Convert 'Fairness in ML' to BibTeX #1044
Conversation
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
|
This should probably go in before #1042 (because when I add citations to that, I really don't want to do so by hand) |
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
| address = {New York, NY, USA}, | ||
| url = {https://doi.org/10.1145/3287560.3287598}, | ||
| doi = {10.1145/3287560.3287598}, | ||
| abstract = {A key goal of the fair-ML community is to develop machine-learning based systems that, once introduced into a social context, can achieve social and legal outcomes such as fairness, justice, and due process. Bedrock concepts in computer science---such as abstraction and modular design---are used to define notions of fairness and discrimination, to produce fairness-aware learning algorithms, and to intervene at different stages of a decision-making pipeline to produce "fair" outcomes. In this paper, however, we contend that these concepts render technical interventions ineffective, inaccurate, and sometimes dangerously misguided when they enter the societal context that surrounds decision-making systems. We outline this mismatch with five "traps" that fair-ML work can fall into even as it attempts to be more context-aware in comparison to traditional data science. We draw on studies of sociotechnical systems in Science and Technology Studies to explain why such traps occur and how to avoid them. Finally, we suggest ways in which technical designers can mitigate the traps through a refocusing of design in terms of process rather than solutions, and by drawing abstraction boundaries to include social actors rather than purely technical ones.}, |
There was a problem hiding this comment.
No; I simply didn't go to the effort of removing it.
There was a problem hiding this comment.
But... shouldn't we? Do we really want to keep abstracts of papers here? It's not like we're going to render those...
There was a problem hiding this comment.
We could trim it; I was just doing the simplest thing based on the citation source.
There was a problem hiding this comment.
If we're not rendering it let's just cut it(?)
There was a problem hiding this comment.
I'm happy to have as few modifications as possible on the copy pasted chunk. Also, this doesn't take that much space, why bother.
There was a problem hiding this comment.
Since we are providing URL, I would personally lean towards skipping most of the extra meta-information here like isbn, abstract, numpages, keywords. But I don't care too much about extra meta-information that doesn't get displayed.
Two things that I think we should care about are:
- consistency in the naming of the bibtex ids (format , where keyword is most of the time the first word in the paper title)
- consistency in the appearance of the references at the end of files (this is something that is actually not too hard to fix later, because we only need to edit .bib file...)
- linking to peer-reviewed versions of papers rather than arXiv whenever peer-reviewed versions are available
There was a problem hiding this comment.
Skipping metainformation is the business of the bibstyle file. Isn't that pretty much the whole point of BibTeX?
There was a problem hiding this comment.
Perhaps #1048 is a good place to continue this conversation. @riedgar-ms I agree except for the abstract (which is neither relevant nor is there even a remote chance of it ever being rendered, plus it's massive). I wrote that in #1048, too, so feel free to object there 😆
There was a problem hiding this comment.
The only issue that I see re. consistency is that these bibtex entries are generating multiple URLs when rendered--multiple URLs are also confusing in their own right (what are we actually referring to?). I'm in favor of dropping "eprint" and "doi" and retaining URL. (I think that's simpler that trying to tweak the stylefile?)
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
MiroDudik
left a comment
There was a problem hiding this comment.
I cannot check how these render... can you take a snippet and post in a comment or one of the replies?
But, based on looking at bibtex entries below, I'd be in favor of something more condensed (e.g., that refers to the venue as FAT* and the publisher as ACM). The easiest way to construct that is to basically take the condensed DBLP entry:
But take the URL from the standard DBLP entry:
Sorry for being a pest, but I think it would be good to refer to academic papers consistently, and this one is not too hard.
| address = {New York, NY, USA}, | ||
| url = {https://doi.org/10.1145/3287560.3287598}, | ||
| doi = {10.1145/3287560.3287598}, | ||
| abstract = {A key goal of the fair-ML community is to develop machine-learning based systems that, once introduced into a social context, can achieve social and legal outcomes such as fairness, justice, and due process. Bedrock concepts in computer science---such as abstraction and modular design---are used to define notions of fairness and discrimination, to produce fairness-aware learning algorithms, and to intervene at different stages of a decision-making pipeline to produce "fair" outcomes. In this paper, however, we contend that these concepts render technical interventions ineffective, inaccurate, and sometimes dangerously misguided when they enter the societal context that surrounds decision-making systems. We outline this mismatch with five "traps" that fair-ML work can fall into even as it attempts to be more context-aware in comparison to traditional data science. We draw on studies of sociotechnical systems in Science and Technology Studies to explain why such traps occur and how to avoid them. Finally, we suggest ways in which technical designers can mitigate the traps through a refocusing of design in terms of process rather than solutions, and by drawing abstraction boundaries to include social actors rather than purely technical ones.}, |
There was a problem hiding this comment.
Since we are providing URL, I would personally lean towards skipping most of the extra meta-information here like isbn, abstract, numpages, keywords. But I don't care too much about extra meta-information that doesn't get displayed.
Two things that I think we should care about are:
- consistency in the naming of the bibtex ids (format , where keyword is most of the time the first word in the paper title)
- consistency in the appearance of the references at the end of files (this is something that is actually not too hard to fix later, because we only need to edit .bib file...)
- linking to peer-reviewed versions of papers rather than arXiv whenever peer-reviewed versions are available
Signed-off-by: Richard Edgar <riedgar@microsoft.com>
|
The current rendering on my machine: Is ACM not considered a canoncial source of bibliographic information? |
MiroDudik
left a comment
There was a problem hiding this comment.
This looks good... (modulo the merge). But if you end up ensuring that there's only one URL associated with each reference (by removing doi and eprint), I definitely won't mind :-)... but it's a low order concern for me.
|
@MiroDudik feel free to make any small adjustments in a new PR! |
Bring civilised citations to the 'Fairness in Machine Learning' page in the documentation. Also do some reformatting of the ReST file, to reduce the length of a selection of lines.
Related to #936