Data Science newsletter – March 24, 2021

Students say that working remotely makes it easier to use phones and notes during exams, and cite constraints in online learning as reasons to explain their behavior.

A study from Imperial College London found a near-200% increase in questions and answers posted to Chegg’s homework help section between April and August 2020.

Experts say the empirical data on Covid cheating is slim, but many students are doing it because during the pandemic remote learning shift they think no one is watching.

TODAY’s Savannah Sellers talks to several college students about the difficulties of online learning and why they are cheating on schoolwork during the pandemic. [video, 4:48]

Oh, good grief, here’s another one today: https://niusdiario.es/sociedad/sanidad/sanidad-reconoce-datos-muertes-ninos-covid-erroneos-contabilizaban-centenarios-como-menores_18_3107220241.html Spanish national COVID stats are wrong because of bad spreadsheet formatting, creating a surge of child/infant deaths that did not actually exist.

The new EPFL Center for Imaging (ECI) follows the Imaging@EPFL initiative that was launched in 2019 to promote interdisciplinary expertise in imaging on campus. “The work done with Imaging@EPFL really laid the bases for a proper center with increased reach and resources,” says Dr. Laurène Donati, ECI’s new Executive Director. The academic governance of the initiative also remains in place: Prof. Michael Unser, the new Academic Director of the ECI, keeps chairing a steering committee composed of five faculty members representing the different Schools.

Shatner announced today that he has joined Los Angeles-based tech company StoryFile as brand ambassador for its conversational video platform StoryFile Life. The sci-fi icon, best known for his role as Captain James T. Kirk in the first-ever Star Trek series, is the first person to use StoryFile Life to create an AI-powered interactive conversational video so family and friends can interact with him for years to come.

A new study from Microsoft, released Monday, found that among the more than 31,000 workers it surveyed, 73% hoped remote work options would continue when the pandemic ends. Even Gen Z applicants were slightly more likely to apply for a job with remote options than for one strictly in an office. But those workers are also facing particular drawbacks.

Gen Z workers, born roughly between the mid-1990s and mid-2010s, responded to Microsoft’s surveys generally by saying they’re more stressed and find they’re struggling more than their peers. They tend to be single, since they’re younger, leading them to feel isolated. And since they’re early in their careers, they don’t have financial means to create a good workspace at home if their employer won’t pay for it. And they’re not having those in-person meetings that sometimes help them land in career advancing projects, or even to get in good with the boss.

Most major global companies no longer plan to reduce their use of office space after the coronavirus pandemic, though few expect business to return to normal this year, a survey by accountants KPMG showed on Tuesday.

Just 17% of chief executives plan to cut back on offices, down from 69% in the last survey in August.

In August of 2019, I published research from now-defunct clickstream data provider, Jumpshot, showing that 50.33% of all Google searches ended without a click to any web property in the results. Today, thanks to new data from SimilarWeb, I’ve got a substantive update to that analysis.

From January to December, 2020, 64.82% of searches on Google (desktop and mobile combined) ended in the search results without clicking to another web property. That number is likely undercounting some mobile and nearly all voice searches, and thus it’s probable that more than 2/3rds of all Google searches are what I’ve been calling “zero-click searches.” Some folks have pointed out that “zero-click” is slightly misleading terminology, as a search ending with a click within the Google SERP itself (for example, clicking on the animal sounds here or clicking a phone number to dial a local business in the maps box) falls into this grouping. The terminology seems to have stuck, so instead I’m making the distinction clear.

A recorded reference to a “hedge” appeared in Old English in 785, around the time the Vikings began to ransack northern England. Back then, a hedge was a boundary planted by man. It was the basic demarcation of property and of ownership, the line between what is mine and what is yours.

The definition of hedge has expanded since then, but not our general understanding of hedge funds. The wall that separates the rest of us from those who buy and sell billions on a daily basis has, if anything, become increasingly fortified—and for good reason. By the time the word had migrated into the lexicon of modern English, the line between “mine” and “not mine” had waxed militaristic, and the definition of hedge came to include arsenals of physical defense, like hedges of archers. At this point in linguistic history, one man’s hedge became another man’s extreme bad luck.

< Another half a millennium would pass before hedging merged with gambling. In 1672, the phrase “to hedge a bet” first appears, with an implication of shaky moral standing. The hedged bet was for rooks, and Shakespearean English abounds with hedge wenches, hedge cavaliers, hedge doctors, hedge lawyers, hedge writers, hedge priests, and hedge wine. Some of the earliest examples of investor hedging appear a few decades later in the coffeehouses of London’s Exchange Alley, where caffeine-addled proto-brokers bet on the movement of equity stakes in the Bank of England, the South Sea Company, and the British East India Company. Investors who owned shares could offset the possibility of future losses by hedging, for example by betting in advance on the stock’s downward movement, a strategy recently duplicated by hedge funds unfortunate enough to have “sold short”—that is, to bet against the price—of GameStop.

According to [Niyant] Narang, Discover Together stresses the discovery aspect of finding new music, while other third-party Spotify services recommend songs to one user that the other likes, or finds common songs between both users’ playlists.

“This is really unique in the sense that it’s not doing either, it’s kind of a little bit of both, where it’s not only mixing your music tastes, but then finding songs that you probably haven’t listened to that you would still love,” said Narang.

Since the algorithm analyzes each user’s top 50 songs, one challenge Reddy noticed was that less relevant genres in a user’s listening history were getting recommended equally as often as more relevant genres, so the team had to make adjustments to factor in genre relevance to ensure that only the most suitable songs would be recommended. Recently, Discover Together passed the milestone of reaching 1,500 playlists made, and Reddy is eagerly awaiting the day 10,000 playlists have been created.

In an ideal world, we would have ample justice-focused technologies from which to choose; however, the reality is usually much more complicated. In practice, the answers to these questions are often disappointing, even for the tools we ultimately choose.

First, it is nearly impossible to find real-world examples of digital tools that check all the boxes of anti-oppression, justice and equity. For example, platforms with end-to-end encryption (a technology that is pro-privacy and anti-surveillance) are often used in right-wing extremist organising, because they have more robust built-in features around privacy than their counterparts. Messaging services like Signal and Telegram, for instance, saw a massive increase in downloads earlier this year, after the purge of far-right content from mainstream social networks and in response to growing concerns around Whatsapp’s privacy practices. But despite the ways that this technology has been used, strong pro-privacy design is nonetheless something we prioritise at The Engine Room as a way of protecting vulnerable communities and frontline activists.

Second, activist groups usually will have complex internal needs and dynamic external visions, for which a specific technology may not exist.

Joy Buolamwini is a researcher at the MIT Media Lab who pioneered research into bias that’s built into artificial intelligence and facial recognition. And the way she came to this work is almost a little too on the nose. As a graduate student at MIT, she created a mirror that would project aspirational images onto her face, like a lion or tennis star Serena Williams.

But the facial-recognition software she installed wouldn’t work on her Black face, until she literally put on a white mask. Buolamwini is featured in a documentary called “Coded Bias,” airing tonight on PBS. She told me about one scene in which facial-recognition tech was installed at an apartment complex in the Brownsville neighborhood of Brooklyn, New York. The following is an edited transcript of our conversation.

Even if you can’t name the tunes, you’ve probably heard them: from the iconic “dun-dun-dun-dunnnn” opening of Beethoven’s Fifth Symphony to the melody of “Ode to Joy,” the German composer’s symphonies are some of the best known and widely performed in classical music.

Just as enthusiasts can recognize stylistic differences between one orchestra’s version of Beethoven’s hits and another, now machines can, too.

A Duke University team has developed a machine learning algorithm that “listens” to multiple performances of the same piece and can tell the difference between, say, the Berlin Philharmonic and the London Symphony Orchestra, based on subtle differences in how they interpret a score.

Data science is a critical yet less recognized driver in the state’s COVID-19 vaccine rollout, according to West Virginia University.

WVU data scientists Brad Price and Katherine Kopp joined with Dariane Drake to form “Data Driven WV” in 2019. The group is helping shape the state in a transformative way as West Virginians count on COVID-19 vaccines to move the Mountain State past a pandemic.

Price and Kopp have implemented data science to support vaccine supply chain efforts by developing dashboards for the Department of Health and Human Resources around populations in West Virginia, including how to identify those who are at-risk.

The Governance lab (The GovLab) at New York University’s Tandon School of Engineering will launch a national online citizen engagement initiative as part of ReinventED: Your Education, Your Voice, a campaign to engage students, parents, caregivers and educators in identifying challenges with today’s education system. Results will be openly published and developed into recommendations targeted at policymakers and philanthropic leaders working in the education space.

Codesigned by the project’s Student Council, 20 students aged 14-21 who helped to write and edit content for the engagement, the initiative will run for three weeks from March 22 through April 11. The students, who hail from 14 states, were recruited in a competitive, national selection process in December 2020. They meet every other week to help draft and refine the content and “focus group” it with their own parents to ensure that topics and language are accessible and relevant.

… What happens instead when we try to multitask, says Gloria Mark, Ph.D., is that our brains switch among tasks, requiring more brain fuel than staying with one task at a time.

“Every activity we do uses a different set of cognitive resources,” says Mark, an informatics professor at the University of California, Irvine. “If I do email, I’m using one set of cognitive resources. If I’m reading a report, I’m using a different set of resources. ”

The more tasks you try to do at any given time, the more cognitive energy you burn.

Today, 25% of figures in biomedical publications contain images of various types, e.g. photos, light or electron microscopy images, x-rays, or even sketches or drawings. Despite being widely used, published images may be ineffective or illegible since details are not visible, information is missing or they have been inappropriately processed. The vast majority of such imperfect images can be attributed to the lack of experience of the authors as undergraduate and graduate curricula lack courses on image acquisition, ethical processing, and visualization.
Here we present a step-by-step image processing workflow for effective and ethical image presentation. The workflow is aimed to allow novice users with little or no prior experience in image processing to implement the essential steps towards publishing images. The workflow is based on the open source software Fiji, but its principles can be applied with other software packages. All image processing steps discussed here, and complementary suggestions for image presentation, are shown in an accessible “cheat sheet”-style format, enabling wide distribution, use, and adoption to more specific needs.