Docstoc

Pervasive and Ubiquitous Mobile Computing

Document Sample
Pervasive and Ubiquitous Mobile Computing Powered By Docstoc
					                   Impact of Advertising in
                   Pervasive and Ubiquitous
                              Eric Wang
                            April 13, 2011
                                CS219

                                              1

Wednesday, April 13, 2011
                            Overview
     • Review of Location/Activity Awareness
     • Understanding Ubiquitous Computing
     • Economics of Mobile Services
       (Advertising)
     • Overview of Industry Trends
     • Implications on Privacy
     • Final Thoughts


                                               2

Wednesday, April 13, 2011
            Recap of Location/Activity
     • Location positioning
           – Outdoor: GPS / A-GPS / WiFi MAC look-up
           – Indoor: WiFi triangulation / BT / RFID
           – Place detection

     • Activity classification
           – User preferences
           – Accelerometer (gait detection)
           – Characterizing motion patterns
                 • Basic: Sitting / Walking / Running
                 • Advanced: Eating / Reading (augmented with
                   additional sensors)                          3

Wednesday, April 13, 2011
                            What is Ubiquitous
     • Ubicomp is the concept of technology
       becoming invisible in our lives

     • Leverages location/activity-awareness

     • Built into things we use (including spaces)
           – Cars, Phones, Homes, Malls

     • Makes technology more natural, powerful,
       and effective

     • Huge economic impact
                                                     4

Wednesday, April 13, 2011
                        Empowering Mobile
     • Location & activity information
       augments user experience

     •    Keeps track of our agendas
     •    Suggests where to go/eat
     •    Suggests what to buy
     •    Reconnect with local people


                                            5

Wednesday, April 13, 2011
                        Empowering Mobile
     • …but at what cost?

     • We can be tracked (for good
       purposes?)
           – Specific locations/routes
           – Detailed shopping habits
           – Web browsing habits



                                            6

Wednesday, April 13, 2011
                   Economics of Services:
     • Ubicomp and advertising share the same
       goal:
           – Targeting the user
           – Evaluate effectiveness/awareness
           – Protecting privacy

     • Advertising supports ubiquitous
       computing
           – Ex: Nearby friend finder app also suggests
             meeting places in the area


                                                         7

Wednesday, April 13, 2011
                   Economics of Services:
     • Why advertising isn’t going away:
           – In a 2004 study, online ads yielded 670%
             more money than paid-for online content
           – Only 5.9% of people willing to pay $4/mo
             for ad-free services

     • Consumers “hate ads,” but not enough
       to pay to remove
     • Advent of ad-supported “free” services

                                                        8

Wednesday, April 13, 2011
                  Targeting Ads to Users
     • Traditional approach:
           – Segment consumer
             population into
             subgroups
                 Demographics
 Location
                 Psychographics
 Age *
                 Behavior

                   
     Benefits Sought


           – More abstract
             segmentation model:
             VALS (Values and

                                           9

Wednesday, April 13, 2011
                  Targeting Ads to Users
     • Better approach:
           – Behavioral targeting
           – Google AdSense tracks history of websites
             visited
                            time: 06/Aug/2008 12:01:32
                            ad_placement_id: 105
                            ad_id: 1003
                            userid: 0000000000000001
                            client_ip: 123.45.67.89
                            referral_url: "http://youtube.com/
                            categories"


           – Some ISPs experimented analyze packets to
             characterize user online activity (browsing, e-
             mail, downloading)
                                                                 10

Wednesday, April 13, 2011
          Ad Targeting with Ubicomp
     • Location sensing: Using WiFi AP / cell
       towers for ads

     • Ads on mobile devices:
           – Skyhook Wireless serves location-based
             ads (e.g. coupon for a store 0.2 miles
             away)
           – Acuity Mobile uses location (current and
             predicted) & time of day to deliver ads to
             mobile devices indoors
                                                          11

Wednesday, April 13, 2011
          Ad Targeting with Ubicomp
     • Traditional-looking ads that
       change
           – Vert developed ad sign on taxis
             that changes based on location
           – Shopping carts equipped with
             location sensors and digital
             displays
           – Video ads inside transit buses *

     • Alaris Media Network
       deployed e-billboards
           – Sensors detect which radio
             stations passing cars have tuned
             in
           – Radio stations  demographics
                                                12

Wednesday, April 13, 2011
       Inference & Context-sensing
     • Predict driver destination
           – Microsoft Multiperson Location Survey collected
             data about 200 drivers, covering 135,000 km
           – Bayesian Inference modeled via four factors
                 • Ground cover, driving efficiency, time, personal dest.
                 • Median error starts ~3km, halfway drops to 2km, then
                   ends 1km




                                                                            13

Wednesday, April 13, 2011
               Context-sensing: Effects
                            Ad targeting under different contexts




     • Bottom-line: Knowing context is immensely useful
           – Better targeted ads
           – Low-power applications
           – New applications

                                                                    14

Wednesday, April 13, 2011
       Context-sensing: Challenges




                                 15

Wednesday, April 13, 2011
             User Attitudes on Location /
                  Activity Awareness
     • Performance of ads
           – Only 1/3 of marketers consider ads
             effective
           – Only 41% of money spent produces a sale


     • Assessing effectiveness
           – Eye-level camera-equipped billboards
             measuring duration of consumer attention
           – Analyze images to infer gender, age
                                                       16

Wednesday, April 13, 2011
             User Attitudes on Location /
                  Activity Awareness
     • Available mobile sensors from Intel
       Research for activity sensing:
           – Microphone, visible light, infrared,
             accelerometer, barometer, thermometer,
             humidity, compass, …


     • Yet, some activities too hard to infer
       from sensors

                                                      17

Wednesday, April 13, 2011
          User Feedback on Ubicomp
     • Experience sampling to directly
       query user about current activity and
       attitude

     • MyExperience (introduced last
       week)
           – Framework for collecting both
             quantitative and qualitative data on
             behavior in the field


     • Proactive Experience Sampling
                                                    18

Wednesday, April 13, 2011
                Today’s Industry Trends
     • Amazon now sells ad-supported $25 cheaper
       version of Kindle (shows up as screensaver /
       bottom of home screen)
           – Ads could freeze the device until user clicks a
             button or dutifully acknowledges the commercial
             message
           – $114 for ad-supported vs $189

     • Pandora’s Android app sends personal data to
       ad servers
           – Sends user GPS, sex, birthdate, Android ID
           – Argues data needed for Pandora music-
             suggestion service, but very easy to uniquely
             identify user
                                                               19

Wednesday, April 13, 2011
                  Implications on Privacy
     • Annoying obligation bolstered by new online
       privacy laws
     • Privacy systems exist  need to be easy to
       configure and transparent to the user
                                  PawS privacy
                                  management
                                  system
                                  •Policy notice
                                  •Choice and
                                  consent
                                  •Locality of person
                                  using data
                                  •User access
                                                        20

Wednesday, April 13, 2011
                            Final Thoughts
     • Past 2 decades:
           – Expansion of pervasive network includes
             global information network on online
             behavior patterns

     • Next 2 decades:
           – Proliferation of pervasive technologies will
             include physical behavior patterns
           – Much better accuracy in modeling user
             interests & preferences
           – Ability for ubicomp systems to target based
             on what people do, not what they click
                                                            21

Wednesday, April 13, 2011
                            Final Thoughts
     • Today’s developers should keep in
       mind
           – Design for ad-supported ubicomp
             services, ensuring a smoothly integrated
             experience
           – If ads are inappropriate, consider open-
             source / subscription / donation business
             model
           – Understand how much info users are

                                                     22

Wednesday, April 13, 2011
                                Thank you.
     •    References:
           – A Framework of Energy Efficient Mobile Sensing for Automatic Human State
             Recognition John Krumm and Eric Horvitz, "Predestination: Where Do You
             Want to GoToday?", IEEE Computer Magazine, vol. 40, no. 4, April 2007, pp.
             105-107.

           – John Krumm, "Ubiquitous Advertising: The Killer Application for the 21st
             Century," IEEE Pervasive Computing Magazine, 10 (1), January - March, 2011.

           – E. Kaasinen. User needs for location-aware mobile services. Personal and
             Ubiquitous Computing, 7(1):70–79, 2003.

           – Langheinrich, M. A Privacy Awareness System for Ubiquitous Computing
             Environments. In Proceedings of Ubicomp 2002. Goteberg, Sweden. pp.
             237-245 2002.
                                                                                          23

Wednesday, April 13, 2011

				
DOCUMENT INFO
Shared By:
Tags:
Stats:
views:34
posted:8/3/2011
language:English
pages:23