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

Semantics derived automatically from language corpora contain human-like biases

scim.ag

Post reply on HN
Print view View original

11–20 of 92 posts

Re: Semantics derived automatically from language corpora contain human-like biases

#11

Post by pasbesoin »

privong wrote: post #8
fpp wrote: post #4

Earlier quoted context omitted.

very interesting topic - could you please share a link to the original paper. best probably to read that one first.

> very interesting topic - could you please share a link to the original paper. Unless the link was changed in the few minutes since you posted your comment, the link for the article is the original Science paper ( http://science.sciencemag.org/content/356/6334/183.full )

(From a Javascript-disabled perspective)

Page with actual link:

http://science.sciencemag.org/content/356/6334/183/tab-pdf

Link to PDF itself:

http://science.sciencemag.org/content/sci/356/6334/183.full....

Re: Semantics derived automatically from language corpora contain human-like biases

#13

Post by fpp »

privong wrote: post #8
fpp wrote: post #4

Earlier quoted context omitted.

very interesting topic - could you please share a link to the original paper. best probably to read that one first.

> very interesting topic - could you please share a link to the original paper. Unless the link was changed in the few minutes since you posted your comment, the link for the article is the original Science paper ( http://science.sciencemag.org/content/356/6334/183.full )

sorry mixed it up.

If I look at glove & WordNet usage e.g. for topic extraction, bagging / clustering or semantic similarity would you say we would need to get rid of such a bias, e.g. create something like a Geiger counter for NLP.

Alternative view - when doing sentiment analysis / classification would you say that such a bias actually helps to identify a type of sentiment in a doc / sentence.

Re: Semantics derived automatically from language corpora contain human-like biases

#14

Post by andreasvc »

yummyfajitas wrote: post #10
randomwalker wrote: post #2

Coauthor here. Some of the press articles about our work didn't have a lot of nuance (unsurprisingly), but in the paper we're careful about what we say, what we don't say, and what the implications are. Happy to engage in informed discussion :)

Do you have any evidence that this effect results in machines making systematically wrong inferences? Near as I can tell, your paper shows that these "biases" result in significantly more accurate predictions. For example, Fig 1 shows that a machine trained on human language can accurately predict the % female of many professions. Fig 2 shows the machine can accurately predict the gender of humans. Normally I'd expec…

I think your questions would be answered by reading the article. Particularly:

"In AI and machine learning, bias refers generally to prior information, a necessary prerequisite for intelligent action (4). Yet bias can be problematic where such information is derived from aspects of human culture known to lead to harmful behavior. Here, we will call such biases “stereotyped” and actions taken on their basis “prejudiced.”"

This definition is not unusual. This is about inferences that are wrong in the sense of prejudiced, not necessarily inaccurate.

Re: Semantics derived automatically from language corpora contain human-like biases

#15

Post by EGreg »

randomwalker wrote: post #2

Coauthor here. Some of the press articles about our work didn't have a lot of nuance (unsurprisingly), but in the paper we're careful about what we say, what we don't say, and what the implications are. Happy to engage in informed discussion :)

Isn't the word bias being redefined by a social justice point of view? Normally bias would be with reference to failing to match reality (eg women in general have physically weaker upper body than men), and not failing to match whatever standard of equality a society wishes were the case eventually.

Re: Semantics derived automatically from language corpora contain human-like biases

#16

Post by andreasvc »

glibgil wrote: post #6
randomwalker wrote: post #2

Coauthor here. Some of the press articles about our work didn't have a lot of nuance (unsurprisingly), but in the paper we're careful about what we say, what we don't say, and what the implications are. Happy to engage in informed discussion :)

Can you provide an example of how this bias might play out in a human-AI interaction?

It appears that the linked paper has examples.

Re: Semantics derived automatically from language corpora contain human-like biases

#17

Post by ylem »

randomwalker wrote: post #2

Coauthor here. Some of the press articles about our work didn't have a lot of nuance (unsurprisingly), but in the paper we're careful about what we say, what we don't say, and what the implications are. Happy to engage in informed discussion :)

Hi! I just read the paper--impressive work! Have you tried any other languages? For example, French or German?

Re: Semantics derived automatically from language corpora contain human-like biases

#18

Post by russdpale »

fpp wrote: post #13
privong wrote: post #8

Earlier quoted context omitted.

> very interesting topic - could you please share a link to the original paper. Unless the link was changed in the few minutes since you posted your comment, the link for the article is the original Science paper ( http://science.sciencemag.org/content/356/6334/183.full )

sorry mixed it up. If I look at glove & WordNet usage e.g. for topic extraction, bagging / clustering or semantic similarity would you say we would need to get rid of such a bias, e.g. create something like a Geiger counter for NLP. Alternative view - when doing sentiment analysis / classification would you say that such a bias actually helps to identify a type of sentiment in a doc / sentence.

Wouldn't this lead to an entire of idea of contextual bias? Times when it could benefit and be used, and times where it is occluded.

Re: Semantics derived automatically from language corpora contain human-like biases

#19

Post by yummyfajitas »

andreasvc wrote: post #14
yummyfajitas wrote: post #10

Earlier quoted context omitted.

Do you have any evidence that this effect results in machines making systematically wrong inferences? Near as I can tell, your paper shows that these "biases" result in significantly more accurate predictions. For example, Fig 1 shows that a machine trained on human language can accurately predict the % female of many professions. Fig 2 shows the machine can accurately predict the gender of humans. Normally I'd expec…

I think your questions would be answered by reading the article. Particularly: "In AI and machine learning, bias refers generally to prior information, a necessary prerequisite for intelligent action (4). Yet bias can be problematic where such information is derived from aspects of human culture known to lead to harmful behavior. Here, we will call such biases “stereotyped” and actions taken on their basis “prejudice…

The usual definition of bias in ML papers is E[theta_estimator - theta]. That is explicitly a systematically wrong prediction.

In any case, the paper suggests that this "bias" or "prejudice" is better described as "truths I don't like". I'm asking if the author knows of any cases where they are actually not truthful. The paper does not suggest any, but maybe there are some?

Post reply on HN