Data Science newsletter – October 2, 2019

GROUP CURATION: N/A

What does it take for a human to trust a robot? That is what Army researchers are uncovering in a new study into how humans and robots work together.

Research into human-agent teaming, or HAT, has examined how the transparency of agents–such as robots, unmanned vehicles or software agents–influences human trust, task performance, workload and perceptions of the agent. Agent transparency refers to its ability to convey to humans its intent, reasoning process and future plans.

New Army-led research finds that human confidence in robots decreases after the robot makes a mistake, even when it is transparent with its reasoning process. The paper, “Agent Transparency and Reliability in Human-Robot Interaction: The Influence on User Confidence and Perceived Reliability,” has been published in the August issue of IEEE-Transactions on Human-Machine Systems.

Foreign governments and corporations could profit from American academic institutions’ failure to safeguard taxpayer-funded biomedical research, according to a set of new reports from a federal watchdog.

The reports, issued by the Department of Health and Human Services inspector general, show that 54% of research institutions funded by the National Institutes of Health — some 1,013 universities and academic centers — failed to publish financial conflict-of-interest policies online. In 2018, according to the report, the NIH conducted just three audits to determine grantee institutions’ own efforts to safeguard their research — down from 28 in 2012.

“The concern, generally speaking, is whether financial interests threaten or distort the use of NIH funds for their intended research purpose or the results of that scientific research,” Erin Bliss, an assistant HHS inspector general.

On Wednesday, Amazon unveiled a slew of new hardware. The gear is, on the face of it, meant to help you listen to music, render raw meat edible, or hold prescription pieces of glass in front of your eyeballs. Most of the stuff also involves getting the Echo, Amazon’s smart speaker, into every nook and cranny of not only your home, but your body. “When stripped of the day’s high gloss,” Sidney Fussel at the Atlantic observed, much of Amazon’s newly announced goods “are just wearable microphones.”

The existing microphones that Amazon has deployed into the world have already resulted in a small slew of privacy scandals, as Fussel notes, from police departments that request footage from owners of Amazon Rings to humans who review audio clips from Alexa speakers. True, Amazon will now give users the option to automatically delete any stored recordings every few months—but as Shira Ovide points out at Bloomberg, Amazon could have chosen to just stop recording voice commands in the first place.

As far as the archaeological record goes, board gaming began with senet, nearly 1,800 years before Tut was even born. About 700 years later, in ancient Sumer, the Royal Game of Ur was important enough that its rules were written down—actually incised on a cuneiform tablet. (Otherwise, rules were almost never set in stone.) The names of many ancient games have slipped through the cracks of history, so researchers identify them by what remains. There’s “33 Circles” from Egypt, “10-Ring” from Bronze Age Crete, and the Middle Eastern “58 Holes.” Since the rules have been lost over time, the way many ancient games were played is based on speculation. With so many games, and so many variations, Browne’s work isn’t merely an archaeological-meets-ludological project. It’s genealogical, too.

Computer scientists say AI research is becoming increasingly expensive, requiring complex calculations done by giant data centers, leaving fewer people with easy access to the computing firepower necessary to develop the technology behind futuristic products like self-driving cars or digital assistants that can see, talk and reason.

The danger, they say, is that pioneering artificial intelligence research will be a field of haves and have-nots. And the haves will be mainly a few big tech companies like Google, Microsoft, Amazon and Facebook, which each spend billions a year building out their data centers.

In the have-not camp, they warn, will be university labs, which have traditionally been a wellspring of innovations that eventually power new products and services.

“The huge computing resources these companies have pose a threat — the universities cannot compete,” said Craig Knoblock, executive director of the Information Sciences Institute, a research lab at the University of Southern California.

Autonomous weapons technologies, which rely on artificial intelligence, are advancing rapidly and without sufficient public debate or accountability. Oversight of increased autonomy in warfare is critically important because this deadly technology is likely to proliferate rapidly, enhance terrorist tactics, empower authoritarian rulers, undermine democratic peace, and is vulnerable to bias, hacking, and malfunction. The top competitors in this arms race are the United States, China, Russia, South Korea, and the European Union. [link to full text]

Next door to the parsonage of the small Christian church (Disciples of Christ) congregation, of which I am a member, live a militantly progressive couple who are estranged from their conservative religious upbringing. For years they exchanged pleasantries with the pastor, before stumbling into a political discussion in which they discovered he was not, to their surprise, a right-winger. “Oh, I get it: You’re not those Christians,” the husband exclaimed. The couple soon became regulars at our church.

I mention this anecdote in connection with new research showing that the political views of conservative Christians — notably the militant Christian right composed mostly of white Evangelicals though with some Catholic “traditionalists” in harness with them — are pushing people who strongly disagree with them away from Christianity (or any other religious faith). Amelia Thomson-Deveaux and Daniel Cox explain:

Researchers haven’t found a comprehensive explanation for why the number of religiously unaffiliated Americans has increased over the past few years — the shift is too large and too complex. But a recent swell of social science research suggests that even if politics wasn’t the sole culprit, it was an important contributor. “Politics can drive whether you identify with a faith, how strongly you identify with that faith, and how religious you are,” said Michele Margolis, a political science professor at the University of Pennsylvania … “And some people on the left are falling away from religion because they see it as so wrapped up with Republican politics.”

The National Science Foundation, the largest government funder of basic social and behavioral research in the United States, is changing how it “positions” some of its research programs in those fields. Arthur “Skip” Lupia, who heads the Social, Behavioral and Economic Directorate at NSF, officially announced the changes in the Behavioral and Cognitive Sciences and the Social and Economic Sciences divisions in a ‘Dear Colleague’ letter released earlier this week. The changes to those two of the directorate’s four divisions are to take effect on October 1.

Lupia talked about emphasizing the public value of social science during a webinar by the Consortium of Social Science Associations two weeks ago. “One way we do that is by making sure it’s conveyed more effectively, make sure that the users understand more about it. The other is to position our programs – our touchpoints with the research community and the public – so people can say, ‘Yeah, that’s the thing I really need.’”

The just-released Intergovernmental Panel on Climate Change (IPCC) special report on the ocean and cryosphere in a changing climate (SROCC) (1) details the immense pressure that climate change is exerting on ocean ecosystems and portrays a disastrous future for most life in the ocean and for the billions of people who depend on it unless anthropogenic greenhouse gas (GHG) emissions are slashed. It reinforces in stark terms the urgency of reducing carbon emissions expressed in a 2018 IPCC report (2). But another just-released report (3) provides hope and a path forward, concluding that the ocean is not simply a victim of climate change, but a powerful source of solutions. Drawing on this report organized by the High Level Panel (HLP) for a Sustainable Ocean Economy, which quantifies and evaluates the potential for ocean-based actions to reduce emissions, we outline a “no-regrets to-do list” of ocean-based climate actions that could be set in motion today. We highlight the report’s analysis of the mitigation potential and the required research, technology, and policy developments for five ocean-based mitigation areas of action: renewable energy; shipping and transport; protection and restoration of coastal and marine ecosystems; fisheries, aquaculture, and shifting diets; and carbon storage in the seabed (see the figure). Make no mistake: These actions are ambitious, but we argue that they are necessary, could pay major dividends toward closing the emissions gap in coming decades, and achieve other co-benefits along the way (3, 4). [full text]

Amazon’s announcements of random devices and the creation of new infrastructures such as the Sidewalk wireless protocol or better neural networks for voice make sense. And the service-centric viewpoint that Amazon has had since day one has helped it by making the infrastructure for a consumer-oriented internet of things ubiquitous.

I saw a lot of people trying to compare the launch of Amazon’s devices and its lack of discipline around those devices to Apple, but the focus on hardware misses the point. Amazon is succeeding because it is willing to throw less-than-perfect hardware over the wall to see what sticks. Remember the original Echo Show? It was hideous, but the next-generation Show devices improved on the design and now the Show products are a huge seller for Amazon.

The Illinois Board of Higher Education this month approved a new undergraduate degree combining computer sciences and animal sciences at the University of Illinois. Incoming freshmen can enroll in the program, known as CS + Animal Sciences, starting in the fall of 2021.

The new degree, first of its kind in the country, expands the list of blended CS + X degrees pioneered through the Department of Computer Science at Illinois and featured in U.S. News & World Report’s upcoming “Best Colleges of 2020” guidebook. CS + Animal Sciences follows the CS + Crop Sciences program as the second such degree to be offered in the College of Agricultural, Consumer and Environmental Sciences.

Lilly Endowment puts more than $840,000 into studies at Purdue and IU into how to teach ethical professional practices in an era of big data and artificial intelligence

September 30th will be my last day in the role of Library Carpentry Community & Development Director. I have been fortunate to meet so many amazing people working in libraries and the research community during this time. Thank you to the IMLS, the California Digital Library, and The Carpentries for this great opportunity. So many members of the community have helped Library Carpentry grow these past couple of years, not to mention the initial hard work that went into starting Library Carpentry.

Confusion around the value of influencer campaigns allows influencer agencies, which manage influencer relationships and sometimes provide creative services, to charge nontransparent markups.

“The lack of transparency is the biggest problem,” said influencer consultant Yuliya Gorenko, who formerly led influencer marketing at L’Oréal in the Ukraine.

But influencer marketing, on track to hit $15 billion in ad spend by 2020 according to Business Insider Intelligence, is quickly becoming a staple of the media plan for many brands.

Despite their concerns, marketers work with influencers and their agencies because they want to reach young, passionate audiences. But many are starting to question the value of their investment and, like in programmatic, will require more clarity before devoting even more media budget to influencers.

Washington, DC November 8, starting at 8 a.m., organized by Duke Center on Law & Technology. “The Data Governance Design Conference (DGDC) will convene policymakers, industry, academia, and legal practitioners to explore models, needs, and enabling environment for data governance.” [save the date]

New York, NY October 17, starting at 9 a.m., New York-Presbyterian Hospital (466 Lexington Ave.). “Learn spatial data science with the author of the sf and sp packages himself, Edzer Pebesma.” [$$$]

Bethesda, MD October 24-25. ” The objective of this meeting is to identify the perspectives, challenges, and opportunities of integrating machine learning with multiscale modeling (ML-MSM) in biomedical, biological, and behavioral systems.” [registration required]

“Employers are welcome to link directly to this page in their job listings. The AAS Employment Committee recommends the following application guidelines for institutions seeking to hire postdoctoral fellows.”

The National Center for Biomedical Information (NCBI) at the National Library of Medicine (NLM) recently moved the five petabytes of public SRA data to the cloud with support from the National Institutes of Health (NIH) Science and Technology Research Infrastructure for Discovery, Experimentation, and Sustainability (STRIDES) Initiative. These data include a variety of genomes, gene expression data, and more. Plans are underway to move the other half of the SRA data, which is controlled-access human genomic data.

Machine learning combined with solution-specific software can dramatically improve the speed, accuracy, and effectiveness of back-office operations and help organizations reimagine how back-office work gets done.

A current trend among mid and large organizations is to implement Robotic Process Automation (RPA) in the back office to minimize manual tasks and achieve efficiencies. While there are specific use cases that make RPA an appropriate technology, there are significant differences with a machine learning approach.

“Google’s new ‘ALBERT’ language model has achieved state-of-the-art results on three popular benchmark tests for natural language understanding (NLU): GLUE, RACE, and SQuAD 2.0. ALBERT is a “lite” version of Google’s 2018 NLU pretraining method BERT. Researchers introduced two parameter-reduction techniques in ALBERT to lower memory consumption and increase training speed.”