I am a retired Elec Engineer. Industrial control and automation, observer and occasional recreational user for research and coding. Mostly DeepSeek.
One thing I have noted is the obsequious content in a conversation. Presumably like click bait to bring the user back? Tells you what you want to hear.
On a bigger issue. Our climate clearly cannot afford the power use. Although I have seen different figures about this. Data centres just serving social media are the real culprits at least for now. What is the betting that breakthroughs are in the wings. I recall when the mapping of the human genome was first proposed it was thought to require immense computing power. It is now done routinely for a few$.
Good stuff, but speaking in the following way will confirm the inherent bias that librarians seem to be living with:
"where existing social inequalities, historical underrepresentation, and the demographics of whoever created and curated the data all leave their marks. A medical AI trained primarily on data from white male patients will perform less accurately when applied to women and people of color."
Less accuracy will occur IF and ONLY IF there ares actual differences in those subpopulations. Assumed difference is just bias. Here is a hint: when you use terms such as "underrepresentation", "social inequalities" you are dealing in rabbit in the hat putting. This is especially true with referencing the "demographics of the" creator of the data set.
This is a silly, silly comment. Male and female humans are incredibly different in their biology. Different races are susceptible to different difficulties. So, yes, medical data based on adult human males of Western European ancestry will be extremely inaccurate if applied to human females of African descent.
You are either a very ignorant human or a bot causing "outrage engagement."
Oh, I am not ignorant at all and I don't seek outrage, but you sure do jump to conclusions and have become outraged. I wrote that iff ( in logic, iff means if and only if) there is a difference - which you seem fixated on as to women (Restatement of the Obvious) - then that difference will matter. My point, which, whoosh, went right over your head was that IF there is a difference then a model only examining one sex will miss that difference. Well, OK, I did not lay out that latter point - I thought it rather obvious. But any one who enters data on only men will miss the differences as to women.
But my point was not about differences. It was about bias of the librarian author. There is the implicit assumption that differences must exist therefore will be missed. Plenty of folks recognize sex differences, and plenty of MDs understand that difference when treating patients. Plenty of published research addresses these differences. Same with other subpopulations.
But I sense I am probably screaming into the void so have at it, Tps. Your turn.
Your knee jerk against bias could be considered for part of the first statement. Assuming authors are inherently, significantly biased to the point of invalidating research is silly. All the rest of the paragraph is reasonable and not "Librarian bias" at all.
I've done historical searches where sources from or covering minority communities are sparse or simply don't exist. I've run medical searches which have no studies of women available at all.
Thank you, Hana. It would be really helpful to have a citations list for posts like this one. Where did these definitions come from?
Excellent read - Thanks.
I am a retired Elec Engineer. Industrial control and automation, observer and occasional recreational user for research and coding. Mostly DeepSeek.
One thing I have noted is the obsequious content in a conversation. Presumably like click bait to bring the user back? Tells you what you want to hear.
On a bigger issue. Our climate clearly cannot afford the power use. Although I have seen different figures about this. Data centres just serving social media are the real culprits at least for now. What is the betting that breakthroughs are in the wings. I recall when the mapping of the human genome was first proposed it was thought to require immense computing power. It is now done routinely for a few$.
Good stuff, but speaking in the following way will confirm the inherent bias that librarians seem to be living with:
"where existing social inequalities, historical underrepresentation, and the demographics of whoever created and curated the data all leave their marks. A medical AI trained primarily on data from white male patients will perform less accurately when applied to women and people of color."
Less accuracy will occur IF and ONLY IF there ares actual differences in those subpopulations. Assumed difference is just bias. Here is a hint: when you use terms such as "underrepresentation", "social inequalities" you are dealing in rabbit in the hat putting. This is especially true with referencing the "demographics of the" creator of the data set.
This is a silly, silly comment. Male and female humans are incredibly different in their biology. Different races are susceptible to different difficulties. So, yes, medical data based on adult human males of Western European ancestry will be extremely inaccurate if applied to human females of African descent.
You are either a very ignorant human or a bot causing "outrage engagement."
Oh, I am not ignorant at all and I don't seek outrage, but you sure do jump to conclusions and have become outraged. I wrote that iff ( in logic, iff means if and only if) there is a difference - which you seem fixated on as to women (Restatement of the Obvious) - then that difference will matter. My point, which, whoosh, went right over your head was that IF there is a difference then a model only examining one sex will miss that difference. Well, OK, I did not lay out that latter point - I thought it rather obvious. But any one who enters data on only men will miss the differences as to women.
But my point was not about differences. It was about bias of the librarian author. There is the implicit assumption that differences must exist therefore will be missed. Plenty of folks recognize sex differences, and plenty of MDs understand that difference when treating patients. Plenty of published research addresses these differences. Same with other subpopulations.
But I sense I am probably screaming into the void so have at it, Tps. Your turn.
Your knee jerk against bias could be considered for part of the first statement. Assuming authors are inherently, significantly biased to the point of invalidating research is silly. All the rest of the paragraph is reasonable and not "Librarian bias" at all.
I've done historical searches where sources from or covering minority communities are sparse or simply don't exist. I've run medical searches which have no studies of women available at all.
This is amazing! Thank you for taking the time to compile the list. Much appreciated.