220 likes | 371 Views
Lessons Learned from the NetSense Smartphone Study Aaron Striegel 1 , Shu Liu 1 , Lei Meng 1 , Christian Poellabauer 1 , David Hachen 2 , Omar Lizardo 2 1 Department of Computer Science and Engineering 2 Department of Sociology University of Notre Dame. Problem Space.
E N D
Lessons Learned from the NetSense Smartphone Study Aaron Striegel1, Shu Liu1, Lei Meng1, Christian Poellabauer1, David Hachen2, Omar Lizardo2 1Department of Computer Science and Engineering 2Department of Sociology University of Notre Dame
Problem Space Segmented Industry StandardsBound Wireless Data Tsunami How do we get good data at a reasonable scale? Indoor Coverage Small Cells New Devices ! 5x Grad Students + 1-3 Friends IoT QoE / Analytics
Why is it Hard? Invasive – Phone Agent Privacy Great Data – Social, Location Raw Cost - $60+ / month Expensive Data Scale - Granularity
Talk Premise Provide a retrospective of the NetSense study and future directions / study evolution On-Going Efforts PhoneLab – Geoffrey Challen SUNY Buffalo LiveLabs – ArchanMisra Singapore Management University
NetSense Cell Phone Study Study how “always-on” networking (i.e. smartphones) impacts how we make / keep friends Provide 200 smart devices to incoming freshmen at Notre Dame (Aug 2011) Sociological Technical Proximity, Facebook, Phone Calls Impact on Usage
Funding “No bucks, no Buck Rodgers” NSF SoCS grant - $750k Three years – 1 year prep + 2 years data 200 users via “smart” devices $75k for devices 200x Nexus S 4G 200 anytime mins Unlimited data, text + =
Approvals Human Subjects Institutional Review Board (IRB) What is being gathered? How often? Consent Forms Deception vs. Transparency Data Security / Access Clear Policies – Overages, Breakages, Repairs Collaborate / Educate Yourself Early
NetSenseAgent – Year 1 • Reflashed with Cyanogenmod • Gingerbread / ICS (2.4+) • “Full” root - monitoring • Proximity via Bluetooth • Mandated w/WiFi on Communications SMS, Call, E-Mail Usage Apps, Tonnage Environment Battery, WiFi APs, B/T Agent application Local SQLite database Threads for “instruments” Periodically phones home
Weekly meetings for 1x year Study Mechanics Cohort selection Notification, response ND Dorm Assignment - July Distribute / manage 200 phones Reflash 200 phones w/CyanogenMod Distribute 200 phones to the student Week before classes
Check-In Architecture Next-generation Play Store based XML Configurability, Rendezvous Server
Active Devices Over Time(Mar 2012) Criterion Must report w/ sufficient data (> 100 reports, on avg per day) Threshold for evaluating “good” users who are fully compliant
Active Devices (Mar 2012) Battery life is tricky to balance Coverage (indoor) plays a huge role in battery life Provider upgrades were “challenging”
Phone Repairs / Effort SD Card (Green) Peaked at end of 1st year Nexus S did not have removable SD card Critical to havealways-on student / liaison to keep tabs
Quiz Response Rate Multiple choice quizzes Do you use WiFi? A. Yes B. No Much worse response rates for long-style tie / behavioral quizzes
Lessons Learned Phone reimbursement / reward for participation Free phone + plan Custom ROM? University-owned phones Purchased phone (subsidized) + free plan Non-compliant users? $100 (Galaxy S3), $200 (Galaxy S4) Subsidy for compliance via pre-paid cards $40 / month for arm band, phone agent, laptop agent Hardship cases? Coercive?
Lessons Learned Points of Contact are Key Undergraduate liaisons S. Pentland @ MIT Paid, < 10 hours per week Configurable Sampling is Essential Probabilistic, Triggered Sampling Don’t Log It – Gone How to Measure User QoE
Lessons Learned Cohort Effects How big of a sample do you need? Is your population biased? Can you quantify the extent of bias? Device Breadth Is a phone enough? How fine of data do you need? Per-packet RSSI? PHY layer? Tower info?
Cohort through May 2015 Cohort of +500 in August 2015* Collaborating Can I get access to your data? Yes* Depending on what data you want Can I run experiments on your phones? SUNY PhoneLab Probably not Singapore LiveLabs Urban Lifestyle Innovation Program
Questions? striegel@nd.edu
Overview • Problem Space • Overview of NetSense Study Goals • Study Process • Funding • IRB Approval • Architecture • Response / Compliance • Lessons Learned / Looking Forward