All posts by admin

Analysis of 791 million actions on social media leads to increase in user’s online and offline activities

Mobilize Center graduate student Tim Althoff, Computer Science graduate student Pranav Jindal, and Mobilize Center faculty Jure Leskovec published a research paper in Proceedings of the Tenth ACM International Conference on Web Search and Data Mining (WSDM 2017), where they investigated the effect of social networks on users’ online and offline behaviors, including user engagement, retention, and physical activity levels.

The results are based on an analysis of data from 6 million people from over 100 countries collected over a 5-year period using the physical activity tracking application Argus. Althoff and colleagues found that participating in the social network increases both in-application activity as well as physical activity levels.  They were also able to take advantage of a natural experiment within the dataset to distinguish between correlation and causation.

Read full article

Participate in challenge to teach a musculoskeletal model to walk using reinforcement learning

Mobilize Center Postdoctoral Fellow Lukasz Kidzinski has organized a challenge to make a musculoskeletal model walk as far as possible in a virtual environment in 5 seconds using a reinforcement learning approach. The winner will be invited to the 2nd Applied Machine Learning Days at EPFL in Switzerland on January 2018, with travel and accommodations paid for (see program for 2017 event here).

The challenge focuses on programming a model of the brain’s motor control unit, which controls movement, within the OpenSim simulation environment. Understanding how the brain functions in normal conditions and for neurological disorders, such as cerebral palsy, multiple sclerosis or stroke, is key to improving treatments for these disorders.

Read more about the evaluation, rules and resources.

Wearable biosensors can measure physiological changes and reveal useful health information

Mobilize Center postdoc Jessilyn Dunn and colleagues published a research paper in PLOS Biology where they recorded over 250,000 daily physiological measurements from 43 participants using multiple wearable devices to investigate the role of these devices to diagnose and analyze disease. Combining the information from these devices and from medical measurements, such as heart rate and oxygen levels, researchers found these devices could be used to identify abnormal physiological signals, in this case arising from the onset of Lyme disease. The signals also showed differences between being insulin-sensitive versus insulin-resistant, raising the possibility of using these devices to help detect the risk for type 2 diabetes.

Read a summary of this work in the Stanford Medicine News

Read full article

OpenSim simulations yield insights into the design of assistive devices to reduce the metabolic cost of running

Mobilize Center researcher Thomas Uchida and OpenSim researchers published a paper in PLOS One where they generated muscle-driven simulations of movement to augment experimental data and provide insights into the design of assistive devices to reduce energy consumption during running.

Using the open-source OpenSim software platform, they simulated 10 human subjects running at 2 and 5 m/s to examine the predicted changes in muscle recruitment patterns and metabolic power consumption with assistive devices.

Results from the simulations yielded observations that can be used to form hypotheses for future experimental studies.

Read full article

Analysis of 600,000+ crisis hotline text messages identifies successful counseling techniques

Mobilize Center graduate student Tim Althoff and Mobilize Center faculty member Jure Leskovec published a research paper in the Transactions of the Association for Computational Linguistics Journal where they conducted a large-scale quantitative analysis of over 600,000 text messages from 15,555 counseling conversations. They developed novel computational methods such as a sequence-based unsupervised conversation model and psycholinguistics-inspired word frequency analyses to learn how various linguistic aspects of conversation are correlated with conversational outcomes.

Results from the analysis yielded five strategies that successful counselors utilize. Read the full article

Women in Data Science (WiDS) conference registration now open

This one-day technical conference provides an opportunity to hear about the latest data science-related research in a number of domains, learn how leading-edge companies are leveraging data science for success, and connect with potential mentors, collaborators, and others in the field.

An outstanding group of speakers has been assembled, including:

  • Diane Greene, Senior Vice President at Google
  • Yael Garten, Director of Data Science at LinkedIn
  • Janet George, Fellow and Chief Data Officer at SanDisk
  • Susan Holmes, Professor of Statistics at Stanford University
  • Megan Price, Executive Director at the Human Rights Data Analysis Group

Learn more and register.

Announcing Workshop on Machine Learning for Health @ NIPS 2016

The Workshop on Machine Learning will take place at the Twenty-Ninth Annual Conference on Neural Information Processing Systems (NIPS 2016) in Barcelona, Spain, on Friday, December 9, 2016. The deadline to submit an extended abstract is October 28, 2016.

The purpose of the workshop is to bring together clinicians, health data experts and machine learning researchers to discuss how statistical learning can enhance both science and the practice of medicine. Mobilize Distinguished Postdoctoral Fellows Jason Fries and Madalina Fiterau are part of the organizing committee.

Learn more about the workshop and abstract submission process.

NIH Rehabilitation Research Conference Highlights Role of Technology and Big Data

The NIH is hosting a conference “Rehabilitation Research at NIH: Moving the Field Forward” on May 25-26, 2016 to discuss the range of rehabilitation research projects funded by the NIH and to gather feedback on the development of the recommendations for the NIH Rehabilitation Research Plan.  Panel discussions include a discussion of telemedicine and smartphone apps in rehab (“Bending the Arc of Technology Toward Rehabilitation and Health) and the use of big data in rehab (“Using Data to Drive Discovery”). Dr. Jennifer Hicks, the Mobilize Center Director of Data Science, will present a talk on how our Center uses data from wearable sensors, research labs, and clinics to understand and improve human mobility.

The full agenda is available here.  The event will also be webcast:

Watch Wednesday’s webcast

Watch Thursday’s webcast

Abby King Discusses the Key for Behavior Change at mHealth Connect

At last week’s mHealth Connect, Abby King highlighted the need to personalize incentives for behavior change. How to select the right approach for motivating an individual, though, should come from an analysis of the individual’s behavior, not from asking the individual about her preference.

Read more about this idea in a piece written by David Shaywitz, the moderator of the panel on which Abby King participated. A recording of the discussion will also be available. To sign up to receive notification of the recording’s availability, click here.   To learn more about mHealth Connect and the issues around wearables that are being tackled, click here.