ScienceDaily: Researchers devise approach to reduce biases in computer vision data sets. “Addressing problems of bias in artificial intelligence, computer scientists from Princeton and Stanford University have developed methods to obtain fairer data sets containing images of people. The researchers propose improvements to ImageNet, a database of more than 14 million images that has played a key role in advancing computer vision over the past decade.”
IdeaStream: Ohio’s Judges Considering Statewide Sentencing Database. “Members of Ohio’s judicial system are calling for more uniformity in sentencing practices across courtrooms. The state’s criminal sentencing commission argues an online database of previous sentences could aid in that effort.”
NiemanLab: Americans of all political stripes expect 2020’s fake news to be biased against their side. “Fake news, misinformation, and disinformation will be major concerns in the 2020 presidential election. According to previous research by the Pew Research Center, half of American adults describe misinformation as a ‘very big problem’ — more than who say the same about climate change, racism, and terrorism (though fewer than who say healthcare affordability, the wealth gap, and drug addiction).”
Governing: University Offers Free Class on Artificial Intelligence Ethics. “The course — developed by [Nathan] Colaner, law professor Mark Chinen and adjunct business and law professor Tracy Ann Kosa — explores the meaning of ethics in AI by looking at guiding principles proposed by some nonprofits and technology companies. A case study on facial recognition in the course encourages students to evaluate different uses of facial-recognition technology, such as surveillance or identification, and to determine how the technology should be regulated.” The course is being offered by Seattle University.
NiemanLab: The fact-checker’s dilemma: Humans are hardwired to dismiss facts that don’t fit their worldview. “Motivated reasoning is what social scientists call the process of deciding what evidence to accept based on the conclusion one prefers. As I explain in my book The Truth About Denial: Bias and Self-Deception in Science, Politics, and Religion, this very human tendency applies to all kinds of facts about the physical world, economic history and current events.”
DigitalNC: We Can Do Better: Making Our Metadata More Equitable. “Over the last few months I’ve been working on a pilot project that looks at how NCDHC staff have portrayed women through metadata (the information that accompanies the images on DigitalNC) over time. This is a small step towards finding unconscious bias in our work and making our metadata more equitable. I’ve accumulated some interesting examples, and I thought I’d share them here.”
Ars Technica: Meta-analysis study indicates we publish more positive results. “While science as a whole has produced remarkably reliable answers to a lot of questions, it does so despite the fact that any individual study may not be reliable. Issues like small errors on the part of researchers, unidentified problems with materials or equipment, or the tendency to publish positive answers can alter the results of a single paper. But collectively, through multiple studies, science as a whole inches towards an understanding of the underlying reality.”