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

James Damore and three others end Google suit

bloomberg.com

Post reply on HN
Print view View original

981–990 of 1001 posts

Re: James Damore and three others end Google suit

#981

Post by TA0x0 »

shadowgovt wrote: post #980
TA0x0 wrote: post #979

Earlier quoted context omitted.

Those were theories to describe specific phenomenas, not facts. Certain facts are indisputable and not open to interpretation: FACT: African-Americans are more likely to commit violent crimes than Asian Americans.

Quite possibly, if one shows one's work on how one arrived at that statement (statistics can be bloody slippery things, and while individual data points are facts, the analysis, interpretation, and summarization of them starts to wander away from "fact," especially when one begins to do curve fitting or assumptions about population distributions). And if one runs with that fact alone and tries to build behaviors on i…

>Quite possibly, if one shows one's work on how one arrived at that statement (statistics can be bloody slippery things, and while individual data points are facts, the analysis, interpretation, and summarization of them starts to wander away from "fact," especially when one begins to do curve fitting or assumptions about population distributions).

Any and every statistical study ever done on race and crime supports the FACT stated above. Per capita, it's an indisputable fact that they commit more violent crimes, raw numbers alone, no curve fitting or extrapolation.

>And if one runs with that fact alone and tries to build behaviors on it, one's likely to fall into a similar trap to the one the phrenologists fell into: assuming that a single fact is descriptive enough of the world to dictate theory or practice.

Which brings us back to my original statement. Are facts racist and sexist?

Re: James Damore and three others end Google suit

#982

Post by shadowgovt »

TA0x0 wrote: post #981
shadowgovt wrote: post #980

Earlier quoted context omitted.

Quite possibly, if one shows one's work on how one arrived at that statement (statistics can be bloody slippery things, and while individual data points are facts, the analysis, interpretation, and summarization of them starts to wander away from "fact," especially when one begins to do curve fitting or assumptions about population distributions). And if one runs with that fact alone and tries to build behaviors on i…

>Quite possibly, if one shows one's work on how one arrived at that statement (statistics can be bloody slippery things, and while individual data points are facts, the analysis, interpretation, and summarization of them starts to wander away from "fact," especially when one begins to do curve fitting or assumptions about population distributions). Any and every statistical study ever done on race and crime supports…

History shows interpretation of facts can be extremely racist and sexist.

Re: James Damore and three others end Google suit

#983

Post by TA0x0 »

shadowgovt wrote: post #982
TA0x0 wrote: post #981

Earlier quoted context omitted.

>Quite possibly, if one shows one's work on how one arrived at that statement (statistics can be bloody slippery things, and while individual data points are facts, the analysis, interpretation, and summarization of them starts to wander away from "fact," especially when one begins to do curve fitting or assumptions about population distributions). Any and every statistical study ever done on race and crime supports…

History shows interpretation of facts can be extremely racist and sexist.

Interpretations are not facts.

There are some facts, like the one I listed above, that are not open to interpretation.

Re: James Damore and three others end Google suit

#984

Post by shadowgovt »

TA0x0 wrote: post #983
shadowgovt wrote: post #982

Earlier quoted context omitted.

History shows interpretation of facts can be extremely racist and sexist.

Interpretations are not facts. There are some facts, like the one I listed above, that are not open to interpretation.

"Water at standard atmospheric pressure boils at 100 degrees C and freezes at 0 degrees C" is not open to interpretation.

Making a terse sociological claim from statistics is almost always open to interpretation. Statistics is the science of interpreting discrete data points, looking for patterns and relations. Lies, damn lies, and all that. The only thing that stands alone without interpretation in statistics is the input data itself, and that tells us nothing.

I suspect that for the specific claim you've made, the evidence supports the claim, but I'd want to see whether they've controlled for conflating the definition of "crimes committed" with "charged with crimes" or "convicted of crimes" to be certain. Or whether the statistics are using the Bayesian or frequentist approach (after all, the claim being made is "are more likely," and that always assumes a giant pile of unstated priors). Or or or, etc.

Even still, whether the statistical claim being made is fact is uninteresting. If one assumes it's fact, the interesting question is "What do you do with it?" In isolation, with that one fact? Nothing. It's not nearly enough info to form theory or policy. Which is why my original response to your original post was about interpretation, not fact. Interpretation is where things get interesting. Facts without interpretation are dead as rocks.

Re: James Damore and three others end Google suit

#985

Post by TA0x0 »

shadowgovt wrote: post #984
TA0x0 wrote: post #983

Earlier quoted context omitted.

Interpretations are not facts. There are some facts, like the one I listed above, that are not open to interpretation.

"Water at standard atmospheric pressure boils at 100 degrees C and freezes at 0 degrees C" is not open to interpretation. Making a terse sociological claim from statistics is almost always open to interpretation. Statistics is the science of interpreting discrete data points, looking for patterns and relations. Lies, damn lies, and all that. The only thing that stands alone without interpretation in statistics is the…

>Making a terse sociological claim from statistics is almost always open to interpretation.

Almost, but in specific cases like above, it is not open to interpretation. It's a raw fact.

>Statistics is the science of interpreting discrete data points, looking for patterns and relations.

And organizing and presenting discrete data, without interpolation.

>The only thing that stands alone without interpretation in statistics is the input data itself, and that tells us nothing.

Wrong. Basic presentation without interpolation is a core pillar of statistics.

>I suspect that for the specific claim you've made, the evidence supports the claim, but I'd want to see whether they've controlled for conflating the definition of "crimes committed" with "charged with crimes" or "convicted of crimes" to be certain.

Even with controlling, African Americans are charged, commit, and are convicted of more crimes than Asian Americans.

>Or whether the statistics are using the Bayesian or frequentist approach (after all, the claim being made is "are more likely," and that always assumes a giant pile of unstated priors).

None of the above, just basic summary statistics.

>Even still, whether the statistical claim being made is fact is uninteresting. If one assumes it's fact, the interesting question is "What do you do with it?"

It's super interesting, considering what the OP asked.

>Nothing. It's not nearly enough info to form theory or policy.

It's more than enough info to form theory, and possibly policy.

>Which is why my original response to your original post was about interpretation, not fact.

And what I've stated above is fact, not interpretation. Back to square one.

>Interpretation is where things get interesting. Facts without interpretation are dead as rocks.

Wrong. Facts without interpretation are as alive as can be, and stand on their own.

Re: James Damore and three others end Google suit

#986

Post by shadowgovt »

TA0x0 wrote: post #985
shadowgovt wrote: post #984

Earlier quoted context omitted.

"Water at standard atmospheric pressure boils at 100 degrees C and freezes at 0 degrees C" is not open to interpretation. Making a terse sociological claim from statistics is almost always open to interpretation. Statistics is the science of interpreting discrete data points, looking for patterns and relations. Lies, damn lies, and all that. The only thing that stands alone without interpretation in statistics is the…

>Making a terse sociological claim from statistics is almost always open to interpretation. Almost, but in specific cases like above, it is not open to interpretation. It's a raw fact. >Statistics is the science of interpreting discrete data points, looking for patterns and relations. And organizing and presenting discrete data, without interpolation. >The only thing that stands alone without interpretation in statis…

> It's more than enough info to form theory, and possibly policy.

Can you give an example? What kind of policy would you form around that statistic?

Re: James Damore and three others end Google suit

#987

Post by TA0x0 »

shadowgovt wrote: post #986
TA0x0 wrote: post #985

Earlier quoted context omitted.

>Making a terse sociological claim from statistics is almost always open to interpretation. Almost, but in specific cases like above, it is not open to interpretation. It's a raw fact. >Statistics is the science of interpreting discrete data points, looking for patterns and relations. And organizing and presenting discrete data, without interpolation. >The only thing that stands alone without interpretation in statis…

> It's more than enough info to form theory, and possibly policy. Can you give an example? What kind of policy would you form around that statistic?

>Can you give an example?

Theory: there are manifestations in African American culture that breeds violence.

Policy: we need to funnel anti-violence education funds towards African American school populations.

Re: James Damore and three others end Google suit

#988

Post by shadowgovt »

TA0x0 wrote: post #987
shadowgovt wrote: post #986

Earlier quoted context omitted.

> It's more than enough info to form theory, and possibly policy. Can you give an example? What kind of policy would you form around that statistic?

>Can you give an example? Theory: there are manifestations in African American culture that breeds violence. Policy: we need to funnel anti-violence education funds towards African American school populations.

Yeah, that's exactly why that's not enough data to put together policy. You might as well measure bumps on people's heads to predict whether they want anti-violence funding; you'll have about as much luck as you will chasing it after skin color or "manifestations in African American culture that breeds violence" (what does that even mean?).

Re: James Damore and three others end Google suit

#989

Post by TA0x0 »

shadowgovt wrote: post #988
TA0x0 wrote: post #987

Earlier quoted context omitted.

>Can you give an example? Theory: there are manifestations in African American culture that breeds violence. Policy: we need to funnel anti-violence education funds towards African American school populations.

Yeah, that's exactly why that's not enough data to put together policy. You might as well measure bumps on people's heads to predict whether they want anti-violence funding; you'll have about as much luck as you will chasing it after skin color or "manifestations in African American culture that breeds violence" (what does that even mean?).

>Yeah, that's exactly why that's not enough data to put together policy.

Except it is, and I just proved it above.

>You might as well measure bumps on people's heads to predict whether they want anti-violence funding

Bumps on people's heads have nothing to do with violence. This is a non sequitur.

>you'll have about as much luck as you will chasing it after skin color

That makes no sense. The fact that the group is more violent is not open to interpretation, or "chasing", as I proved above.

>manifestations in African American culture that breeds violence" (what does that even mean?).

https://www.merriam-webster.com/dictionary/manifestation

Re: James Damore and three others end Google suit

#990

Post by shadowgovt »

TA0x0 wrote: post #989
shadowgovt wrote: post #988

Earlier quoted context omitted.

Yeah, that's exactly why that's not enough data to put together policy. You might as well measure bumps on people's heads to predict whether they want anti-violence funding; you'll have about as much luck as you will chasing it after skin color or "manifestations in African American culture that breeds violence" (what does that even mean?).

>Yeah, that's exactly why that's not enough data to put together policy. Except it is, and I just proved it above. >You might as well measure bumps on people's heads to predict whether they want anti-violence funding Bumps on people's heads have nothing to do with violence. This is a non sequitur. >you'll have about as much luck as you will chasing it after skin color That makes no sense. The fact that the group is m…

> Bumps on people's heads have nothing to do with violence. This is a non sequitur.

I submit to you it's exactly as irrelevant as skin color. I submit to you if you take the population that you have described, change only the category from "African American" to another category and nothing else, and kept the same population of people with no other parameters changed, you would not see a change in the violence rate.

Believing that measurable correlation implies causation is one of the major failings that I see often encouraged in the computer science space. It underpins some of the greatest failings of machine learning.

You made claims about population and then claims about interventions that suggest that the population membership is causal. What if the causal issue is poverty? Poverty in the United States is so deeply correlated with being a member of a racial category that it is extremely hard to disambiguate effects that show up in race from effects that show up in income and asset levels. This is what I mean when I say lies, damn lies, and statistics.

Because if the cause is poverty and not "African American culture," then violence interventions for African Americans are like "get your energy level up" interventions for people who are starving. You'll waste good money treating the wrong problem. You'll also miss people that also have the same problem but don't get any intervention because they don't fit the poorly-defined template.

Post reply on HN