You’re Not Stuck In Traffic You Are Traffic,
Source: The Journal of Design and Science, a collaboration between the Media Lab and MIT Press to try to bring design to science and science to design, to create a new kind of design for complex system self-adaptive systems and to bring this design to science.
Similarly, I think,
I am excited to be part of a growing community of researchers and practitioners concerned with fairness, accountability, and transparency in Machine Learning. Therefore I want to share links to some of the recources that have shaped my view of the field and ideas on new projects.
Goal #1: focus either on using Generative Neural Network or a Variational Autoencoder. Currently exploring the possibility of using GANs to better model the mapping between a contruct and observed data space. The work is inspired by some of the research done on evaluating fairness based on mappings between data spaces and distances between the initial datapoints and the outputs of a classifier system: On the (im)possibility of fairness
In Semi-supervised learning scenarios the data produced by a trained GAN generator network gets used as training data for a new classifier. Consider that we can condition the latent variables on the input of the GAN on a single or multiple protected parameters. Then the generated GAN output will belong to a specific subcluster inside the original real world data distribution. In this work we will focus on using these subclusters to create a metric for fairness of the target classifier and potentially feed that into a loss function to adjust the model parameters during training.
FAT/ML, a growing community of practitioners focussing on Fairness, Accountability and Transparency in ML
AI NOW, researching the social impacts of artificial intelligence now to ensure a more equitable future
Solon Barocas's class on Ethics and Policy in Data Science at Cornell, all course materials are available online and it is an amazing compilation of resources in the field.
CS 294: Fairness in Machine Learning by Moritz Hardt - the focus is on understanding and mitigating discrimination based on sensitive characteristics, such as, gender, race, religion, physical ability, and sexual orientation. Recent years have shown that unintended discrimination arises naturally and frequently in the use of machine learning and algorithmic decision making.
Weapons of Math Destruction by Cathy O’Neil
Raw Data Is an Oxymoron", a combination of essays from some of the top researchers in the field
The Ethics and Governance of Artificial Intelligence initiative, a collaboration between the Berkman Klein Center and MIT Media Lab
The projects at AI Advance, a Community Convening at Harvard Law School to advance the Ethics and Governance of Artificial Intelligence Initiative
J.N. Matias Thesis at MIT: Governing Human and Machine Behavior in an Experimenting Society