"small, n = me, data" - deborah estrin
TRANSCRIPT
Small, n=me, data
Deborah EstrinProfessor, Computer Science, Cornell NYC Tech
Professor, Public Health, Weill Cornell Medical CollegeCo-founder, Open mHealth
[email protected] done with collaborators from
Cornell, UCLA, openmhealth.org, ...
1
1Saturday, August 17, 13
Agenda
• Shared, digital, life
• Prelude: mobile health as n=me data
• Small data: beyond mobile and beyond health
• Proposed market/system architecture with individual as nexus of control
• Proposed shared testbed
2Saturday, August 17, 13
2
n=medata
third pillar of personalized, precision, medicine“Big data”
(EHRs, Web mining)
“n=me data” (mHealth,
digital traces)+ +“omics”
3
3Saturday, August 17, 13
mobile apps generate data
4Saturday, August 17, 13
5
Passively-recorded activity and location traces
UI: E. Wang
5Saturday, August 17, 13
6
h"p://ginger.io/the-‐pla1orm/
Communication and activity data: Ginger.io check engine light
6Saturday, August 17, 13
7
• mobile carriers- location/activity - call records
• cable box/home gw- TV patterns (sleep,hearing)- internet mediated patterns- household focus
• utilities (elec,water)- diurnal rhythms- appliance use- household focus
• smart cars- location/activity
• search- state of mind- topic/concern
• social media and email- social patterns- interaction- mood
• e-commerce, payments- consumption/input- patterns
• games/music/videos- cognitive state- indicator/influencer
Beyond mobilesmall data: digital traces from diverse consumer services
7Saturday, August 17, 13
8
measure, manage, incentivize, improve: wellbeing, consumption, personal/family logistics behavior change, community resiliency
• aging independent living seniors
• newly independent living young adults
• newcomers to a neighborhood/city
• personal profiles in social media, games
• n=me health and wellness outcomes
• ...
Beyond ‘health’...life:
8Saturday, August 17, 13
Proposed small data socio-technical architecture
individual as nexus for fusion of their data streamsapps run over data in personal data vault
• Subscriber access to their individual data traces--data liberation!
•programmatic, realtime, opt-in through personal data APIs
• Raw data shared with subscriber only
•avoids a range of privacy and regulatory concerns
• Fuel new market of third-party personal informatics apps/services
• some apps will run in PDV; others externally
9Saturday, August 17, 13
10
Individual as nexus: discussion
• Each data source has shared/other origins
• Individual has control over their corpus of data streams to correlate, fuse
• App/service utility derives from lack of anonymity
• Selective sharing embodied in apps
10Saturday, August 17, 13
small data: key challenges
• Getting the data– Personal data APIs (data liberation to the consumer)– Convincing/incentivizing service providers
• Data processing, inference, fusion, modeling– diverse, noisy, lossy data– signal processing, machine learning, natural language...
• Data and API standards– app model and economy
• Personal data vaults– Security models and mechanisms, usability– Policy questions re. ownership, access, rights
• Testbed for prototypes and pilots– economy of scale in a shared testbed for rapid iterative exploration– secure and private data handling, IRB, methods, tools 11
11Saturday, August 17, 13
mpireproposed Testbed for Small Data and personal informatics
participant recruitment, incentives, management--across large and diverse participant populations
experiment configuration, control, coordination, analysis, administration
IRB study management templates
aggregated datasets
Current collaborators: Intel Labs, Ericsson Research, IBM Research, ATT Research, Verizon, Time Warner Cable
Network and cloud APIs to personal-data
Deployment framework for end-user apps
Secure personal-data vaults
Data processing, fusion, inference, modeling12
12Saturday, August 17, 13
open architecture to promote modularity, data interoperability and software reusability
Can we reuse open mhealth architecture/APIs/Registry?
activity classification
graphing significant changes in mobility
data storage (e.g., PHR)
proprietary component with omh API (RunkeeperEntra Glucometer)
mood App (PAM, moodmap...)
13Saturday, August 17, 13
http://smalldata.tech.cornell.edu/
1414Saturday, August 17, 13