-
Notifications
You must be signed in to change notification settings - Fork 2
Jones, Gabriel
License
bigdata-i523/hid104
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
--- owner: hid: 104 name: Jones, Gabriel url: https://github.com/bigdata-i523/hid104 paper1: abstract: > We breifly analyze the history of data to show how having Lots of Data hardly differs from data storage and analysis in the early days of SQL, or even before computers. We then explain how Big Data represents a paradigmatic shift from conventional data analysis. We then begin to look at the potential limits of Big Data to assert that this paradigmatic shift does not mean the end of science. We conclude that misunderstanding Big Data prevents organizations from capitalizing on its potential and can lead them to spurious answers. author: - Jones, Gabriel chapter: Theory hid: - 104 status: Oct 28 17 100% title: What Separates Big Data from Lots of Data? url: https://github.com/bigdata-i523/hid104/tree/master/paper1 paper2: review: Nov 10 2017 abstract: > Since its origins, Big Data has promised an unimaginable potential to revolutionize the world. Scholars have wisely noted that it represents a paradigmatic shift from conventional norms of data, but the public has quickly latched onto provocative but unrealistic narratives that deify big data as omniscient, infallible, and devoid of bias. Confiding in such narratives diminishes the integrity of credible science and poses serious ethical challenges, but these challenges are more likely overlooked because the very narratives seem to reject the need for ethical discussion. The authors argue that such blind optimism will cause irreversible damage to society if left unchecked. First we debunk the fallacious narratives people tend to tell about big data, offering a more realistic discussion of its merits and its limitations. We then explore how analytical or algorithmic bias and sampling bias, two problems that statisticians have faced since long before the onset of big data, present pitfalls for deriving knowledge from data. We examine how the ethical implications of these pitfalls can cause serious damage in society. We conclude that Big Data analysis must obey the principles of transparency, clear and appropriate objective definition, and self-correcting feedback mechanisms. author: - Jones, Gabriel - Millard, Mathew hid: - 104 - 216 status: Nov 10 17 100% title: "Big Data = Big Bias? The Fallibility of Big Data" chapter: Theory url: https://github.com/bigdata-i523/hid104/tree/master/paper2 project: abstract: > While Big Data can make the world a better place, blind optimism in its infallibility can cause irreversible damage to society if left unchecked. With the mission of ensuring accountability, we debunk the fallacious narratives people tend to tell about Big Data, offering a more realistic discussion of its merits and its limitations. We then explore how analytical or algorithmic bias and sampling bias, two problems that statisticians have faced since long before the onset of Big Data, present pitfalls for deriving knowledge from data. We examine how the ethical implications of these pitfalls can cause serious damage in society. We determine that effective, credible, and ethically sound Big Data analysis must obey the principles of transparency, clear and appropriate objective definition, and self-correcting feedback mechanisms. We examine case studies where academicians and businesses have tested algorithms to study how well they exhibit these principles. We then implement our own test to check for potential algorithmic bias in Google. Based on evidence that certain individuals have been corrupted in part by Google searches allegedly bias against racial minorities, we hypothesize that Google's algorithms systematically exhibit biases against minority groups. We test this hypothesis by examining how Google search suggestions associate certain negative words with names that typically belong to minority groups. We conclude that while our study alone cannot prove or disprove our argument, the evidence in our analysis contradicts our hypotheses, thus suggesting that no systematic bias is exhibited. We discuss end by discussing what the results could mean for future studies of potential algorithmic bias in Google. author: - Jones, Gabriel - Millard, Mathew hid: - 104 - 216 status: 100% title: Big Bias? An Analysis of Google Search Suggestions type: report url: https://github.com/bigdata-i523/hid104/tree/master/project chapter: Technology
About
Jones, Gabriel
Resources
License
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published