 1000 patients. All it does is cause unfair delays; the original 1000 can be pre-empted by newcomers who get earlier appointments since their place in line is not being maintained.

Re: What Data Can’t Do

#28

Post by currymj »

carlosf wrote: post #4

I am increasingly worried with people applying ML in everything without any rigour. Statical inference generally only works well in very specific conditions: 1 - You know the distribution of the phenomenon under study (or make an explicit assumption and assume the risk of being wrong) 2 - Using (1), you calculate how much data you need so you get an estimation error below x% Even though most ML models are essentially…

i think this actually gets at what makes applied ML distinct from statistics as a practice, even though there is a ton of overlap.

statisticians make assumptions 1 and 2, and think of themselves as trying to find the "correct" parameters of their model.

people doing applied ML typically assume they don't know 1 (although they might implicitly make some weak assumptions like sub-gaussian to avoid fat tails, etc.) and also typically don't care about being able to do 2. and they don't care about their parameters; in a sense to an ML practitioner, every parameter is a nuisance parameter.

instead you assume you have some reliable way of evaluating performance on the task you care about -- usually measuring performance on an unseen test set. as long as this is actually reliable, then things are fine.

but you are right that in the face of a shifting distribution or an adversary crafting bad inputs, ML models can break down -- but there is actually a lot of research on ways to deal with this, which will hopefully reach industry sooner rather than later.

Re: What Data Can’t Do

#29

Post by nonameiguess »

erichahn wrote: post #22
carlosf wrote: post #4

I am increasingly worried with people applying ML in everything without any rigour. Statical inference generally only works well in very specific conditions: 1 - You know the distribution of the phenomenon under study (or make an explicit assumption and assume the risk of being wrong) 2 - Using (1), you calculate how much data you need so you get an estimation error below x% Even though most ML models are essentially…

Isn't the point of ML exactly that you don't know the underlying distribution? How is this ever assumed in any way? ML is not parametric statistics.

(Some) ML is non-parametric, but there are always some questions you need to be able to answer about your data. At bare minimum, is the generating process ergodic, what is the error of your measurement procedure, how representative of the true underlying distribution is your sampling procedure? All use of data should start with some exploratory analysis before you ever get to the modeling stage.

Once you have a model, at minimum understand how to tune for the tradeoffs of different types of error and don't naively optimize for pure accuracy. At the obvious extremes, if you're trying to prevent nuclear attack, false negatives are much more costly than false positives, if you're trying to figure out whether to execute someone for murder, false positives are much more costly than false negatives. Understand the relative costs of different types of error for whatever you're trying to predict and proceed accordingly.

Re: What Data Can’t Do

#30

Post by lifeisstillgood »

mdip wrote: post #13

Kind of love the initial story in the article about 48-hour wait times. I had a stint writing conferencing software for quite some time, and every once in a while we'd come across a customer requirement that had capabilities which were obvious to us developers "would be misused". As a result, we did the "Thinking, Fast and Slow" pre-mortem to help surface other ways that the system could be attacked (along with what…

conferencing as in 'ComicCon' or 'Zoom'? Can you give an example?
Post reply on HN