… The idea of using phones for health behavior came to pediatric hematologist Nirmish Shah MD one day while he attempted to talk to a 16-year-old sickle cell disease patient as she snapped selfies of herself with the doctor. Her mom and toddler sister nearby both had their noses to screens as well. “I need to change how I do this,” Shah thought to himself.
Pediatric hematologist Nirmish Shah MD
Pediatric hematologist Nirmish Shah MD is director of Duke’s sickle cell transition program.
Twenty health apps later, he’s running phase II clinical trials of phone-based interventions for young sickle cell patients that encourage them to stay on their medication schedule and ask them often about their pain levels.
One tactic that seems to work pretty well is to ask his patients to send in selfie videos as they take their meds each day. The catch? The female patients send a minute or so of chatty footage a day. The teenage boys average 13 seconds, and they’re grumpy about it.
Clearly, different activities may be needed for different patient populations, Shah said.
Mobile dating apps that allow users to filter their searches by race – or rely on algorithms that pair up people of the same race – reinforce racial divisions and biases, according to a new paper by Cornell researchers.
As more and more relationships begin online, dating and hookup apps should discourage discrimination by offering users categories other than race and ethnicity to describe themselves, posting inclusive community messages, and writing algorithms that don’t discriminate, the authors said.
“Serendipity is lost when people are able to filter other people out,” said Jevan Hutson ‘16, M.P.S. ’17, lead author of “Debiasing Desire: Addressing Bias and Discrimination on Intimate Platforms,” co-written with Jessie G. Taft ’12, M.P.S. ’18, a research coordinator at Cornell Tech, and Solon Barocas and Karen Levy, assistant professors of information science. “Dating platforms have the opportunity to disrupt particular social structures, but you lose those benefits when you have design features that allow you to remove people who are different than you.”
The paper, which the authors will present at the ACM Conference on Computer-Supported Cooperative Work and Social Computing on Nov. 6, cites existing research on discrimination in dating apps to show how simple design decisions could decrease bias against people of all marginalized groups, including disabled or transgender people.
People dealing with chronic conditions are keener to share personally-generated data than people that don’t have a chronic disease, Deloitte’s 2018 Survey of U.S> Health Care Consumers learned. This and other insights about the patient journey are published in Inside the patient journey, a report from Deloitte that assesses three key touch points for consumer health engagement.
Understanding important high-resolution features of cells in tissues remains a challenge. Recently developed technologies, including many supported through NIH programs, allow researchers to explore the organization of large numbers of cells at the individual cell level. These advances opened the possibility to study and map the organization of all cells within tissues or organs across the human body. While HuBMAP is not anticipated to map the entire body, it will get the work started, providing a framework for more complete mapping and making data available to the research community for further study.
Through the research awards, HuBMAP investigators will:
Generate, standardize, and validate extensive data sets on cell organization and variability using existing technologies;
Develop new tools and techniques to construct high-resolution tissue maps; and
Coordinate program activities, manage HuBMAP data, and build an atlas of tissue maps.
The company’s tech gives pathologists insights to improve the accuracy of diagnostics. Some 300 clinical and research facilities are now using the software. Under development are specific AI-enabled, disease-specific modules that help automate insights, the first of which is expected to hit the market by December.
With the United States engaged in a “long-term strategic competition” with China and Russia, which are mounting persistent cyber attack campaigns that pose long-term risks to America, the U.S. military will act to deter aggression, cyber or otherwise, according to a new policy, known as the Department of Defense Cyber Strategy, from the U.S. Department of Defense.
“China is eroding U.S. military overmatch and the nation’s economic vitality by persistently exfiltrating sensitive information from U.S. public and private sector institutions,” a summary of the DOD strategy stated. “Russia has used cyber-enabled information operations to influence our population and challenge our democratic processes. Other actors, such as North Korea and Iran, have similarly employed malicious cyber activities to harm U.S. citizens and threaten U.S. interests. The United States’ growing dependence on the cyberspace domain for nearly every essential civilian and military function makes this an urgent and unacceptable risk to the nation.”
As such, DOD “must take action in cyberspace during day-to-day competition,” to defend U.S. interests and maintain the U.S. military advantage, the summary stated. This action includes conducting intelligence operations in cyberspace as well as preparing the military’s cyber capabilities.
Microsoft Ignite, Microsoft’s annual developer conference, wrapped up last week and many of the big announcements focused on artificial intelligence and machine learning. The keynote presentation from Microsoft’s Cloud AI lead Eric Boyd showcases the major developments, or you can check out his accompanying blog post for a written summary.
In this post, I’ll dive down into the details of the specific announcements and provide links to the associated recorded presentations from Ignite and other technical documents.
Two weeks ago, AncestryDNA and the popular music-streaming service Spotify unveiled a service that promised to use your very own genetic material to build a personalized playlist. Such a playlist, their website stated, would be totally unique. The sonic essence of your very own cultural inheritance. “If you could listen to your DNA, what would it sound like?” they asked.
Online critics immediately roasted the companies for both the puffed-up gesture toward bleeding-edge genealogical experience — strap yourself into yourself and enjoy the ride! — and the dubious promise of tech-enabled self-discovery. Like so many companies that collect biometric data to personalize their services, Spotify and AncestryDNA’s playlist generator felt both intrusive and gimmicky — the kind of low-level troubling we’ve come to associate with an increasingly invasive consumer culture.
Artificial intelligence is guiding a growing number of decisions in criminal justice, education, health care and other areas, with the potential to significantly alter people’s lives.
To examine the real-world consequences of this trend, a group of researchers at Cornell’s Ithaca and New York City campuses have formed the Artificial Intelligence, Policy, and Practice initiative. Over the next three years, members hope to bring together a community of scholars with expertise in computing, the law, social science, communications and philosophy to create opportunities to collaborate on related research.
“AI is being used for particularly high-stakes decisions that affect people’s opportunities to succeed in life, and often has disproportionate impact on people from marginalized communities,” said Karen Levy, assistant professor of information science, one of the four researchers who formed the group. The others are Solon Barocas, assistant professor of information science; Jon Kleinberg, the Tisch University Professor of Computer Science; and Helen Nissenbaum, professor of information science at Cornell Tech.
An EPFL scientist has developed a novel approach that boosts the chances of finding extraterrestrial intelligence in our galaxy. His method uses probability theory to calculate the possibility of detecting an extraterrestrial signal (if there is one) at a given distance from Earth.
With so many violations — and so few repercussions — senior executives from Google, Apple, Amazon, and Twitter, among other firms, were summoned before the Senate Committee on Commerce, Science, and Transportation last week to explain why data privacy breaches continue, and to discuss some remedies. Sen. John Thune (R., S.D.), the committee chairman, said it’s no longer a question of whether there needs to be a federal law to protect consumer data privacy but “what shape the law should take.” Another hearing is planned later this month.
Urs Gasser, LL.M. ’03, is executive director of the Berkman Klein Center for Internet & Society at Harvard University and a professor of practice at Harvard Law School. His research and teaching focus on information law and policy, and he writes frequently about privacy, data protection, and the regulation of digital technology. Gasser discussed the state of data privacy with the Gazette via email and suggested what might be done to protect users from companies that profit from people’s data.
Online “To get a shirt, you must make five pull requests (PRs) between October 1–31 in any timezone. PRs can be to any public repo on GitHub, not just the ones highlighted. The PR must contain commits you made yourself. PRs reported by maintainers as spam or that are automated will be marked as invalid and won’t count towards the shirt. This year, the first 50,000 of you can earn a T-shirt (compared with 30,000 in 2017).”
New York, NY November 16-18. “Join MIT Hacking Medicine and the Mt. Sinai Dept. of Rehabilitation and Human Performance in New York City for one of the largest healthcare hackathons in the world!” Deadline to apply is October 31.
The “Challenge seeks to evaluate the power of statistical and machine learning models as a systematic and cost-effective means for catalyzing compound-target interaction mapping efforts by prioritizing most potent interactions for further experimental evaluation.” Registration required.
“ACM, the Association for Computing Machinery, today announced the inaugural ACM Charles P. “Chuck” Thacker Breakthrough in Computing Award (“ACM Breakthrough Award”). Named in honor of the late Chuck Thacker, an engineer and 2009 ACM A.M. Turing Award recipient, the new award is designed to recognize those whose work embodies the same revolutionary impact that exemplified Thacker’s career. A cash prize in the amount of $100,000 will accompany the award, with financial support provided by Microsoft. Nominations can now be submitted via an online submission form and will be accepted through January 15, 2019.”
To gain insight into users’ expectations and mental models around the many types of individualized recommendations offered on sites, we ran a remote moderated usability study with 8 participants located across the United States. In each session, participants completed facilitator-assigned tasks on 2–3 websites on which they had accounts and also answered recommendation-related questions in an interview.
Our study participants were highly attuned to the fact that sites commonly track their browsing patterns, purchase histories, and other sources of data to present individually personalized suggestions. Overall, these recommendations were appreciated and seen as instrumental for narrowing down the options available on a site. To reap this benefit, users were willing to sacrifice some privacy; they expected many of their actions to be tracked and analyzed.
PyTorch 1.0 accelerates the workflow involved in taking breakthrough research in artificial intelligence to production deployment. With deeper cloud service support from Amazon, Google, and Microsoft, and tighter integration with technology providers ARM, Intel, IBM, NVIDIA, and Qualcomm, developers can more easily take advantage of PyTorch’s ecosystem of compatible software, hardware, and developer tools. The more software and hardware that is compatible with PyTorch 1.0, the easier it will be for AI developers to quickly build, train, and deploy state-of-the-art deep learning models.
Fuzz testing refers to a technique used to find and expose vulnerabilities in a program via generation of unexpected inputs. Black box fuzzers operate quickly, and are best for exposing shallow bugs. White box fuzzers, in contrast, are more comprehensive in their approach, traversing the “entire source code tree.”
Consequently, as programs grow large, the quality of this technique worsens. Grey-box fuzzing provides a middle ground in the world of fuzz testing. David Rosenberg, Data Scientist in the CTO office at Bloomberg LP and Adjunct Associate Professor at CDS, with Siddharth Karamcheti, CTO Research Intern at Bloomberg, and Gideon Mann, Head of Data Science in the CTO office at Bloomberg, presented improvements on grey-box fuzz testing via Thompson sampling and bandit-based optimization in their recent work.
Computational researchers can reproduce experiments with a single click on a Web application called ReproServer. This new platform was developed by Remi Rampin (Moore-Sloan Research Engineer), Fernando Chirigati (Postdoctoral Research Associate at NYU Tandon School of Engineering), Vicky Steeves (Research Data Management and Reproducibility Librarian), and Juliana Freire (Professor of Computer Science, Engineering, and Data Science).
ReproServer builds on these researchers’ past work, ReproZip. ReproZip “allows users to automatically and transparently capture all the dependencies of a computational experiment in a single, distributable bundle, that can be used to reproduce the entire experiment in another environment.” It works automatically with existing applications, meaning researchers don’t need to concern themselves with reproducibility as they design experiments. But, before ReproServer, ReproZip users still had to download and install software to reproduce an experiment.
The new addition of ReproServer to ReproZip now means users who want to reproduce experiments can do so with a single click.