exploring emergent consumer experience: a topological data analysis approach

© Novak and Hoffman 2015 | http://postsocial.gwu.edu

Exploring Emergent Consumer

Experience: A Topological Data

Analysis Approach

Tom Novak

Donna Hoffman

The George Washington University

Center for the Connected Consumer

CCB 2015 Annual Complexity in Business Conference, November 12-13, 2015

1

http://postsocial.gwu.edu


Ready or Not, The Smart Home is Coming Soon!

http://www.jklossner.com/TechToons/

2



› Setting the Stage

› A New Framework

› Uncovering the Possibility Space with Topological Data Analysis (TDA)

› TDA Application: Smart Home Sensors and Activities

› TDA Application: IFTTT Rules

Agenda

3



Setting The Stage

4



Internet Phase 1

Internet of Information

(Web)

Internet Phase 2

Internet of People

(Social)

Internet Phase 3

Internet of Things

(Post-Social)

Research Focus online experience social mediaconsumer experience of the

assemblage

Catchphrase

“Nobody knows you’re a dog” “On the Internet, everybody knows you’re a dog”

“On the Internet of Things, nobody knows you’re a

fridge”

5

From the Internet to IoT


© Novak and Hoffman 2015 | http://postsocial.gwu.edu 6

The Evolution of InteractivityInternet Phase 1

Internet of Information(Web)

Internet Phase 2Internet of People

(Social)

Internet Phase 3Internet of Things

(Post-Social)

Nature of Interactivity

Many to many interaction between people and content via Web interfaces.

Google

Many to many interaction directly between people via social networks.

Facebook

C2M, M2M, M2P, C2C interactions; device interactions autonomous; Digital → physical. Interactions highly heterogeneous, ongoing and evolve over time.

??

Identity of the Internet

Shop online, browse web pages, search for information are all largely static.

Global Collective Identity

Shift to smarter apps to enable more sophisticated and complex interactions. Balance of power shifts from marketer to consumer.

Global and Personal Collective Identity

C2C interactions recede compared to evolving heterogeneous interactions of C2M, M2M, M2P in overlapping assemblages.

New Personalized Consumer Experiences Will Emerge in the Context of the Unique Identities of These New Assemblages



Why Now? The 7 Technology Laws

Technology Law Description

#1 Moore’s Law Processing power. Transistor density on integrated circuits doubles every 12-24 months.

#2 Kryder’s Law Storage power. The density of information on hard drives doubles every 13 months.

#3 Gilder’s Law Communications power. Total bandwidth of communication systems doubles every 6 months.

#4 Kurzweil’s Law Accelerating returns. The time interval between salient technology events shorter as time passes.

#5 Weiser’s Law Instant adaptation. As technology becomes ubiquitous, people instantly adapt to new technology and take it for granted.

#6 Meeker’s Law 10x Multiplier Effect. With each new technology cycle, the number of devices increases tenfold.

#7 Metcalfe’s Law Network power. The value of a network is proportional to the square of the number of users.



The Consumer Internet of Things (IoT)

The collection of everyday objects in the physical environment, embedded with technology including sensors, actuators and the ability to communicate wirelessly with the Internet. These devices interact and communicate with themselves and each other –and with humans.

-- (Hoffman and Novak 2015)



The Internet of Things is Going to Be Huge

250 Million Connected Cars by 2020 (Telefonica)

$19 Trillion opportunity by 2020

(Cisco)

“100% IoT” by 2020 (CES)

Intel’s IoT group had 19% increase + $2.1

billion revenue in 2014

6.4 billion by 2016 and 21 billion by 2020 (Gartner)



A lot of hype, but so far not a lot of adoption

Industry research suggests that consumer adoption of smart devices, the “Internet of Things,” is inevitable (Acquity Group 2014; Affinova 2014), with ⅔ of consumers saying they plan to buy at least one smart home device in the next five years.

Adoption rates are low:16% own one device and 4% own two or more (Gartner)6% use smart home tech (Nielsen)4% own one device (Acquity)

Even the most aggressive projections suggest that only 30% are expected to purchase a smart thermostat (one of the most obvious smart home applications) 5 years from now, with much lower rates of adoption for other smart home devices.

But the Consumer IoT Has an Adoption Problem

10



Even Early Adopters are Having Trouble



Awareness. 87% of consumers have never heard of the “Internet of Things” and if they have, they don’t really know what it is.

Price, Security, Privacy and a Loss of Control. Oh, and smart home device prices are too high, there are serious concerns about security and privacy, and consumers worry that smart devices may develop scary minds of their own.

Product Value and Performance. Industry research shows that the main barriers are a lack of awareness and a perception that the value proposition is missing. Many current products simply aren’t ready for prime time.

3 Barriers to Adoption of Smart Devices

12



For the consumer IoT to expand beyond upscale consumers and techy DIYers willing to suffer, we need to understand the value it offers to consumers.

The current focus is on individual products (thermostat, light bulb, refrigerator) and specific “use cases” (turn on the lights when I get home).

Many consumers are having a hard time seeing the value from that focus.

We believe that value is embedded in what smart devices means to consumers and want to shift the conversation to smart device identity and consumer experience.

For this, for this we need a new framework...

Cracking the Value Code

13



Framework

14



What kind of insights can we derive about emergent

consumer experience in the IoT from all these

heterogeneous interactions?

These interactions create a whole that is more than

the sum of the parts – a set of recurrent

“assemblages” (Hoffman and Novak 2015).

Just as the web needed new frameworks for

understanding consumer experience (Hoffman and

Novak 1996), the IoT will need new frameworks to

understand the consumer experience that emerges

from these interactions.

Interactivity is Evolving and New Consumer

Experiences are Emerging

15



The Smart Home Consumer Starts with A Device or Two



The Consumer’s Focus Shifts Once There Are 3-4 Devices



People Start to Want the Devices to Talk to Each Other



Assemblage Theory. A comprehensive theory from the neo-realist school of philosophy which explains the processes by which the identity of a whole - a whole that is more than the sum of the parts - emerges from on-going interactions among its parts .

(Deleuze and Guattari 1988; DeLanda 2002, 2006, 2011; Harman 2008).

Smart Devices - An Assemblage of Heterogeneous Parts



CONSUMER DEVICEFrom assemblage theory, interaction

involves paired capacities (Hoffman and

Novak 2015; DeLanda 2006)

capacity to affect

(IF “trigger”)

+ capacity to be affected

(THEN “action”)

The entities in an IoT interaction are either

consumers or devices (sensors, beacons,

smart products, hubs, wearables, etc.).

Interaction is the Fundamental Unit of Analysis in the

Post-Social Internet of Things (IoT)

IF I trigger a beacon

by entering a room

THEN Philips Hue

turns on the lights

DEVICE CONSUMER

IF my leak detector is

triggered by a broken

pipe

THEN my smart

home hub sends me

a text

DEVICEDEVICE

IF a motion sensor

detects activity

THEN my security

camera starts recording

Hoffman and Novak (2015) “Emergent Experience and the Connected Consumer in the Smart Home Assemblage and the IoT”



Some Interactions Define “Use Cases”

When the garage door opens or closes -

Then turn the kitchen lights red.

“Safety and Security”



In a 2010 White Paper, IBM introduced four areas of smart home capabilities demonstrating early adoption: 1) entertainment and convenience, 2) energy management, 3) safety and security, and 4) health and welfare

In 2015, Lowe’s markets an Iris “Safe & Secure” kit, and a “Comfort & Control” kit.

In 2015, SmartThings markets the benefits of system as “Home Security,” “Peace of Mind,” and “Limitless Possibilities”

These benefits derive from specific use cases.

Use Cases Dominate Current Smart Home Marketing

22



But - Uses Cases Don’t Explain What Can Emerge

Safety and Security

Awareness of Energy Usage

Control and Convenience

“My house cares

about me”



The Home and Consumer Develop Emergent

Capacities From Interactions

Consumer Smart Home

remotely control camera

stream live videoview video live stream

move cameratime 1



New Capacities Emerge From Interactions

Consumer Smart Home



feel secure

move camera

enable securitytime 2

time 1



New Capacities Emerge From Interactions

Consumer Smart Home



feel secure

move camera

enable security

grant delivery person entry open front doortime 3

time 2

time 1



A Framework for Consumer Experience in the IoT

27

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2648786


http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2648786


Assemblage theory assumes an underlying

topological space of possibilities.

An actual assemblage is an individual singularity

that has been realized in this possibility space.

The possibility space contains universal

singularities - topological invariants – that structure

the possibility space and influence the population of

individual singularities that will emerge (DeLanda 2006, p30).

Thus, consumer experience is not only concerned

with “what is” for a given consumer (a single

consumer-IoT assemblage), but also with the

underlying structure of “what could be” for all

consumers (the larger population of all consumer IoT

assemblages).

Uncovering the Possibility Space of Consumer Experience

from Interactions in the IoT



Uncovering the Possibility Space with TDA

29



Interactivity is Evolving and New Consumer Experiences

are Emerging

What kind of insights can we derive

about emergent consumer experience

in the IoT - the “possibility space” -

from actual interactions?

These interactions represent a lot of

digital “big data” – very high

dimensionality data consisting of often

millions of ongoing interactions among

complex, heterogeneous component

devices (and consumers!).

30



Predictive Analytics approach: Fit predictive models to the data. But the complexity of the data means

hypothesis testing is often challenging. We need to know what questions to ask. Are we asking the right

questions? With big data, insights can be slow.

Conventional approaches for reduction and visualization: Use linear and nonlinear dimension

reduction techniques such as PCA, MCA, and MDS. But, even if they work, are sensitive to distance metrics

and do not preserve topological structures of the data.

Data-Driven Discovery Approach: Hypothesis-free approach based on computational topology to

qualitatively analyze functions on very high-dimensional data and visualize the data structure in low-

dimensional topological spaces. Topological data analysis (TDA) reveals structures in the data that have

invariant properties and can propel insight and improve hypothesis-generation and predictive modeling;

“digital serendipity” (Singh 2013).

132 million data points

Making Sense of of Digital Big Data from the IoT

31



TDA Framework for Generating Topological Networks

Data

Metric and Lenses

Cover Image

Cluster Inverse Image

Nodes and Edges

Network Visualization

Computational topology technique on complex high-dim data using unsupervised machine learning and network visualization.

Result is 3-dim topology of simplicial complexes (discrete, combinatorial objects) in which groups of data are represented as nodes that contain rows that are similar to each other in the high-dim space and edges connect nodes that share rows (Carlsson 2009; Lum, et.al. 2012).



TDA draws on theory of topological spaces and simplicial complexes (algebraic topology); implementation invokes computational topology (computational geometry, computational complexity theory, and computer science) – e.g. see Carlsson2009; Lum, et.al. 2012; Singh, Memoli and Carlsson 2007.

TDA applies a function (lens) to a data set and builds a compressed summary of the data.

A visual network of nodes (representing data points) connected by edges is created using four types of parameters:

Metric (measure of similarity)

Lenses (functions on the data)

Bin resolution

Bin overlap

How TDA Works

Slide image courtesy of Ayasdi, Inc. http://ayasdi.com/

Implementation: Ayasdi 3.0 Software Platform (ayasdi.com, Ayasdi, Inc., Menlo Park, CA) and Python Mapper (Singh, Memoli and Carlsson 2007; http://danifold.net/mapper/).


http://ayasdi.com/

http://ayasdi.com/

http://danifold.net/mapper/


Basic Steps of Implementation: 1. Specify a Metric and Real-

Valued Functions to Calculate an Image of the Data

We choose a metric to represent the distance or similarity among rows.

We choose real-valued functions (lenses) (i.e. the image) to be applied to the data.

Here, we map data points to their y-coordinate value.

y-coordinate function



http://ayasdi.com/


Basic Steps of Implementation: 2. Cover Image

We put the data into overlapping bins based on a chosen resolution(number of groups for each function) and gain(degree of overlap of each group within each function).

The functions use the metric to transform each row in the dataset into a single point that fall into overlapping bins.

y-coordinate function



http://ayasdi.com/


Basic Steps of Implementation: 3. Cluster Inverse Image

Cluster the data in the cover based on the metric. Rows that fall within each bin are independently clustered, using a measure of similarity (the metric) on the original data (i.e. the inverse image of the functions).

Ayasdi uses single-linkage clustering with a fixed heuristic for the choice of the resolution parameter.

Each within-bin cluster becomes a node in the network. Nodes represent sets of data points that are similar with respect to the metric.



http://ayasdi.com/


Basic Steps of Implementation: 4. Network Visualization

Edges connect nodes with rows in common.

Since the data were divided into overlapping bins, a data point can be in multiple nodes.

The resulting nodes and edges construct a topological network of the simplicial complex.

A visualization is created whose patterns can be interpreted for meaning.



http://ayasdi.com/


Metrics: correlation, Euclidean distance, cosine, hamming, categorical cosine, user-defined…

Functions: mean, variance, density, centrality, PCA, MDS, user-defined…

Supervised Machine Learning Models: TDA of machine learning outputs with outcome variables can enhance models through discovery of systematic error and construction of local models as opposed to a single global model

Metrics and Functions



http://ayasdi.com/


TDA Application: Smart Home Sensors and Activities

For a smart home to be smart, it needs to be able to understand what the people in the home are doing.



TDA Example #1: Smart Home Sensors and Activities

ID day/time activity sensor

1

2

Phone Call

Wash Hands

Cook

Eat

Clean

Phone Call

Wash Hands

Cook

Eat

Clean

24 participants were asked to perform 5 activities in a fixed sequence, in an apartment equipped with motion and item sensors:

1. Make a phone call to retrieve a voicemail with a recipe

2. Wash their hands

3. Cook a pot of oatmeal according to the directions in the voicemail

4. Eat the oatmeal

5. Clean the dishes

A total of 24 sensors recorded motion (11 sensors), items (10 sensors), and water and burner usage (3 sensors) as the participants performed the 5 activities in sequence. Time was recorded for each sensor event.

Data source: Dataset 1, ADL Activities, CASAS, Washington State University, http://casas.wsu.edu/datasets/

D. Cook and M. Schmitter-Edgecombe, Assessing the quality of activities in a smart environment.

Methods of Information in Medicine, 2009.

tim

e

IDDAY/TIME ACTIVITY SENSOR


http://casas.wsu.edu/datasets/


Map of Smart Home Sensor Locations

Phone book sensor

Oatmeal

Raisins

Brown

Sugar

Bowl Spoon

Medicine

container

Cabinet sensor

Pot

Burner

Hot

Water

Cold

Water

Motion Sensors Item Sensors

Gas & Water Sensors

Phone sensor

Data source: Dataset 1, ADL Activities, CASAS, Washington State University,





Sample of Smart Home Event Stream Data

ID DATE TIME SENSOR ACTIVITY (“ground truth”)

1 2008-02-27 12:50:10.25482 AD1-B wash hands

1 2008-02-27 12:50:25.21543 AD1-B wash hands

1 2008-02-27 12:50:38.708635 M17 wash hands

1 2008-02-27 12:50:40.5204 AD1-B wash hands

1 2008-02-27 12:50:42.521531 M17 wash hands

1 2008-02-27 12:50:43.533263 M17 cook

1 2008-02-27 12:50:56.242 I07 cook

1 2008-02-27 12:51:01.601718 D01 cook

1 2008-02-27 12:51:03.797739 I02 cook

1 2008-02-27 12:51:05.950145 I03 cook

1 2008-02-27 12:51:09.800081 I01 cook

1 2008-02-27 12:51:10.4779 M17 cook

1 2008-02-27 12:51:11.473605 I04 cook

1 2008-02-27 12:51:12.497614 I05 cook

1 2008-02-27 12:51:16.729926 D01 cook

There are total of 6425 sensor events from 24 participants.

Data source: Dataset 1, ADL Activities, CASAS, Washington State University,





ID activity sensor

Current

event

time

(seconds)

Window

start time

(seconds)

Window

duration

(seconds) AD

1_

A

AD

1_

B

AD

1_

C

D0

1

E01

I01

I02

I03

I04

I05

I06

I07

I08

M0

1

M0

7

M0

8

M0

9

M1

3

M1

4

M1

5

M1

6

M1

7

M1

8

M2

3

aste

risk

1 cook I07 166 133 33 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 167 135 32 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook M17 170 137 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 173 137 36 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 173 140 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 176 151 25 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 176 153 23 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 178 153 25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 180 154 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 180 154 26 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-A 182 156 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 185 156 29 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-A 188 158 30 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 188 160 28 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 190 161 29 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook M17 194 161 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook I05 199 162 37 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-A 200 163 37 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 207 164 43 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook I05 207 166 41 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 213 167 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 215 170 45 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 219 173 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook M17 221 173 48 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

TIME SENSOR FOR CURRENT EVENT

43

Step 1: Create Binary Variables for Each of 24 Sensors

Sensor AD1-A (burner) was triggered at 215 seconds for participant ID = 1

Time

begins with

0 for each

participant.



ID activity sensor

Current

event

time

(seconds)

Window

start time

(seconds)

Window

duration

(seconds) AD

1_

A

AD

1_

B

AD

1_

C

D0

1

E01

I01

I02

I03

I04

I05

I06

I07

I08

M0

1

M0

7

M0

8

M0

9

M1

3

M1

4

M1

5

M1

6

M1

7

M1

8

M2

3

aste

risk

1 cook I07 166 133 33 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 167 135 32 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook M17 170 137 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 173 137 36 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 173 140 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 176 151 25 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 176 153 23 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 178 153 25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 180 154 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 180 154 26 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-A 182 156 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 185 156 29 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-A 188 158 30 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-B 188 160 28 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 190 161 29 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook M17 194 161 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook I05 199 162 37 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook AD1-A 200 163 37 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 207 164 43 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook I05 207 166 41 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 213 167 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook AD1-A 215 170 45 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

1 cook M17 219 173 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

1 cook M17 221 173 48 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0

TIME SENSOR FOR CURRENT EVENT

44

Step 2: Define Sliding Window of the Past 20 Events

Fixed window from time ti-19 to ti sums over the last 20 events, for 24 sensors

win

do

w o

f 20 e

ven

ts



ID activity sensor

Current

event

time

(seconds)

Window

start time

(seconds)

Window

duration

(seconds) AD

1_

A

AD

1_

B

AD

1_

C

D0

1

E01

I01

I02

I03

I04

I05

I06

I07

I08

M0

1

M0

7

M0

8

M0

9

M1

3

M1

4

M1

5

M1

6

M1

7

M1

8

M2

3

aste

risk

AD

1_

As

AD

1_

Bs

AD

1_

Cs

D0

1s

E01

s

I01

s

I02

s

I03

s

I04

s

I05

s

I06

s

I07

s

I08

s

M0

1s

M0

7s

M0

8s

M0

9s

M1

3s

M1

4s

M1

5s

M1

6s

M1

7s

M1

8s

M2

3s

aste

risk

_s

1 cook I07 166 133 33 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 1 0 0

1 cook M17 167 135 32 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 1 0 0

1 cook M17 170 137 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 10 0 0 0

1 cook AD1-A 173 137 36 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 10 0 0 0

1 cook M17 173 140 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 1 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 10 0 0 0

1 cook AD1-A 176 151 25 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 1 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-B 176 153 23 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 1 0 0 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook M17 178 153 25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 2 1 0 0 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-A 180 154 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 1 0 0 0 1 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-B 180 154 26 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 2 0 0 0 0 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-A 182 156 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 2 0 0 0 0 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-B 185 156 29 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 3 0 0 0 0 0 2 1 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-A 188 158 30 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 3 0 0 0 0 0 2 0 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-B 188 160 28 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 4 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook M17 190 161 29 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 5 4 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook M17 194 161 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 5 4 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook I05 199 162 37 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 4 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-A 200 163 37 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 4 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 207 164 43 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 6 4 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook I05 207 166 41 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 4 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 213 167 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 6 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-A 215 170 45 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 219 173 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 7 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 221 173 48 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 6 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0

TIME SENSOR FOR CURRENT EVENT SENSOR SUMS IN SLIDING WINDOW

45

Step 3: Create Sums of Sensor Events in Sliding Window

Sum of the number of times each sensor was triggered in the sliding window

See: Cook, Krishnam and Rashidi (2013), Krishnan and Cook (2014), Cook and Schmitter-Edgecombe (2009)



ID activity sensor

Current

event

time

(seconds)

Window

start time

(seconds)

Window

duration

(seconds) AD

1_

A

AD

1_

B

AD

1_

C

D0

1

E01

I01

I02

I03

I04

I05

I06

I07

I08

M0

1

M0

7

M0

8

M0

9

M1

3

M1

4

M1

5

M1

6

M1

7

M1

8

M2

3

aste

risk

AD

1_

As

AD

1_

Bs

AD

1_

Cs

D0

1s

E01

s

I01

s

I02

s

I03

s

I04

s

I05

s

I06

s

I07

s

I08

s

M0

1s

M0

7s

M0

8s

M0

9s

M1

3s

M1

4s

M1

5s

M1

6s

M1

7s

M1

8s

M2

3s

aste

risk

_s

1 cook I07 166 133 33 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 1 0 0

1 cook M17 167 135 32 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 1 0 0

1 cook M17 170 137 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 10 0 0 0

1 cook AD1-A 173 137 36 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 10 0 0 0

1 cook M17 173 140 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 1 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 10 0 0 0

1 cook AD1-A 176 151 25 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 1 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-B 176 153 23 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 1 0 0 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook M17 178 153 25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 2 1 0 0 0 2 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-A 180 154 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 1 0 0 0 1 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-B 180 154 26 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 2 0 0 0 0 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 9 0 0 0

1 cook AD1-A 182 156 26 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 2 0 0 0 0 1 2 1 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-B 185 156 29 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 3 0 0 0 0 0 2 1 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-A 188 158 30 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 3 0 0 0 0 0 2 0 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-B 188 160 28 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 4 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook M17 190 161 29 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 5 4 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook M17 194 161 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 5 4 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook I05 199 162 37 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 4 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-A 200 163 37 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 4 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 207 164 43 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 6 4 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook I05 207 166 41 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 4 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 213 167 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 6 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0

1 cook AD1-A 215 170 45 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 219 173 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 7 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0

1 cook M17 221 173 48 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 6 4 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0

TIME SENSOR FOR CURRENT EVENT SENSOR SUMS IN SLIDING WINDOW

46

Step 4: Use the Data for Topological Data Analysis (TDA)

1) TDA uses rectangular matrix of events (rows) by 3 time and 24 sensor sums variables (columns)

2) Actual activity is “ground truth” (i.e., what we know and want the TDA to be able to detect).

3) The specific sensor triggered and ID are explanatory variables



Rectangular data matrix of processed sensor events using a simple baseline approach:

› Rows: current sensor event is the unit of analysis

› Columns: sums of 24 sensor events over a 20 event sliding window, time of current

event, time of first event in window, duration time of window

Distance among sensor events (rows) was obtained using the normalized correlation

metric. Columns were normalized to a mean of 0 and variance of 1, and Pearson

correlation of each pair of rows was obtained.

Ayasdi TDA lens was the x and y coordinates of a 2-dimenisonal embedding of a k-

nearest neighbors graph of the data.

TDA of Smart Home Sensor and Activity Data

Objective: Use TDA to produce a topological summary of the stream of sensor events.



Apartment Activities (Colored by Time of Events)

Metric: Normalized Correlation

Lens: K-nearest neighbors (resolution 30, gain 3)

*We used the Ayasdi 3.0 software platform

(ayasdi.com, Ayasdi, Inc., Menlo Park, CA) to

perform the TDA on the CASAS data

Blue early times

Green intermediate times

Red later times



Apartment Activities (Colored by # of Phone Call Events)

88% of Phone Call

events are in Group 1

12% of Phone Call


Look in phone book,

start phone call

M13 (dining) 48%

M14 (dining) 14%

M08 (dining) 6%

M07 (dining) 6%

M01 (dining) 6%

M09 (dining) 5%

Phone book 4%

Phone on 3%

Hang up phone,

approach kitchen

M13 (dining) 51%

M14 (dining) 29%

Phone off 21%

Note: individual sensors are shown that are triggered significantly more

often for this activity in a circled group (hypergeometric p-value <.05)



Apartment Activities (Colored by # of Wash Hands Events)

59% of Wash Hands


20% of Wash Hands


21% of Wash Hands


Near Sink

Hot Water 19%

M16 (kitchen) 13%

M15 (kitchen) 12%

M14 (dining) 12%

Near Sink

Hot Water 13%

M14 (dining) 13%

M15 (kitchen) 12%

M16 (kitchen) 10%

At Sink

Hot Water 20%

M17 (stove) 46%

M18 (sink) 26%



Apartment Activities (Colored by # of Cook Events)

98% of Cook




TDA Shows There are Two Parts to Cooking

Preparing Oatmeal

Burner 22%

Oatmeal 3%

Raisins 3%

Brown sugar 3%

Bowl 3%

Measuring spoon 3%

M17(stove) 45%

Cooking Oatmeal

Burner 33%

M17 (stove) 54%

Meal Preparation

Meal Cooking



Apartment Activities (Colored by # of Eat Events)

24% of Eat


28% of Eat


47% of Eat


Get Medicine and Water

Cold Water 13%

Medicine container 6%

Cabinet 9%

Brown sugar 3%

M16 (kitchen) 10%

M15 (kitchen) 8%

Walk to Table and Eat


M13 (dining) 26%

M14 (dining) 23%

M15 (kitchen) 11%

M16 (kitchen) 11% Eat

M14 (dining) 45%

M13 (dining) 40%

M15 (kitchen) 5%

M16 (kitchen) 5%



Apartment Activities (Colored by # of Clean Events)

68% of Clean

events are in Group 517% of Clean


15% of Clean


Put items away and start cleaning

Hot Water 24%

Cabinet 6%

Oatmeal 5%

Bowl 4%

Measuring spoon 3%

Raisins 3%

M18 (sink) 24%

Cleaning

Hot Water 32%

Cold Water 8%


M18 (sink) 11%Cleaning

Hot Water 29%

Cold Water 15%


M18 (sink) 11%

M15 (kitchen) 4%



Implications of TDA Analysis of Smart Home Data

TDA recovers known activities based on a simple baseline method for processing event streams using sliding windows of counts of sensors, plus start, stop and duration time of the window.

TDA provides clear visual input on the performance of a given approach to processing event stream data. For example, cooking has a clear structure with 2 stages.

Next step is to improve activity recognition by modifying how the event stream is processed, for example by using alterative window sizes, fixed vs. variable windows, exponential time decay on event sums, sensor dependency, past contextual information (e.g. Krishnan and Cook 2014).

Longer term: Use machine learning to predict the activities from the sensor sums, and use TDA to understand the errors in prediction.



TDA Application: IFTTT Rules

What is the structure of the possibility space underlying the way people use IFTTT?



Most Popular IFTTT Trigger and Action Channels

IFTTT Trigger Channel IFTTT Action Channel

Feed (RSS feeds) 25% Twitter 15%

Instagram 9% Email 11%

Date & Time 8% SMS 8%

Weather 8% Evernote 7%

Facebook 5% Facebook 7%

Gmail 4% Dropbox 6%

YouTube 2% Facebook Pages 4%

Twitter 2% Google Drive 4%

WordPress 2% Tumblr 3%

Tumblr 2% Pocket 3%

Total of top 10 trigger channels 68%

Total of top 10 action channels 74%



IFTTT Data: 120,253 Rules and 1101 Variables

“Turn on my lights when I get close to my home”(Title Words)

IOS Location(Trigger Channel)

You enter an area(Trigger Words)

Phillips Hue(Action Channel)

Turn on the lights(Action Words)

IF

THEN

86 binary variables

69 binary variables 103 binary variables

280 binary variables

563 binary variables

Data Description

120,253 IFTTT rules were created by 60,230 IFTTT users over a 3 year period from mid 2011 to mid 2014. (8404 unique rules)

68% of users created 1 rule22% of users created 2-3 rules8% of users created 4-9 rules2% of users created 10+ rules (23,496 total rules or 20% of data)

1101 binary variables for trigger channels, action channels, trigger words, action words & title words.

132 million data points



1. What IFTTT rules are people creating?

2. Are there interesting and important patterns that underlie the IFTTT rules that have been created?

3. Do these patterns suggest emergent themes?

Let’s take a look at some conventional data reduction and visualization approaches first…

Preliminary Research Questions



Network Graph of 553 Nodes for 120,253 IFTTT Rules Using All IFTTT Trigger Channels/Triggers and Action Channels/Actions

Network graph cannot reveal clear patterns in these complex data.

Only option would be to significantly reduce the number of nodes shown.



PCA and K-Means Clustering on the Component Scores

Some general features, but complexity is not revealed:

Component 1 identifies groups of rules about photos (+) versus email and SMS (-)

Component 2 separates weather and smart home rules (+) and new rules and feeds (-)



Multiple Correspondence Analysis of the Burt Matrix

Some gross features, but no complexity:

dimension 1 separates weather (+) and photos (1) while dimension 2 contrasts reading (+) and the smart home devices and Android devices (-).



Conventional dimension reduction and visualization approaches have difficulty revealing clear patterns in these complex data of 1100+ vars.

Need about 400 dimensions to account for most of the variance in the data. Took Stata 6 hours to compute the MCA solution.

General features that grossly separate rules are apparent, but it’s difficult to see the more subtle behavioral patterns underlying rule creation, let alone potential emergent themes.

Topological data analysis offers an approach to visualization of network structure that is more revealing and useful.

TDA Provides a Different Approach



Ayadsi* TDA Solution of All 120,253 IFTTT Rules

Each node is a cluster of rules.

Nodes connect if they have

rules in common.

Color indicates the number of

rules in each node:

2790 nodes containing

118,226 rules

(98.3% of all rules)

1305 nodes containing 2028 rules

(1.7% of all rules)

1

rule500+ rules

Metric: Hamming

Lenses: MDS 1 and MDS 2 (resolution 60, gain 1.6)

*We used the Ayasdi 3.0 software platform

(ayasdi.com, Ayasdi, Inc., Menlo Park, CA) to

perform the TDA on the IFTTT data



TDA Reveals the Possibility Space of IFTTT Rules

8

IF a new feed item

THEN send me an SMS or email notification

13,437 rules

Send me feed items

1

IF a new feed item

THEN tweet/post it or save it for later

22,227 rules

Share or save feed items

7

IF you like/tag content

THEN upload it to the cloud

4,235 rules

Store my likes

5

IF time/weather/SMS/email/location event

THEN control a device or log event

12,668 rules

Internet of Things/QS

4

IF time/weather/location event


14,265 rules

Send me event info

3

IF a new email or Craigslist post meets search conditions


7,128 rules

Send me content I’m looking for

2

IF new email, article, video

THEN save it for later

5,869 rules

Save content for later

8

IF news social media content


35,882 rules

Share or save social media

6



IFTTT Feed Rules:

Located in Two Distinct Flares in Full Network

IF a new feed item


13,437 rules

Send me feed items

1

IF a new feed item


22,227 rules

Share or save feed items

7

News Feeds are

used as triggers in

two very different

ways.

What happens if

we only analyze

the 29,990 IFTTT

rules that have

news feed as a

trigger?

Immediately consume news

Distribute or archive news



IFTTT Feed Rules:

Three Different Types in Subset Network

IF a new feed item matches my search


10,758 rules

Send me feed items I’m looking for

1

IF a new feed item

THEN share it on social media

9,530 rules

Share feed items with others

7b

IF a new feed item

THEN save it in Pocket, Buffer, Delicious etc.

6,997 rules

Save feed items for later

7a

Metric: Hamming

Lenses: MDS 1 and MDS 2 (resolution 30, gain 1.6)

Cluster 1 remains the

same (send me).

Cluster 7 splits into 7a

(save) and 7b (share). Immediately consume news

Distribute news

Archive news



In the full topological summary of 120,253 IFTTT rules, we can use

categorical variables as data lenses to separate the IFTTT rules into

groups with distinct structures.

› Social media channel lens - whether the IFTTT rule included a

social media channel (e.g. Facebook, Twitter, YouTube, etc.).

Does the structure of IFTTT rules vary based on whether the rule

includes a social media channel?

› Time lens - the year the IFTTT rule was created.

Does the structure of IFTTT rules change over time?

Using Data Lenses to Include Known Structure



Using a Data Lens to Separate Networks:

Social and Non-Social IFTTT Rules

54,300 Non-Social Rules

(45% of rules do not use a Social Media channel)65,953 Social Rules

(55% of rule use a Social Media channel)

Metric: Hamming

Lens: MDS 1 & 2 (resolution 60, gain 1.9)

Data Lens: Social Media (2 groups)



Non-Social Rules Social Rules

Send an SMS or email if

a new feed item

Save new feed

item to read later

Save new

feed item to

cloud

Save new photo

or file to cloud

Add event to cloud from a

location or SMS/email trigger

Save tagged

or favorited

content to

cloud

Smart Home and Wearables

Day, Time and Weather

Create a

note from

email

Send email to

notify me when

something

happens

Send me email or SMS if

Craigslist or Reddit post

Send me

email or

SMS if

tagged in

Facebook

Schedule

tweet/update by

day and time

Save Facebook

or Instagram

content to cloud

Save Instagram pic

to cloud, or send to

FB or Flickr

Send Instagram

content to Twitter or

blog

Send content from

one social platform

to another

Tweet, blog,

post new

feed item





Non-Social Rules Social Rules

Smart Home and

Automation

Collect and

Save

Notify Me

Notify Me

Social

Automation

Collect and

Save

Broadcast

and Share





Non-Social

IFTTT

Rules

Social

IFTTT

Rules

Year 12011-12

Year 22012-13

Year 32013-14

Metric: Hamming

Lens: MDS 1 & 2 (resolution 50, gain 2.1)

Data Lens: Social Media (2 groups), Year (3 groups)


Three Years of IFTTT Rules



Non-Social

IFTTT

Rules

Social

IFTTT

Rules

Year 12011-12

Year 22012-13

Year 32013-14

Collect & Save

Notify Me

Automation The basic structure of IFTTT emerged in

year 1, but became more organized and

interconnected in years 2 and 3.Broadcast & Share



TDA provides a way to visualize what emerges from the IoT using

interaction events as the unit of analysis.

IFTTT is an assemblage that emerges from the capacities of the

components (i.e. triggers and actions) exercised in the interaction over

time between apps and devices that are connected through the individual

rules.

Based on our assemblage theory framework (Hoffman and Novak 2015),

the topology represents the possibility space (DeLanda 2006, 2011)

underlying the potential capacities of the IFTTT assemblage.

Supports productive hypothesis generation and subsequent predictive

modeling.

Implications of TDA Analysis of IFTTT Rules



Exploring Emergent Consumer Experience: A Topological Data Analysis A

Tom Novak and Donna Hoffman

postsocial.gwu.edu



Acknowledgments

The Ayasdi 3.0 software platform for

topological data analysis (ayasdi.com)

was used to construct all networks of

the IFTTT data. The authors

acknowledge the support of Devi

Ramanan, Global Head – Product

Collaborations, Ayasdi Inc., Menlo

Park, CA

IFTTT data were collected from a crawl

during May - June 2014 and are used

with permission of IFTTT.com, San

Francisco, CA.

Smart apartment data are from the

Center for Advanced Studies in

Adaptive Systems (CASAS), School of

Electrical Engineering and Computer

Science, Washington State University.

http://ailab.wsu.edu/casas/


http://ailab.wsu.edu/casas/