machine learning family routines

197
Scott Davidoff Dissertation Defense FAMILY CONTROL SMART HOME THE PROJECT ON Routine as Resource for the Design of Learning Systems Scott Davidoff Dissertation Defense

Upload: scott-davidoff

Post on 08-Jul-2015

342 views

Category:

Technology


0 download

DESCRIPTION

An end-to-end HCI project trajectory: How to learn a model of family logistical routines, and present that information to families to help them feel more in control of their lives.

TRANSCRIPT

Page 1: Machine Learning Family Routines

Scott DavidoffDissertation Defense

FAMILY CONTROL SMART HOMETHE PROJECT ON

Routine as Resource for the Design

of Learning Systems

Scott Davidoff Dissertation Defense

Page 2: Machine Learning Family Routines

Scott DavidoffDissertation Defense

A pattern of behavior that is followed

repeatedly, but is subject to change

if conditions change

Winter 1964

“”

1

Page 3: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Koestler 1967

Routines reduce our attention needs

2

Page 4: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Zerubavel 1981

Reduce attention needed for tasks

3

Page 5: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Wakkary & Maestri 2007

Free attention for bigger challenges

4

Page 6: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Wolin + Bennett 1984

5

Page 7: Machine Learning Family Routines

Scott DavidoffDissertation Defense

The cyclic nature of routine makes it

a natural target for machine learning

Page 8: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Machine learning can use this order

7

Page 9: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Suchman 1983, Tolmie et al. 2002

Idiosyncrasies are hard to model

8

Page 10: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Certain human routines can be

modeled, increasing the scope of

activity recognition

Machine learning

Opportunities

9

Page 11: Machine Learning Family Routines

Scott DavidoffDissertation Defense

The ability to use learned routines in

end-user applications would solve a

variety of human problems

HCI

Opportunities

10

Page 12: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Dual-income family logistics

11

Page 13: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Sequence of place and

transportation (rides) that occur on

daily, weekly, and seasonal cycles

“”

12

Page 14: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Darrah et al. 2000

Managing details can be difficult

13

Page 15: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Frissen 2000

Routines give a feeling of control

14

Page 16: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Beech et al. 2004

Life does not always follow routines

15

Page 17: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Breakdowns lead to loss of control

16

Page 18: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Perry et al. 2001, Ling + Campbell 2003

A constant need to follow updates

17

Page 19: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Gneezy + Rustichini 1998, Darrah 2009

A constant source of anxiety

18

Page 20: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can learn a model of family

logistical routines, and present that

information to families to help them

feel more in control of their lives

19

Page 21: Machine Learning Family Routines

Scott DavidoffDissertation Defense 20

Page 22: Machine Learning Family Routines

Scott DavidoffDissertation Defense 21

Page 23: Machine Learning Family Routines

Scott DavidoffDissertation Defense

2We can use sensing

and modeling to

synthesize missing

information resources

3Show how to use the

model and evaluate the

impact of the

information

1We can use fieldwork to

identify missing but

needed information

resources

Fieldwork Modeling Validation

22

Page 24: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Create an ordered list of places and rides

Attend to the details of a plan as it unfolds

Planning and coordinationLogistics

Coordinate

Plan

23

Page 25: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Fieldwork1

24

Page 26: Machine Learning Family Routines

Scott DavidoffDissertation Defense

1805

22 Davidoff et al Ubicomp 2006Months

Months

Months0606

12Families

Families

Families

Davidoff et al Ubicomp 2007

Davidoff, Dey + Zimmerman CHI 2009

45 Months24Families

25

Page 27: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Semi-Structured Interviews

Davidoff et al. Ubicomp 2006

26

Page 28: Machine Learning Family Routines

Scott DavidoffDissertation Defense scott davidoff, min kyung lee, anind dey + john zimmerman

Needs Validation

Page 29: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff et al. Ubicomp 2007

User Enactments

28

Page 30: Machine Learning Family Routines

Scott DavidoffDissertation Defense

User Enactments

Davidoff et al. Ubicomp 2007

29

Page 31: Machine Learning Family Routines

Scott DavidoffDissertation Defense

GPS

Phone Calls

Email

SMS

Calendars

Davidoff, Dey + Zimmerman CHI 2009

30

Page 32: Machine Learning Family Routines

Scott DavidoffDissertation Defense

528 phone interviews

109 activity Interviews

108 calendar months

Davidoff, Dey + Zimmerman CHI 2009

31

Page 33: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Less than 20% of days go exactly as

planned1

Routines are not documented

People have incomplete knowledge

of other people’s routines

People make plans that depend on

incorrect information

2

3

4

32

Page 34: Machine Learning Family Routines

Scott DavidoffDissertation Defense 33

Page 35: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Family E

Routine

Deviation, Scheduled

Deviation, Unscheduled

34

Page 36: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Family EAll Families

Routine

Deviation, Scheduled

Deviation, Unscheduled

35

Page 37: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Less than 20% exactly as planned

36

Page 38: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Routineness of Family E Activities

37

Page 39: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Routineness of Family E Activities

38

Page 40: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Routineness of Family E Activities

39

Page 41: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Routines are not documented

Fieldwork Finding

Davidoff, Dey + Zimmerman CHI 2010

40

Page 42: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

August September October November

41

Page 43: Machine Learning Family Routines

Scott DavidoffDissertation Defense

People have incomplete knowledge

of other people’s routines

Fieldwork Finding

Davidoff, Dey + Zimmerman CHI 2010

42

Page 44: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

43

Page 45: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

44

Page 46: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

45

Page 47: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

46

Page 48: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

47

Page 49: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

48

Page 50: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

49

Page 51: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey + Zimmerman CHI 2010

50

Page 52: Machine Learning Family Routines

Scott DavidoffDissertation Defense

People make plans that depend on

incorrect information

Fieldwork Finding

Davidoff, Dey + Zimmerman CHI 2010

51

Page 53: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Orthodontist

Scouts

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

View of Plan: Dad

Check-up

Track Practice

52

Page 54: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Orthodontist

Scouts

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

View of Plan: S16

Paper Route

Track Practice

53

Page 55: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Orthodontist

Scouts

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

Who Will do the Paper Route?

Check-up

Track Practice

Paper Route

54

Page 56: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Information gaps can break down

coordination

No resources exist to find needed

information

Davidoff, Dey + Zimmerman CHI 2010

55

Page 57: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Modeling2

56

Page 58: Machine Learning Family Routines

Scott DavidoffDissertation Defense

2 31Ride

Detection

Driver

Prediction

Predict

Lateness

57

Page 59: Machine Learning Family Routines

Scott DavidoffDissertation Defense

In Situ

Observation

Maintain Current

Behaviors

Ubiquitous

Sensing

GPS

58

Page 60: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Parent drives kid to an activity

Parent drives kidfromactivity

Pick-ups and drop-offsRide

Pick-up

Drop-off

59

Page 61: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Ride Detection

Page 62: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Day Care

Work

Using GPS to Sense a Drop-Off

Parent Child 61

t3 t4 t5 t6t2t1 t7

Page 63: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Day Care

Work

Using GPS to Sense a Drop-Off

Parent Child 62

t3 t4 t5 t6t2t1 t7

Page 64: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Day Care

Work

Using GPS to Sense a Drop-Off

Parent Child 63

t3 t4 t5 t6t2t1 t7

Page 65: Machine Learning Family Routines

Scott DavidoffDissertation Defense

t3Day Care

t4 t5 t6t2

Work

t1 t7

Using GPS to Sense a Drop-Off

Parent Child 64

Page 66: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Day Care

Work

Using GPS to Sense a Pick-Up

Parent Child 65

t3 t4 t5 t6t2t1 t7

Page 67: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Day Care

Work

Using GPS to Sense a Pick-Up

Parent Child

t3 t4 t5 t6t2t1 t7

Page 68: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Day Care

Work

Using GPS to Sense a Pick-Up

Parent Child

t3 t4 t5 t6t2t1 t7

Page 69: Machine Learning Family Routines

Scott DavidoffDissertation Defense

GPS Ride Detection Evaluation

Family Precision Recall

How many of the sensed rides

are right

How many of the total rides

are sensed?

Pick-up Drop-off Pick-up Drop-off

A .991 .987 .912 .910

B .966 .962 .979 .981

C .913 .824 .971 .921

D .878 .873 .980 .944

E .931 .684 .959 .985

Mean .936 .866 .960 .950

.901 .955

68

Page 70: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Smarter power management

Wi-Fi, Bluetooth, cell tower ID

Apply heuristics to modeling

Cultural norms, Individual Behavior

Single Location Sensor

GPS Sampling Rate

69

Page 71: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Driver Prediction

Page 72: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Name Meaning Values

Ln Location of pick-up or drop-off Place ID

RType Ride type Pick-up, Drop-off

DoW Day of week 0,1,2,3,4,5,6

ToD Discretized time of day (15 min) 1,2,3…96

drivert-j Driver for the last 5 rides to Ln Mom, Dad

Driver distribution model [0,1]

Driver Prediction Model Features

71

Page 73: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Driver Prediction Performance x

Number of Training Weeks of Data

1 24 1 24 1 24 1 24 1 24 1 24 1 24

1 24 1 24 1 24 1 24 1 24 1 24 1 24

72

Page 74: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Driver Prediction Performance x

Number of Training Weeks of Data

1 24 1 24 1 24 1 24 1 24 1 24 1 24

1 24 1 24 1 24 1 24 1 24 1 24 1 24

73

Page 75: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Driver Prediction Performance x

Number of Training Weeks of Data

1 24 1 24 1 24 1 24 1 24 1 24 1 24

1 24 1 24 1 24 1 24 1 24 1 24 1 24

74

Page 76: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Lateness Prediction

Page 77: Machine Learning Family Routines

Scott DavidoffDissertation Defense 76

TnowTideal

Lateness in Minutes

N

u

m

b

e

r

o

f

E

x

a

m

p

l

e

s

Distribution of Parent Late Arrivals

N = 83

Page 78: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Name Meaning Values

R Whether the parent

remembers

True, False

Lchild Location of the child Place ID

Lcurr Current location of a

parent

Place ID

J Driver prediction model Mom, Dad

77

Page 79: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Name Meaning Values

R Whether the parent

remembers

True, False

Lchild Location of the child Place ID

Lcurr Current location of a

parent

Place ID

J Driver prediction model Mom, Dad

T If the parent is traveling True, False

Lstart Starting location of a

parent

Place ID

78

Page 80: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Name Meaning Values

R Whether the parent

remembers

True, False

Lchild Location of the child Place ID

Lcurr Current location of a

parent

Place ID

J Driver prediction model Mom, Dad

T If the parent is traveling True, False

Lstart Starting location of a

parent

Place ID

Empirical cumulative

distribution (ecdf)of

on-time arrivals to Lchild

at time TnowTideal

[0,1]

79

Page 81: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Name Meaning Values

R Whether the parent

remembers

True, False

Lchild Location of the child Place ID

Lcurr Current location of a

parent

Place ID

J Driver prediction model Mom, Dad

T If the parent is traveling True, False

Lstart Starting location of a

parent

Place ID

Empirical cumulative

distribution (ecdf)of

on-time arrivals to Lchild

at time TnowTideal

[0,1]

D Destination of a parent Place ID

80

Page 82: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Name Meaning Values

R Whether the parent

remembers

True, False

Lchild Location of the child Place ID

Lcurr Current location of a

parent

Place ID

J Driver prediction model Mom, Dad

T If the parent is traveling True, False

Lstart Starting location of a

parent

Place ID

Empirical cumulative

distribution (ecdf)of

on-time arrivals to Lchild

at time TnowTideal

[0,1]

D Destination of a parent Place ID

81

Page 83: Machine Learning Family Routines

Scott DavidoffDissertation Defense 82

tideal- 30

A’ = 0.659

.6

.5

Page 84: Machine Learning Family Routines

Scott DavidoffDissertation Defense 83

tideal

A’ = 0.801

Page 85: Machine Learning Family Routines

Scott DavidoffDissertation Defense 84

tideal+10

A’ = 0.826

Page 86: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can learn a model of family

logistical routines, and present that

information to families to help them

feel more in control of their lives

85

Page 87: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Validation3

86

Page 88: Machine Learning Family Routines

Scott DavidoffDissertation Defense 87

Page 89: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Do parents understand the view?

Do parents perceive that the

information is valuable?

Do parents feel more in control

having this view?

88

Page 90: Machine Learning Family Routines

Scott DavidoffDissertation Defense

1257

45Davidoff et al Ubicomp 2006

Davidoff et al Ubicomp 2007

Davidoff, Dey + Zimmerman CHI 2009

Davidoff et al. CHI 2010Months

Months24

24Families

Families

Families

Davidoff et al. Ubicomp 2011

(in preparation)

89

Page 91: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff et al. Ubicomp 2011 in preparation

Experience prototype: Doctor

90

Page 92: Machine Learning Family Routines

Scott DavidoffDissertation Defense 91

Page 93: Machine Learning Family Routines

Scott DavidoffDissertation Defense 92

Page 94: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff et al. Ubicomp 2011 in preparation

Experience prototype: Kitchen

93

Page 95: Machine Learning Family Routines

Scott DavidoffDissertation Defense

It's nice to have a map of the day. I

don't literally write all this stuff down,

so sometimes it's just hard to keep

in my head.

Visual Distills Details

94

−P2“”

Page 96: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Because the calendar [is] all words

and numbers, so you have to really

think about …how long everything

takes…

Visual Simplifies Calculation

95

−P10

Page 97: Machine Learning Family Routines

Scott DavidoffDissertation Defense

…With the visual you can see who's

doing what at what time. You can

make a decision about who's

activity to change.

Visual Simplifies Calculation

96

−P10“”

Page 98: Machine Learning Family Routines

Scott DavidoffDissertation Defense

I would be able to know where

everyone was gonna be, instead of

having to ask around.

Visual Clarifies Intentions

“”97

−P11

Page 99: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Venkatesh + Davis 2003

Technology Acceptance Model-3

98

Page 100: Machine Learning Family Routines

Scott DavidoffDissertation Defense

…helps me do my job “as a parent.”

…would want my family to use it.

…more in control of details, more

ready for changes

Perception of Control

Behavioral Intention

Perceived Usefulness

99

Page 101: Machine Learning Family Routines

Scott DavidoffDissertation Defense 100

Perceived

Usefulness

Behavioral

Intention

Perception of

Control

Supports

Planning

Supports

Awareness

7

6

5

4

3

2

1

Page 102: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can learn a model of family

logistical routines, and present that

information to families to help them

feel more in control of their lives

101

Page 103: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Contributions

102

Page 104: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Suchman 1983, Tolmie et al. 2002

Certain routines can be modeled

103

Page 105: Machine Learning Family Routines

Scott DavidoffDissertation Defense

2 31Ride

Detection

Driver

Prediction

Predict

Lateness

104

Page 106: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Commodity GPS with

real power constraints

105

Page 107: Machine Learning Family Routines

Scott DavidoffDissertation Defense 106

Page 108: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Breakdowns lead to loss of control

107

Page 109: Machine Learning Family Routines

Scott DavidoffDissertation Defense 108

Perceived

Usefulness

Behavioral

Intention

Perception of

Control

Supports

Planning

Supports

Awareness

7

6

5

4

3

2

1

Page 110: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Gneezy + Rustichini 1998, Darrah 2009

A constant source of anxiety

109

Page 111: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Empower Parents

Davidoff et al. Ubicomp 2007

110

Page 112: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Future Work

111

Page 113: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Improve algorithms

Use algorithms to enable other

family coordination systems

Add additional sensor inputs

Extend to other forms of routine

112

Page 114: Machine Learning Family Routines

Scott DavidoffDissertation Defense Neustaedter + Brush CHI 2006 113

Page 115: Machine Learning Family Routines

Scott DavidoffDissertation Defense

6pm

Orthodontist

Neustaedter + Brush CHI 2006 114

Page 116: Machine Learning Family Routines

Scott DavidoffDissertation Defense

6pm

Ortho

Paper Route

Neustaedter + Brush CHI 2006 115

Page 117: Machine Learning Family Routines

Scott DavidoffDissertation Defense Brown et al. Ubicomp 2007 116

Page 118: Machine Learning Family Routines

Scott DavidoffDissertation Defense Brown et al. Ubicomp 2007 117

Page 119: Machine Learning Family Routines

Scott DavidoffDissertation Defense

FAMILY CONTROL SMART HOME

The project on

Page 120: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Thanks for your time, brain power, and support

Page 121: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Thanks to our sponsors

NSF IIS 1017429

Google Research Award

Page 122: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Thanks to our supporters

Page 123: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Thanks to my collaborators

Page 124: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Thanks to my committee

Page 125: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Thanks to my advisors

Page 126: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Questions

Page 127: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Image Creditshttp://thesituationist.files.wordpress.com/2008/06/traffic.jpg

http://kaleidoscope.cultural-

china.com/chinaWH/upload/upfiles/2009-

06/22/unique_morning_exercises_a_daily_routine_for_countles

s_chinese67f17841b1f4bfae458c.jpg

http://travellingcam.files.wordpress.com/2006/11/japan-self-

defense-force-marching-festival-2006-7.jpg

http://thhsmusic.com/Images/2007_Marching_Unit.jpg

http://0.tqn.com/d/golosangeles/1/0/-/H/-/-/2007-juggling3.jpg

http://popten.net/wp-content/uploads/2009/03/traffic-jam.jpg

http://www.lateott.com/images/control_key.jpg

http://stuffunemployedpeoplelike.com/wp-

content/uploads/2009/05/carpool.jpg

http://larchmontgazette.com/commentary/columns/soccermom.j

pg

http://www.schoolbrief.com/SchoolBrief/images/parent.jpg

http://kmcavoy.edublogs.org/files/2010/11/6a00d83451b3c669e

200e54f94c4bb8833-500wi-xl20xk.jpg

http://www.weshipyourcar.com/LinkPageImages/early-flight.gif

http://memory.loc.gov/master/mss/mwright/04/04003/0030d.jpg

http://www.mnn.com/sites/default/files/user-

60/school%20bus_0.jpg

http://arlingtonkids.files.wordpress.com/2011/01/school-bus.jpg

http://4.bp.blogspot.com/_K-

aTC4UHVR0/S7tDj1UfVSI/AAAAAAAAAB4/QTg9MTxuecY/s16

00/confused.jpg

http://cdn.sheknows.com/articles/10-Qualities-SM-Art-Img.jpg

http://i.dailymail.co.uk/i/pix/2009/10/25/article-0-

059F2EC8000005DC-613_468x329.jpg

http://static.tvguide.com/MediaBin/Galleries/Editorial/090511/fat

guys/fatguys-honeymooners8.jpg

http://www.ministryofmanipulation.com/images/8bcube.jpg

http://http//vandoverviewpoints.com/wp-

content/themes/Yamidoo/yamidoo/scripts/timthumb.php?src=htt

p://vandoverviewpoints.com/wp-

content/uploads/2010/04/Family_back.jpg&w=390&h=600&zc=

1

http://www.wmich.edu/registrar/assets/images/photos/calendar2

.jpg

http://www.mediabistro.com/fishbowlDC/files/original/1924_chec

king_facts.jpg

http://www.sbac.edu/~media/images/questionmark.jpg

http://img.dailymail.co.uk/i/pix/2008/02_04/textDM2502_468x35

2.jpg

http://farm3.static.flickr.com/2387/2137448527_eb889a39cb.jpg

http://www.zoodles.com/blog/wp-content/uploads/2010/08/lost-

child4-300x225.jpg

126

Page 128: Machine Learning Family Routines

Scott DavidoffDissertation Defense

QUESTION

SLIDES

127

Page 129: Machine Learning Family Routines

Scott DavidoffDissertation Defense

ROUTINE

MODELING

RELATED WORK

128

Page 130: Machine Learning Family Routines

Scott DavidoffDissertation Defense

FORMULAS

129

Page 131: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Driver Distribution

130

Page 132: Machine Learning Family Routines

Scott DavidoffDissertation Defense

LWDT

131

Page 133: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Bayesian Network

132

Page 134: Machine Learning Family Routines

Scott DavidoffDissertation Defense

RIDE SENSING

TECHNICAL

VERSION

133

Page 135: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff, Dey& Zimmerman CHI 2011

134

Page 136: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Two Kinds of Drop-Offs

135

Page 137: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Two Kinds of Drop-Offs

t3t2 ,t1 ,

136

Page 138: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Two Kinds of Drop-Offs

t3t2 ,t1 ,

137

Page 139: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Two Kinds of Pick-Ups

138

Page 140: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Pick-Up

t3t2 ,t1 ,

139

Page 141: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Pick-Up

t3t2 ,t1 ,

140

Page 142: Machine Learning Family Routines

Scott DavidoffDissertation Defense

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

141

Page 143: Machine Learning Family Routines

Scott DavidoffDissertation Defense

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

142

Page 144: Machine Learning Family Routines

Scott DavidoffDissertation Defense

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

143

Page 145: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

144

Page 146: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

145

Page 147: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

146

Page 148: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

147

Page 149: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Drop-Off

148

Page 150: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Pick-Up

149

Page 151: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Pick-Up

150

Page 152: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Pick-Up

151

Page 153: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Pick-Up

152

Page 154: Machine Learning Family Routines

Scott DavidoffDissertation Defense

P C

5:15

Day Care

5:30 5:45 6:005:00

Work

4:45 6:15

Using GPS to Sense a Pick-Up

153

Page 155: Machine Learning Family Routines

Scott DavidoffDissertation Defense

SCHOOL HALF

DAY

BREAKDOWN

154

Page 156: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Baseball

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

View of Plan: S16

Track Practice

155

Page 157: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Baseball

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

View of Plan: S16

Track Practice

156

Page 158: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Baseball

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

Dad Started Baseball League

Track Practice

157

Page 159: Machine Learning Family Routines

Scott DavidoffDissertation Defense

OTHER

APPLICATIONS

158

Page 160: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Orthodontist

Scouts

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

Detect conflict with unlabeled routine

Track Practice

Check-up

Paper Route

159

Page 161: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Baseball

School

Dad Work

Home

4:00 5:00 6:00 7:003:002:00 8:00

S16Dad

Dad Started Baseball League

Track Practice

160

Page 162: Machine Learning Family Routines

Scott DavidoffDissertation Defense

ACTIVITY

SCATTERPLOT

161

Page 163: Machine Learning Family Routines

Scott DavidoffDissertation Defense

RIDE SENSING

ERRORS

162

Page 164: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Smarter power managementGPS Sampling Rate

163

Page 165: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Smarter power management

Wi-Fi, cell tower, bluetooth

Single Location Sensor

GPS Sampling Rate

164

Page 166: Machine Learning Family Routines

Scott DavidoffDissertation Defense

5:15

Home

5:30 5:45 6:005:00

Work

4:45 6:15

Simultaneous Departure Errors

School

Parent Child165

Page 167: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Smarter power management

Wi-Fi, cell tower, bluetooth

Apply heuristics to modeling

Cultural norms, Individual Behavior

Single Location Sensor

GPS Sampling Rate

166

Page 168: Machine Learning Family Routines

Scott DavidoffDissertation Defense

1:00pm

Home

2:00 3:00 4:0012:00

Work

11:00 5:00

Lost Signal During Daytime

School

Parent Child167

Page 169: Machine Learning Family Routines

Scott DavidoffDissertation Defense

DRIVER

PREDICTION

ERRORS

168

Page 170: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Driver Prediction Performance x

Number of Training Weeks of Data

169

Page 171: Machine Learning Family Routines

Scott DavidoffDissertation Defense

USER

ENACTMENT

PICS

170

Page 172: Machine Learning Family Routines

Scott DavidoffDissertation Defense

User Enactments

Davidoff et al. Ubicomp 2007

171

Page 173: Machine Learning Family Routines

Scott DavidoffDissertation Defense

User Enactments

Davidoff et al. Ubicomp 2007

172

Page 174: Machine Learning Family Routines

Scott DavidoffDissertation Defense

UNUSED PICS

173

Page 175: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Koestler 1967

Routines are not documented

174

Page 176: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Darrah et al. 2000, Frissen 2000, Beech et al. 2004

People can’t see all the facts

175

Page 177: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Davidoff et al. Ubicomp 2006

Managing Rides Causes Anxiety

176

Page 178: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Pentland + Reuter 1994

Graceful

177

Page 179: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Wolin& Bennett 1984

Path for exercise

178

Page 180: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Medved 2004

Confident

179

Page 181: Machine Learning Family Routines

Scott DavidoffDissertation Defense

HCI Approach to Machine Learning

180

Page 182: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Dual-income families

181

Page 183: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Frissen 2000, Fiese et al. 2002, Davidoff et al. 2007

Routines help parents feel in control

182

Page 184: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Frissen 2000, Fiese et al. 2002, Davidoff et al. 2007

Just surviving the day

183

Page 185: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Darrah et al. 2000, Frissen 2000, Beech et al. 2004

Non-Routine Events Cause Anxiety

184

Page 186: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Gneezy + Rustichini 1998, Darrah 2009

A constant source of anxiety

185

Page 187: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Gneezy + Rustichini 1998, Darrah 2009

A constant source of anxiety

186

Page 188: Machine Learning Family Routines

Scott DavidoffDissertation Defense

How do routines provide support?

Davidoff et al. Ubicomp 2007

187

Page 189: Machine Learning Family Routines

Scott DavidoffDissertation Defense

LONG-FORM

APPROACH

188

Page 190: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can use fieldwork to identify

missing but needed information

resources

Fieldwork

189

1

Page 191: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can use sensing and modeling

to synthesize missing information

resources

Modeling

190

2

Page 192: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can develop an application to

deliver synthesized resources and

evaluate how it helps coordination

Show how to use the model and

validate that the application

potentially solves the problem by

Build and validation

191

3

Page 193: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can exploit routine as a

resource to design learning systems

Thesis Statement

192

Page 194: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Understanding how routines support

daily life generate new design ideas

Generative Resource

193

Page 195: Machine Learning Family Routines

Scott DavidoffDissertation Defense

Model information about family

transportation routines to

supplement family knowledge

Information Resource

194

Page 196: Machine Learning Family Routines

Scott DavidoffDissertation Defense

We can exploit routine as a

resource to design learning systems

Thesis Statement

195

Page 197: Machine Learning Family Routines

Scott DavidoffDissertation Defense

context-aware, adaptive, intelligent,

cognitive, expert, decision-support

196

system