Author: N/A; Publisher: Wired; Publication Year: 2018. The following video discusses how cops in Los Angeles (& many places) are using predictive algorithms in policing. Proponents say it is effective at reducing crime, but many people argue that it helps police justify racial profiling. The data is biased because policing statistics have always disproportionately targeted minorities, so these algorithms can not…
Author: Wenhong Chen, Anabel Quan-Haase; Publisher: Social Science Computer Review; Publication Year: 2018. The following article addresses new issues created by big data, such as biases, subjectivities, and forms of oppression. They define 4 major aspects of big data ethics and politics: 1). Potential biases in big data collection and interpretation, 2). Community and citizen concerns of big data (mis)use in public life and for journalistic purposes, 3). Media…
Author: Vidhi Chugh; Publisher: Medium; Publication Year: 2021. The following article describes how as a data scientist, if you want the product you are working on to be “[artificial intelligence (AI)] for good,” then it is inevitable to encounter some situations such as biased data, biased models, or biased implementation. This article illustrated forms of what data bias can be and there is no one solution for all. But…
Author: Ugonma Nwankwo, Michael Pisa; Publisher: Center for Global Development; Publication Year: 2021. The following article discusses how evidence-based policy making is ineffective when it relies on biased information, a potential source for this is bias in the datasets. In the U.S., women make up 18% of data scientist jobs and in lower-income countries, that stat is even worse. In the data value chain, which includes collection, publication, uptake and…
Author: Tolga Bolukbasi, Kai-Wei Chang, James Zou, et al.; Publisher: Arxiv; Publication Year: 2016. The following article discusses how machine learning applied blindly runs the risk of amplifying biases in data. Word embedding, a popular framework for representing text data as vectors that has been used in many machine learning and natural language processing tasks, poses such a risk. The authors demonstrate that even word embeddings…
Author: N/A; Publisher: Synced; Publication Year: 2020. The following news article talks about Genderify, an artificial intelligence (AI)-powered tool designed to identify a person’s gender by analyzing their name, username or email address — has been completely shut down. Genderify was met with fervent criticism on Twitter, with many decrying what they saw as built-in biases. Entering the word “scientist”…
Author: Rumman Chowdhury; Publisher: Accenture Technology; Publication Year: 2019. In the following video, Dr. Ruman Chowdury explains how “[artificial intelligence (AI)] is information about people meant to understand trends about human behavior.” There are 2 kinds of bias: bias in data and models and bias in the imperfect world. Ethics is not just about improving technology but improving the society behind the technology. Technologists…
Author: Robin Hauser; Publisher: TED; Publication Year: N/A. In the following talk, Robin Hauser talks about unconscious bias in artificial intelligence (AI) algorithms. As part of producing her new movie about unconscious bias, she became interested in finding out whether it would be possible to create AI without bias. As she came to find out, it is oftentimes harder to create unbiased algorithms for multiple…
Author: Prabhakar Krishnamurthy; Publisher: Medium; Publication Year: 2019. The following article discusses how most datasets suffer from bias which can affect conclusions drawn from the data in a way that is discriminatory. This paper describes different types of bias and how it may arise. Knowing the sources of bias can help us mitigate their effect or improve processes to collect data to use in modeling. 2 additional…
Author: Peter Haas; Publisher: TEDx Talks; Publication Year: 2017. The following article features Peter Haas, who actually works in robotics at Brown University, who is afraid of robots. The example of how to train a model to classify wolf and Husky illustrates that there exists bias in the data set that was fed to the algorithm. This example proved the importance of researchers/developers who working on artificial…