Data Science newsletter – April 26, 2019

GROUP CURATION: N/A

The University of Maryland is getting ready to open a new computer science center in College Park.

The Brendan Iribe Center for Computer Science and Engineering, named after the Oculus cofounder who has made multiple gifts toward tech education at the university, will officially be dedicated on April 27 at the corner of Baltimore Avenue and Campus Drive. The 215,600-square-foot building looks to provide room for students to work together and test out new ideas in classrooms and a variety of instructional space. Plus, there are common spaces and study-break areas with natural light.

There’s heartening news for our location privacy out of Massachusetts this week. The Supreme Judicial Court, the state’s highest court, ruled that police access to real-time cell phone location data—whether it comes from a phone company or from technology like a cell site simulator—intrudes on a person’s reasonable expectation of privacy. Absent exigent circumstances, the court held, the police must get a warrant.

In Commonwealth of Massachusetts v. Almonor, police had a phone carrier “ping” the cell phone of a suspect in a murder case—surreptitiously accessing GPS functions and causing the phone to send its coordinates back to the phone carrier and the police. This real-time location data pinpointed Mr. Almonor’s phone to a location inside a private home. The state argued it could warrantlessly get cell phone location data to find anyone, anytime, at any place as long as it was less than six hours old. A trial court disagreed and the state appealed.

In February, former NASA astronaut Mark Kelly announced his decision to run for the US Senate in Arizona — a move that he had been thinking about for the last couple of months. As the husband of former US Rep. Gabrielle Giffords (D-AZ), Kelly is no stranger to politics. If elected, he will join a very small group of astronauts who have transitioned from an orbital office to one on Capitol Hill.

The Verge spoke to Kelly about his path from astronaut to Senate candidate and how he plans to incorporate his scientific experience into politics.

The University of Maryland has announced a new $1 million initiative designed to promote diversity in the computer science department.

The Iribe Initiative for Inclusion and Diversity will expand services already available from the Maryland Center for Women In Computing, such as tutoring programs and summer camps. The services, previously offered through the center and used by female students, will now be marketed to a larger group of underrepresented minority groups.

When grand plans for a new Chicago-based Discovery Partners Institute were announced in October 2017, it caught many University of Illinois faculty members off guard.

The project had been quietly discussed in top administrative circles for months, but professors at Urbana and the other two UI campuses had many questions about what exactly it would be, how it would be governed, and what it would mean for their students and academic programs.

After more than a year of intense faculty consultation, the Urbana campus’s Academic Senate on Monday gave overwhelming approval to the institute, on a temporary basis. The vote was 104-3, with three abstentions.

Like many computer science students, Chelsey Serrano ’22 is finishing up a two-part CS intro sequence with help from close friends, who struggle together through late nights and stressful deadlines. But, as a first-generation and female student, Serrano more often than not finds herself in the minority; data shows that in 2018, close to 30 percent of CS degrees were awarded to women and 12 percent to historically underrepresented groups on campus — lagging behind the increasingly diverse demographics of the student body.

As the computer science community works to become more inclusive to underrepresented groups, its policies can make it difficult at times for all students to feel at home in the department. The sometimes severely strict no-collaboration policy, paired with a growing want for equal representation in CS, can create obstacles to a fully welcoming community. The increasing presence of affinity groups allows for students from underrepresented groups to find a place in the community, but some students say there is still more to be done. While no one student or voice can be representative of the experiences of everyone in the University’s largest department, stories from CS students illuminate both a sense of camaraderie and obstacles that can hinder a sense of belonging.

Researchers now have access to a flood of educational data on students that they hope will offer insights on how to improve the learning experience. Will it work?

The Trump administration’s attempt to add a citizenship question to the 2020 Census reaches the Supreme Court on Tuesday. Thus far, three federal judges have ruled against the Trump administration, most recently in Maryland. The Supreme Court will consider not only whether the administration violated administrative law, but also whether its attempt violated the Constitution.

A crucial issue in the case is whether adding this question for the first time since 1950 will hurt the ability of the census to accurately count the American population. In particular, critics of the administration fear the question will dissuade some U.S. residents, especially immigrants, from answering the census.

Research suggests these fears are justified. Working separately, we have used surveys and experiments to show that the citizenship question would make people less likely to respond to the census and provide complete information if they do respond. This is particularly true for Latinos and immigrants.

Hoffman discussing fairness and bias especially as it relates to violence latent in discussions of data discrimination

An Amazon.com Inc. team auditing Alexa users’ commands has access to location data and can, in some cases, easily find a customer’s home address, according to five employees familiar with the program.

The team, spread across three continents, transcribes, annotates and analyzes a portion of the voice recordings picked up by Alexa. The program, whose existence Bloomberg revealed earlier this month, was set up to help Amazon’s digital voice assistant get better at understanding and responding to commands.

Team members with access to Alexa users’ geographic coordinates can easily type them into third-party mapping software and find home residences, according to the employees, who signed nondisclosure agreements barring them from speaking publicly about the program.

Artificial intelligence and machine learning models can be found in almost every aspect of modern life. News-ranking algorithms determine which information we see online, compatibility algorithms influence the people we date, and ride-hailing algorithms affect the way we travel. Despite the pervasiveness of these life-changing algorithms, we don’t have a universal understanding of how they work or how they’re shaping our world.

So, a team of researchers—including two Northeastern University professors—says that it’s time to study artificially intelligent machines the way we study humans.

David Lazer one of the authors of the paper, is University Distinguished Professor of political science and computer and information sciences at Northeastern.

A new paper published Wednesday in the scientific journal Nature calls upon scientists from across various disciplines to unite in studying machine behavior. For years, scientists have studied the function, causes, development, and evolutionary history of human behavior. With intelligent machines doing more and more of our collective ‘thinking,’ the same interdisciplinary approach needs to be applied to understanding machine behavior, the authors say.

Over the past few years, big-name venture capitalists and flashy PR campaigns have boasted about replicating the Silicon Valley tech boom in second-tier cities around the country. But so far there is no stampede to the hinterland.

The big picture: Existing U.S. tech hubs are not only holding on to their imprimatur as primary magnets of top tech talent, but increasing it, persuading the best candidates to take jobs and stay put. Second-tier cities remain just that.

What’s happening: A small number of U.S. cities — Boston, Austin, and of course the soup of cities making up Silicon Valley — still host almost all the Big Tech headquarters, while parceling out mere morsels of satellite offices or data centers elsewhere, away from the action.

Mozilla today released the 2019 Internet Health Report, an analysis that brings together insights from 200 experts to examine issues central to the future of the internet. This year’s report chose to focus primarily on injustice perpetuated by artificial intelligence; what NYU’s Natasha Dow Schüll calls “addiction by design” tech, like social media apps and smartphones; and the power of city governments and civil society “to make the internet healthier worldwide.”

For years, laboratories have chafed against testing being, literally and figuratively, an out-of-sight, out-of-mind transaction. Now a new, highly visible era in genetics may be pushing testing the other way, into the hands of consumers who value entertainment as well as medical information. Anyone who wants to write a book about this shift has a ready-made title: From Basement to Big Top.

It’s not that clinical testing is becoming an actual circus. But ever since the first consumer genetic tests entered the market in 2007—in a nonphysician-ordered, SNP array technology way—labs, physicians, and regulatory agencies have had plenty to juggle. Today that includes relatively affordable sequencing, DNA ancestry searches, and patient empowerment. Throw in a little Silicon Valley verve, and we arrive at the present: presumably healthy consumers who want a peek at their own genetic profiles.

Some call this consumer-directed testing. Others prefer consumer-facing testing.

Erratic funding for basic research, coupled with concerns that science has fallen down the priority list for politicians, has contributed to a shift in researchers’ attitudes to lobbying.

New York, NY May 1, starting at 7 p.m., Columbia University.

London, England June 11, starting at 6 p.m., The Information Lab (25 Watling Street). “In our latest Data Book Club event, we’ll be hearing from Sir David Spiegelhalter about his book The Art of Statistics. The book guides us through the essential principles we need in order to derive knowledge from data, showing us why data can never speak for itself.” [$$]

“Oren Etzioni is chief executive of the non-profit Allen Institute for Artificial Intelligence (AI2) in Seattle, Washington, and is on leave from the nearby University of Washington. He offers some recommendations for how to stem the outflow of artificial-intelligence (AI) researchers from academia to industry — a loss that is damaging academia’s ability to teach incoming undergraduates.”

As part of their recent announcement of a new web design system for US government websites, the General Services Administration has also introduced a new typeface called Public Sans.

Over the past year, machine learning models have dramatically improved scores across many language understanding tasks in NLP. ELMo, BERT, ALICE, the model formerly known as BigBird (now MT-DNN), and OpenAI GPT have advanced a surprisingly effective recipe that combines language modeling pretraining on huge text datasets with simple multitask and transfer learning techniques that adapt the resulting model to downstream applications.

GLUE, released a year ago, is a benchmark and toolkit for evaluating recipes like these (think the Great British Baking Show meets Sesame Street). GLUE is a collection of nine (English) language understanding tasks — things like textual entailment, sentiment analysis, and grammaticality judgments — and was meant to cover a big enough swath of NLP so that the only way to do well at it was to build tools so general that they would help with most new language understanding problems that might come along.

We set out to see if we could build a better word cloud: retaining their visual appeal, but making them more comprehensible. Our thinking is if we select words more carefully and organize them by concept, that would lead to better understanding of the underlying topics of a document. We aren’t the first to suggest this. Jeff Clark, a master of text visualization, has a version of this idea for books, and the TopicPanorama project of Wang et al. has a word cloud composed of words drawn from up to four topics.

We tested the effectiveness of this idea through a sequence of careful studies. We showed that word clouds can be more understandable with just two major changes to how they are built today.

A standing desk shouldn’t replace sitting, but complement it. Here’s the right way to use one, plus accessories to make it even more effective.