Research Stack

What is ResearchStack?

ResearchStack is an SDK and UX framework for building research study apps on Android. It is designed from the ground up to meet the requirements of most scientific research, including capturing participant consent, extensible input tasks, and the security and privacy needs necessary for IRB approval.

Is it similar to Apple's ResearchKit(TM)?

Yes, an overarching goal of ResearchStack is to help developers and researchers with existing apps on iOS more easily adapt those apps for Android. Though the correspondence of features between the two SDKs isn’t one-to-one, the two SDKs will offer enough shared functionality and a common framework and naming scheme to greatly speed up adaptation of ResearchKit apps to Android (and ResearchStack apps to iOS) and the procedural aspects of running a study on a new platform (such as IRB approval and secure connectivity with a data collection backend).

What features are available?

Researchers and developers can leverage ResearchStack to build a basic Android app without having extensive knowledge of the Android platform. Using JSON and HTML files, a simple app can be built that will include the onboarding flow (Study Overview/Consent/Sign Up) and an Activities screen with scheduled tasks for the user. Graphing libraries facilitate data visualization. Researchers will also be able to reuse existing ResearchKit AppCore resources, with some minor changes. ResearchStack also has tools to allow easy on-disk encryption and strives to be extensible so that researchers can implement or modify screens they need.

Digital Marshmallow Test

Powered by Apple’s ResearchKit and Android’s ResearchStack application development frameworks, the DMT uses the smartphone to understand individuals who have trouble controlling their impulses, particularly with regard to unhealthy behaviors, such as overeating or drinking alcohol, in order to help them regain control when they need it most. The app consists of surveys, game-like tasks and passive sensor data such as accelerometer, aiming to understand how individuals regulate their behavior in different settings and at different times of the day. The ultimate goal is that the DMT will eventually lead to non-invasive mobile interventions to identify and help those at greatest risk for impulsive responding and enhance temptation resilience before serious problems can occur. This is collaborative work with Northwell Health.

Beehive

The study of human behavior currently faces three major challenges: recruiting participants, conducting frequent and longitudinal data collection, and utilizing smart and adaptive interventions. We developed a novel system, Beehive, to tackle the intersection of these challenges. Beehive is a research platform developed for the study of human behavior by leveraging the power of mobile ubiquitous technology. Using this platform, researchers will be able to recruit more participants, collect objective data and apply smart interventions to improve daily life, in the wild. Beehive seamlessly integrates with smartphones making it easy for participants to sign up for and participate in a study by using the mobile phone that is already in their pockets. To learn more about ongoing experiments with Beehive, contact the researchers involved.

Immersive Recommendation

Our daily lives are mediated by a myriad of digital services. As a result we continuously generating digital traces that can be used to chronicle, characterize, and influence our behaviors and preferences. Digital services from search to e-commerce and entertainment leverage these data to target advertising, personalize recommendations and adapt services.

In our research we explore a personalization approach in which individuals actively harness their own digital traces to reframe personalization. We focus in particular on algorithms, systems, and rich user interactions that make personalization immersive, through the use of diverse personal data sources, and intentional, through transparency and user-modeling input that allows the end user to amplify their aspirational-behaviors relative to historically-observed behavior.

Ancile

People today generate a tremendous amount of data as a byproduct of their daily activities through the use of mobile phones, wearable devices, and online services. The availability of this “microscale” data creates exciting opportunities for solving complex planning problems (e.g.,using location data to decide where to place a new café on campus to best serve the Cornell community) but it also raises serious concerns about security and privacy.

The Ancile Project is developing a new software platform for managing microscale data in a privacy-sensitive manner. This platform will enable researchers to investigate how microscale data can be used to solve planning problems and improve resource allocation. Results from this research could provide deeper insights about the aggregate behavior of the Cornell community and could help improve use of facilities, dining, and transportation services. The first set of Ancile-based research studies will be conducted on the new Cornell Tech campus in Roosevelt Island. To learn more, contact the researchers involved.

Retrospective Data Learning

Retrospective data learning mines an person’s digital traces generated through online interactions for clues to their past. Service providers such as Google and Amazon implicitly record users’ location, search queries, email, or calendar. By analyzing these retrospective data, we aim to learn about temporal patterns and deviations that reveal individual behavior changes across many contexts.

One example application is to create new digital biomarkers for health and health behavior from patients’ retrospective data. We are working to demonstrate the use of personal historical digital traces as a source of signal for identifying severity and changes in state of a disease, and response to treatment, with ultimate use in treatment management.

Presently, we have built a tool for research study participants to access their individual search and location history from Google Takeout and filter it for inclusion in IRB approved research; with the utmost care taken to provide security and privacy for the patient. We work with clinical collaborators to recruit patients diagnosed with conditions of interest, collect data via takeout, then analyze search patterns and location trajectories. We are exploring how these data reveal personal symptom development and variability in daily mobility by comparing patients’ historical data with their clinical history to develop a detection algorithm for various aspects of disease state and behavior.

Pushcart is an effortless and compatible passive data collection tool that provides objective, high frequency, household-level data on food consumption and nutrition. Pushcart can serve as a proxy for variety of nutritional metrics on the household level including: total calories consumed and division to different food groups (e.g. fruits, vegetables, sugar) continuously on a weekly basis. Pushcart provides an objective measure of nutritional habits by analyzing online shopping email receipts from grocery stores (e.g. Amazon) and prepared-food delivery services (e.g. Seamless).

Pushcart is plug-and-play compatible with any service that generates email receipts. There is no single technology platform or grocery service to which Pushcart is tied and the service provider does not have to participate. The consumer unilaterally signs up for Pushcart by permitting all emailed receipts to be automatically sent to the platform for processing.

LIMBR is a suite of applications that we have developed based on the innovations developed under this funding for effective pain reporting, and self-directed management of medication and exercise tracking LIMBR is comprised of tools to: (i) facilitate self reporting of pain, medication, and affect, (ii) deliver individually tailored exercise videos, (iii) provide a platform for one-on-one chatting with a health coach, and (iv) collect data on the participants’ patterns of movement.

We tested LIMBR in a prospective pilot study, in 100 patients with acute or chronic low axial discogenic back pain at the Hospital for Special Surgery. The prevalence and disability associated with discogenic low back pain are significant and there is a general lack of widely accepted conservative treatments for the condition. A majority of previous studies that have assessed the efficacy of self-directed exercise based rehabilitation programs did not include comprehensive metrics to confirm whether participants actually participated in the exercises prescribed. In this pilot study, we aimed to determine whether proxies of compliance with a daily, self directed rehabilitation program for discogenic low back pain are correlated with participant reported improvements in pain and function.

The specific aim is to explore how mHealth applications and data tracking can promote adherence to self-directed rehab and engagement in low back pain self-management and improve pain and function. The feasibility of an AVA based mobile interface designed to help patients with chronic low back pain track ADLs, medication, affect and pain intensity is also evaluated.