Assignment due 10/30 on "A Convex Framework for Fair Regression."
Link to the assignment here.
A Convex Framework for Fair Regression
List the 2-3 primary publication venues related to the topic of the paper you read. These may be conferences or journals. For each, give your reasoning about what makes you think this is a primary publication venue.
Make a clustered* list of at least 10 (more is better!) papers from the venues you cited in part 1 that you think are worth exploring that seem to either build on the work from your original paper or present an alternate solution to the same problem. These papers must be from authors other than the original authors of your paper. In your list, include at least the title and authors of these papers and a link to each paper, as well as a short note about how the paper relates to your original paper (1-2 sentences).
Learning Optimal and Fair Decision Trees for Non-Discriminative Decision-Making
Sina Aghaei, Mohammad Javad Azizi, Phebe Vayanos
The authors build upon the in-processing approach taken by Berk et al. on preventing discrimination, and also use the general approach of using a flexible framework for models to better define fairness.
Optimization Hierarchy for Fair Statistical Decision Problems
Anil Aswani, Matt Olfat
This paper introduces a hierarchy for deciding which optimization technique to use for various types of models. The authors emphasize how the methods of one of their own models is similar to the one introduced by Berk et al., in that it is in-training and has a similar regularizer. However, they note that despite the benefits that come with its convexity, the model of Berk et al. is insufficient for large datasets.
Fair Regression for Health Care Spending
Anna Zink, Sherri Rose
The authors apply the work of Berk et al. specifically to the issue of risk assessment in health insurance and the undercompensation of those groups, leading to their missing health coverage.
Avoiding Resentment Via Monotonic Fairness
Guy W. Cole, Sinead A. Williamson
The authors of this paper use the same fairness-by-regularization approach as Berk et al., but do so in the context of classification instead of regression.
General Fair Empirical Risk Minimization
Luca Oneto, Michele Donini, Massimiliano Pontil
Similarly to the others, this paper follows the model of Berk et al. in that they apply fairness regularizers in-process. However, they diverge from Berk et al. in that they consider the situation where either the protected variable or the response variable has a real value.
Fair Algorithms for Learning in Allocation Problems
Hadi Elzayn, Shahin Jabbari, Christopher Jung, Michael Kearns, Seth Neel, Aaron Roth, Zachary Schutzman
The authors analyze notions of fairness in the context of resource allocation (as opposed to the classification or regression). The specific situation they use is the allocation of police to specific districts, where the goal is the number of crimes caught.
A General Framework for Fair Regression
Jack Fitzsimons, AbdulRahman Al Ali, Michael Osborne, Stephen Roberts
This paper uses the work of Berk et al. to further research into group fairness in the context of decision trees.
Incorporating Latent Meanings of Morphological Compositions to Enhance Word Embeddings
Yang Xu, Jiawei Liu, Wei Yang, Liusheng Huang
This paper does not build on the research done by A Convex Framework for Fair Regression, but is very relevant to our specific research topic because it discusses NLP in the context of language. Current state-of-the-art word embedding methods typically capture semantic information based on words. The newest models go one step deeper, to morphemes; yet the authors propose that even this is not deep enough. Instead, word embeddings should be trained on the latent meanings **of morphemes.
What Kind of Language is Hard to Language-Model?
Sebastian J. Mielke, Ryan Cotterell, Kyle Gorman, Brian Roark, Jason Eisner
This paper also does not directly reference or build on the work done by A Convex Framework for Fair Regression, but it is extremely significant to our own topic. Our research analyzes the varying efficacy of current NLP models on different languages, and this paper does the same.
Aldrian Obaja Muis, Naoki Otani, Nidhi Vyas, Ruochen Xu, Yiming Yang, Teruko Mitamura, Eduard Hovy
Again, this paper is not fundamentally similar to A Convex Framework for Fair Regression, but it still uses fairness as it studies the impacts of low-resource data on the models that use them for training.
Cluster (group) these papers by how related they are to each other. E.g. some will extend the work in one direction while others will go in a different direction. Try to put some high level order in the related work space.
Concerning specific instances of fairness
Concerning language and word embeddings
What Kind of Language is Hard to Language-Model?
Incorporating Latent Meanings of Morphological Compositions to Enhance Word Embeddings
Using Berk et al. to advance specific areas
Decision trees: A General Framework for Fair Regression
Classification: Avoiding Resentment Via Monotonic Fairness
Resource allocation: Fair Algorithms for Learning in Allocation Problems
Furthering the same line of research
Optimization Hierarchy for Fair Statistical Decision Problems
General Fair Empirical Risk Minimization
Learning Optimal and Fair Decision Trees for Non-Discriminative Decision-Making