Data Science newsletter – October 16, 2020

GROUP CURATION: N/A

An electronic nudge to clinicians—triggered by an algorithm that used machine learning methods to flag patients with cancer who would most benefit from a conversation around end-of-life goals—tripled the rate of those discussions, according to a new prospective, randomized study of nearly 15,000 patients from Penn Medicine and published today in JAMA Oncology.

Early and frequent conversations with patients suffering from serious illness, particularly cancer, have been shown to increase satisfaction, quality of life, and care that’s consistent with their values and goals. However, today many do not get the opportunity to have those discussions with a physician or loved ones because their disease has progressed too far and they’re too ill.

“Within and outside of cancer, this is one of the first real-time applications of a machine learning algorithm paired with a prompt to actually help influence clinicians to initiate these discussions in a timely manner, before something unfortunate may happen,” said co-lead author Ravi B. Parikh, MD,

In late 2019, Betelgeuse, the star that forms the left shoulder of the constellation Orion, began to noticeably dim, prompting speculation of an imminent supernova. If it exploded, this cosmic neighbor a mere 700 light-years from Earth would be visible in the daytime for weeks. Yet 99% of the energy of the explosion would be carried not by light, but by neutrinos, ghost-like particles that rarely interact with other matter.

If Betelgeuse does go supernova soon, detecting the emitted neutrinos would “dramatically enhance our understanding of what’s going on deep inside the core of a supernova,” said Sam McDermott, a theorist with the Fermi National Accelerator Laboratory.

The interviews and documents obtained by Science show [Deborah] Birx replaced a functional, if imperfect, CDC data system—well understood by hospitals and state health departments—with an error-ridden and unreliable filter on hospital needs that sometimes displays nonsensical data, such as negative numbers of beds. Such problems could hamper effective distribution of federal resources during an anticipated fall and winter spike in COVID-19 and flu cases, CDC officials say.

“This is the surreal part of it: They are attempting to replicate something we built over 15 years. And they are failing,” says a high-level CDC official with personal knowledge of the system. “Either Birx isn’t looking at the data, or she’s looking away—because it’s an absolute disaster.”

Tufts University School of Engineering is collaborating with Noodle Partners, a leading online program manager (OPM), to launch a new online Master of Science in Data Science program and a Post-Baccalaureate in Computer Science. The programs are expected to launch in January 2021 with classes beginning in Fall 2021.

The Master of Science program in Data Science is designed to prepare students who have earned bachelor’s degrees in STEM fields for advanced careers in data analysis and data-intensive science. The program focuses on statistics and machine learning, with courses in data infrastructure and systems, data analysis and interfaces, and theoretical elements.

The Post-Baccalaureate program in Computer Science is open to individuals with at least a bachelor’s degree in any discipline (BA or BS) and one college-level introductory computer course.

Scientists from Oxford University have used a machine learning algorithm to discover every lion has its own identifiable and trackable roar.

Previous research has shown that lions roar to communicate with other members of their pride and scare off foes. But we still know little about how they recognize which animal has made the call.

More and more companies are leveraging technological advances in machine learning, natural language processing, and other forms of artificial intelligence to provide relevant and instant recommendations to consumers. From Amazon to Netflix to REX Real Estate, firms are using AI recommenders to enhance the customer experience. AI recommenders are also increasingly used in the public sector to guide people to essential services. For example, the New York City Department of Social Services uses AI to give citizens recommendations on disability benefits, food assistance, and health insurance.

However, simply offering AI assistance won’t necessarily lead to more successful transactions. In fact, there are cases when AI’s suggestions and recommendations are helpful and cases when they might be detrimental. When do consumers trust the word of a machine, and when do they resist it? Our research suggests that the key factor is whether consumers are focused on the functional and practical aspects of a product (its utilitarian value) or focused on the experiential and sensory aspects of a product (its hedonic value).

In an article in the Journal of Marketing — based on data from over 3,000 people who took part in 10 experiments — we provide evidence supporting for what we call a word-of-machine effect: the circumstances in which people prefer AI recommenders to human ones.

Scientists at the Lawrence Berkeley National Laboratory, or Berkeley Lab, have designed an Automated Recommendation Tool, or ART, that seeks to revolutionize the field of bioengineering by employing a systematic and automated approach to synthetic biology.

The field of synthetic biology entails generating “novel valuable molecules,” such as biofuels or anti-cancer drugs, according to a study conducted by Berkeley Lab researchers. Scientists face limitations, however, because of the years-long effort it requires to understand and manipulate a cell’s biochemistry to produce a desired outcome.

ART pairs synthetic biology with machine learning algorithms to allow for the automated metabolic engineering of a cell’s DNA and the resultant creation of cells, according to a Berkeley Lab press release.

“People are making synthetic milk, burgers, spider silk, collagen and biomaterials,” said Héctor García Martin, study co-author and staff scientist in Berkeley Lab’s Biological Systems and Engineering, or BSE, division. “This is just a tool that makes bioengineering easier because it really takes out the problem of ingesting all that data and deciding what the next step is.”

The COVID-19 (coronavirus disease 2019) pandemic is a collective stressor unfolding over time; yet, rigorous empirical studies addressing its mental health consequences among large probability-based national samples are rare. Between 18 March and 18 April 2020, as illness and death escalated in the United States, we assessed acute stress, depressive symptoms, and direct, community, and media-based exposures to COVID-19 in three consecutive representative samples from the U.S. probability-based nationally representative NORC AmeriSpeak panel across three 10-day periods (total N = 6514). Acute stress and depressive symptoms increased significantly over time as COVID-19 deaths increased across the United States. Preexisting mental and physical health diagnoses, daily COVID-19–related media exposure, conflicting COVID-19 information in media, and secondary stressors were all associated with acute stress and depressive symptoms. Results have implications for targeting public health interventions and risk communication efforts to promote community resilience as the pandemic waxes and wanes over time. [full text]

Every single patient that heads to a RWJBarnabas Health facility will be soon be screened for social determinants of health and, if needed, referred to services for ongoing support, officials announced Tuesday.

SDOHs are social or environmental factors in a patient’s life—such as their access to healthy food, safe housing or transportation—that can ultimately impact their health.

Officials from the New Jersey-based health system say the program, called Health Beyond the Hospital (HBTH), will be the first “end-to-end, universally applied, culturally-tailored and fully integrated” SDOH program in the country.

September 2020 was the warmest September since global record keeping began in 1880, NOAA’s National Centers for Environmental Information, NCEI, reported October 14.

The month was just 0.02 degrees Celsius warmer than the previous record, held jointly by September 2015 and 2016. NASA and the European Copernicus Climate Change Service also rated the month as the warmest September on record, and the Japan Meteorological Agency rated it as the third-warmest September on record. Minor differences in rankings often occur among various research groups, the result of different ways they handle data-sparse regions such as the Arctic.

The 197 radio astronomy dishes of the Square Kilometre Array (SKA) in South Africa will sit within a radio-quiet zone the size of Pennsylvania where even a cellphone is forbidden, to preserve the array’s views of the heavens. Yet that precaution won’t save the telescope, due to be completed in the late 2020s, from what may soon be overhead: tens of thousands of communications satellites beaming down radio signals straight from the heavens. “The sky will be full of these things,” says SKA Director General Phil Diamond.

The rocket company SpaceX has already launched hundreds of Starlink satellites, the first “megaconstellation” intended to provide internet service to remote areas. The satellites have aroused the ire of optical astronomers because of the bright streaks they leave across telescopes’ fields of view. Now, radio astronomers are worried, too. This week, SKA released an analysis of the impact that Starlink and other constellations would have on the array. It finds they would interfere with one of the radio channels SKA plans to use, hampering searches for organic molecules in space as well as water molecules used as a key marker in cosmology.

SpaceX is promising to address the concern. But radio astronomers are also seeking regulations.

Brazil’s largest research facility focused on artificial intelligence (AI) has been launched yesterday (14) through a collaboration between the private and public sector.

Announced in February 2019, the Artificial Intelligence Center (C4AI) is the result of investments made by IBM, the São Paulo Research Foundation (FAPESP) and the University of São Paulo (USP).

C4AI will aim to tackle five major challenges related to health, the environment, the food production chain, the future of work and the development of Natural Language Processing technologies in Portuguese, as well as projects relating to human wellbeing improvement as well as initiatives focused on diversity and inclusion.

Almost 40 million people around the world have contracted Covid-19, and more than 1 million have died from the virus. The devastation has rippled even further, thanks to a global recession and rising political unrest. And as all of this unfolds, new research indicates that the governments around the world have exploited the pandemic to expand their domestic surveillance capabilities and curtail internet freedom and speech.

The human and digital rights watchdog Freedom House today published its annual Freedom on the Net report, which tracks the ebb and flow of censorship laws, net neutrality protections, internet shutdowns, and more around the world. This year’s report, which covers the period from June 2019 through May 2020, encompasses not only the Covid-19 pandemic but also the trade war between the US and China, which has resulted in a dramatic acceleration of the cyber sovereignty movement. Combined with numerous other geopolitical clashes that have impacted digital rights, global internet freedom has been broadly curtailed in 2020.

How the world’s greatest public health organization was brought to its knees by a virus, the president and the capitulation of its own leaders, causing damage that could last much longer than the coronavirus.

Online October 19-23. [registration required]

Online October 28, starting at 1 p.m. Eastern time. “Age of AI and Big Data, an initiative designed to develop curricula to foster character and ethical values in future leaders, preparing them to respond appropriately to the challenges posed by rapidly evolving technologies, such as artificial intelligence and Big Data management.” [registration required]

“In this blog post we discuss a mental model for RL, based on the idea that RL can be viewed as doing supervised learning on the “good data”. What makes RL challenging is that, unless you’re doing imitation learning, actually acquiring that “good data” is quite challenging. Therefore, RL might be viewed as a joint optimization problem over both the policy and the data. Seen from this supervised learning perspective, many RL algorithms can be viewed as alternating between finding good data and doing supervised learning on that data. It turns out that finding “good data” is much easier in the multi-task setting, or settings that can be converted to a different problem for which obtaining “good data” is easy. In fact, we will discuss how techniques such as hindsight relabeling and inverse RL can be viewed as optimizing data.”

“Lightning makes coding complex networks simple. Spend more time on research, less on engineering. It is fully flexible to fit any use case and built on pure PyTorch so there is no need to learn a new language.”

Institutions throughout the world have had to adapt to the Covid-19 pandemic. Many scanning centres shut their doors during lockdown, and have had to reopen gradually, and carefully. We surveyed several labs from around the world – to find out the challenges they’ve experienced and, in a few cases, the opportunities afforded.