The Automagic Box of Beauty
A Prototypical Smart Device as Use Case Example for User-centered Decision
Support via the Hub-of-all-Things
Helen Oliver
University of Cambridge, Computer Laboratory, William Gates Building, 15 JJ Thomson Avenue, Cambridge, U.K.
Keywords: Smart Devices, User-Centered Decision Support, Internet of Things, Generativity, Personal Data Stores,
Ubiquitous Computing, Sensor Networks, Smart Homes, Usage Context-based Design.
Abstract: In this position paper we present the Automagic Box of Beauty, a prototype smart cabinet which enables the
user to track their rate of consumption of toiletries. By transmitting product consumption data from the box
to the Hub-of-all-Things (HAT), a platform for personal data that is fully owned and controlled by the
individual, we open the potential for decision support for the user that hitherto has only been available at the
enterprise level. By returning ownership of personal data to the individual, the HAT enables horizontal
integration of information that has until now been held in vertical silos. We show how, by contextualizing
the product consumption data with data from a variety of other sensors and sources, the system will support
individual users in making decisions – in this use case example, decisions about replenishment and selection
of the products in the box.
1 INTRODUCTION
This article describes the Automagic Box of Beauty
(Beautybox), a prototype smart device which
demonstrates an example use case for user-centered
decision support within the Hub-of-all-Things
(HAT) (Ng, 2014).
The HAT is an open platform for personal data
stores which are fully owned and controlled by the
individual (Ng, 2014). Developers will find in the
HAT an attractive market with much demand for
applications to enable user-centered decision
support. The HAT will also provide a platform for
individuals to market their contextualized personal
information to other stakeholders (World Economic
Forum, 2013), giving companies unprecedented
opportunities to understand the contexts which
influence consumer decisions.
While there is an established need for contextual
intelligence in the Internet of Things (Vermesan et
al., 2014), the HAT will enable users to benefit from
unprecedented access to their own personal data and
unprecedented opportunities to do their own
contextual analysis of their behaviour patterns
according to their own ideas about what contextual
information is meaningful (Dourish, 2004).
The Beautybox is a smart cabinet which contains
everyday toiletries, tracking the products and their
rate of consumption, and uploading the data to the
HAT platform, the endpoint being a single database
instance for a single user. By measuring the
individual user’s rate of consumption of a product,
the system will be able to predict when that product
needs to be replenished and enable automation of
reordering without user intervention. An application
is in progress to provide this functionality, and will
also allow the user to specify what action to take if a
product is out of stock (list other products to
substitute, specify acceptable substitute products by
ingredient, and so on).
At this stage of the work in progress, the user is a
volunteer known as a Digital Person Zero (DP0)
who has consented to donate their personal data to
this research effort (Ng, 2014). The motivation in
creating the Beautybox is to make a start in
connecting the DP0’s home with HAT-enabled
devices – beginning with the bathroom, and adding
devices until the whole house is covered by a HAT
sensor network. By uploading timestamped data
from the DP0’s entire sensor network as well as their
choice of application data and social media data, it is
possible to contextualize and make sense of
behaviour patterns. For example, imagine that on
particular nights, certain lights are switched on and
91
Oliver H..
The Automagic Box of Beauty - A Prototypical Smart Device as Use Case Example for User-centered Decision Support via the Hub-of-all-Things.
DOI: 10.5220/0005330000910096
In Proceedings of the 4th International Conference on Sensor Networks (SENSORNETS-2015), pages 91-96
ISBN: 978-989-758-086-4
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
are not turned off. Context is provided by looking at
the other data from those nights and the mystery is
solved: it always happens after watching a horror
movie on Netflix.
At the time of writing, two Beautyboxes have
been successfully deployed to their respective DP0s
and work is in progress to develop a decision
support application using the Beautybox data. In this
article we will describe the Beautybox, and the
application we are in the process of developing for
the Beautybox data.
In this section, we have described what the HAT
project is, what the Beautybox is, what a DP0 is and
the current stage of deployment of the Beautybox,
and how we intend to use the data generated by the
Beautybox.
In section 2, we will explain the scope of this
position paper and what this article is and is not
about. In section 3, we will place the Beautybox in
its larger context as a smart device. In section 4, we
will describe the materials and methods used in
creating the Beautybox, walk through a single user
session with the Beautybox, and describe the
resulting state of the system. In section 5, we will
describe the work currently in progress to develop
an application to demonstrate the data analysis
potential of this use case. In section 6, we will
discuss the wider implications of this work. In
section 7 we will state our position about the impact
that we believe the HAT project can have.
2 SCOPE
In this section, we will explain the scope of this
position paper and what this article is and is not
about.
This article describes work in progress on the
Automagic Box of Beauty, a prototype smart device
to demonstrate a use case for the HAT platform.
The contribution of this article is to show a use
case of the Beautybox as a motivating example for
user-centered decision support, enabling individual
users to capture contextual intelligence (Vermesan et
al., 2014) (Dourish, 2004) about their own daily
activities, as part of an ecosystem of smart devices
which offer almost limitless potential for
generativity (Zittrain, 2008) for the individual as
well as for application developers, businesses and
other stakeholders.
The Beautybox itself is a prototype, with all the
limitations of a prototype, and we do not present it
as a design object in its own right. We also do not
claim that the Beautybox is innovative as a smart
device. Smart cabinets already exist on the market,
and more advanced inventory readers have been in
the realm of student projects within the last five
years.
The innovation is in the Beautybox’s role in
enabling user-centered decision support and its place
in the HAT ecosystem, which we will describe in
sections 5 and 6.
3 BACKGROUND
In this section, we will place the Beautybox in its
larger context as a smart device.
The current popularity of personal activity
tracking devices such as the Fitbit proves that
quantifying personal behaviour has mainstream
appeal.
However, personal data requires a device to
capture it (Ng, 2014); without the device, there is no
data. This is the Beautybox’s place in the HAT
ecosystem.
Fitbits aside, we must first ask: what would
motivate an individual user to add to their life yet
another conduit through which data may flow out to
destinations unknown? Why would someone ever
give data away to a company? The question is
rhetorical. Of course, in order to make any purchase
online we have to give at least some data to a
company – name, address, phone number, credit
card number – but that is just the data we know we
are giving away. Many things can also be inferred
from metadata, which we may not be conscious of
disclosing. As the Canadian Privacy Commission
informs the public, "every time you add information
about yourself, it’s like filling in a survey… all
without your knowledge, consent, or having ever
been asked…" (Canadian Privacy Commission,
quoted by Robin Hamman, at
http://www.cybersoc.com/2007/11/video-on-privac
.html#.VFtUD0u6q0s accessed 6 November 2014).
Companies aggregate data and metadata to make
inferences about us, but the insights gained are
revealed only to the company, not to the individual
who generated the data. Since the data is aggregated
and anonymized, it ceases to be personal enough to
be useful to the individual (other than the
sufficiently motivated stalker). The individual
receives benefits in the form of loyalty cards, but is
not allowed to know what the company knows. If
the data were truly owned by the individual, the
individual would be issuing loyalty cards to the
company, not the other way round.
SENSORNETS2015-4thInternationalConferenceonSensorNetworks
92
Whatever its users’ varying opinions about
privacy, the Fitbit is so popular because – to state the
seemingly obvious – the user benefits from being
able to view and analyze the data that comes from it,
and from the ability to combine it with data from
other applications (Ng, 2014).
How does the Beautybox compare to the Fitbit in
this regard? The difference is that the data collected
by the Beautybox is sent not to a corporate endpoint,
but to the individual’s HAT. Unlike the pseudo-
anonymized data that companies hold, all the data in
the individual user’s HAT will be personally
identifiable, with the user having complete and fine-
grained control of which personal data they disclose
and to whom they disclose it. To ensure trust, device
manufacturers and companies using HAT data will
require certification of their scrupulosity in
maintaining the privacy of the data entrusted to them
by HAT users.
4 MATERIALS AND METHODS
In this section, we will describe the materials and
methods used in creating the Beautybox, walk
through a single user session with the Beautybox,
and describe the resulting state of the system.
Figure 1: The Beautybox with open lid, showing the
barcode scanner mounted on the front, and the assembly
compartment on the right.
Figure 1 shows what the user sees when they
open the Beautybox. The Beautybox is housed in a
Perspex box assembly 39cm long by 28.8cm high by
20cm wide. An interior panel divides off a 5cm wide
compartment at one end of the box, which houses
the microcontroller and wiring assembly. On the end
of the box is a removable panel held on with Velcro
tape finished with strips of ribbon, enabling access
to the assembly compartment.
Figure 2: The barcode scanner and weighing scale
mounted inside the Beautybox, with the weighing platter
removed and the cables threaded through apertures in the
instrument panel.
The Beautybox contains two sensors, which can
be seen in Figure 2: a weighing scale on the bottom
of the box (Salter 1036 “Disc” electronic kitchen
scale) and a barcode scanner (Adafruit) mounted
near the top edge of the front of the box. The
weighing scale was adapted by the addition of a
specially designed circuit board to send the weight
sensor data to the Arduino Mega microcontroller.
To hide the scale and extend its weighing
surface, a 33.5cm x 19cm piece of 5mm black
Perspex was affixed on top with Velcro strips; this
Perspex weighing platter became the floor of the
box. The cables for both sensors are threaded
through apertures to the other side of the instrument
panel.
Figure 3: Beautybox with transparent end panel, showing
the assembly compartment.
TheAutomagicBoxofBeauty-APrototypicalSmartDeviceasUseCaseExampleforUser-centeredDecisionSupportvia
theHub-of-all-Things
93
The assembly compartment is shown in Figure 3.
The Arduino Mega is mounted with Velcro and has
an Adafruit WiFi Shield mounted on top of it to
provide network connectivity and push the sensor
data to the HAT endpoint. The barcode scanner’s
PS2 cable, as well as the load cell connectors, are
wired to the Arduino, as are the LED light that
displays system status to the user (Figure 4). The
power is supplied by 2 x 3.5V lithium batteries
(Figure 3) and the power supply operated by a lever
switch (Figure 4).
Figure 4: LED light displaying system status to the user.
4.1 Walkthrough: using the Beautybox
In this subsection, we will walk through a single
user session with the Beautybox.
To activate the system, the user turns the power
switch to ON. The LED light turns red as the system
initializes.
The LED light turns blue as the WiFi Shield
connects to the user’s wireless network.
The LED light turns red as the system initializes
the weighing scale and records the starting weight of
the box contents.
The LED light turns green when the barcode
scanner becomes available.
The user takes an item out of the box and scans
the barcode. The LED light turns red while the
system polls for a change in weight readings. Weight
is calculated by a moving average over 250 readings
taken at 100-millisecond intervals.
If the system times out before a weight change is
detected, the LED turns green and the barcode
scanner becomes available again. Otherwise, if a
weight change is detected, the LED turns blue and
the system uses the HAT REST API to send the
barcode and item weight to the HAT endpoint. Item
weight is the absolute difference between the latest
and previous weights of the box contents.
When the database transaction is complete, the
LED turns green to indicate that the barcode scanner
is available.
When the session has finished, the user manually
powers off the system.
4.1.1 Resulting State of the System
In this subsection, we will describe the resulting
state of the system.
The DP0’s HAT now contains two new items of
sensor data: the timestamped barcode identifying the
product used, and the timestamped item weight.
The simplest analysis that the user can do with
this sensor data is to track the rate of consumption of
each product. As the product – for example shampoo
– is taken out, used, and put back into the
Beautybox, the item weight will go down, and then
suddenly go up when the product is replaced.
5 WORK IN PROGRESS
In this section, we will describe the work currently
in progress to develop an application to demonstrate
the data analysis potential of this use case.
We are currently developing a simple application
to demonstrate a use case of analysis of the
Beautybox data. Using the combined barcodes,
timestamps, and item weights, the average daily
usage of a product is simple to calculate, and the
system can predict the amount of time before a
product will need to be replaced.
With this information, we will implement
automated reordering functionality – like a
subscription service, but more efficient and less
expensive because the reorder will be placed just-in-
time rather than at regular intervals.
Not everyone wishes to use the same shampoo
every time, and products sometimes become
unavailable. In this situation, a use case for decision
support becomes apparent. Using the barcode, it is
possible to search publicly available APIs such as
the Tesco API and find product information; the
search criteria depend on the user’s reasons for
selecting that particular shampoo in preference to
others.
Supposing a user suffers from scalp problems
and must use a very specific treatment shampoo.
The shampoo is expensive, difficult to find, and
subject to withdrawal by the manufacturer at
unpredictable intervals. The user might know of a
list of acceptable alternative products and specify the
product names. But what if the user needs help to
identify an alternative product? In that case, the
system can search for a shampoo that contains the
required active ingredients (coal tar, salicylic acid)
or without the undesired ingredients (proteins,
silicones).
Not every user has such specific preferences.
SENSORNETS2015-4thInternationalConferenceonSensorNetworks
94
With the finished application, users will be able to
specify how important it is for them to replace the
Beautybox items with the exact product each time.
Perhaps a user will accept any shampoo as a
substitute, or wants to receive a random shampoo
(possibly within a certain price range) each time.
Chen et al. (Chen et al., 2013) define the usage
context of a product as “all aspects describing the
context of product use that vary under different use
conditions and affect product performance and/or
customer preferences for the product attributes”.
Taking the application a step further, we will work
with each DP0 to contextualize the Beautybox data
according to their respective preferences and needs
(Dourish, 2004). For example, imagine the box
contains a hair-straightening product and the user
runs out of it before the system has a chance to
reorder it. Why has this happened? On examination,
the straightening product shows a very low daily
usage rate until there is a sudden spike in
consumption. Imagine this user has also been
recording weather data in the HAT, making it
possible to combine it with the Beautybox data.
Looking at the combined weather and Beautybox
sensor data, it turns out that the straightening
product was only used on rainy days with a chance
of thunderstorms, but otherwise never touched. We
could therefore add functionality to the application
by contextualizing product usage with weather,
anticipating increased usage of the straightening
product for the week ahead if thunderstorms are
predicted.
6 DISCUSSION
In this section, we will discuss the wider
implications of this work.
Because the variety of user preferences for
contextualization is potentially limitless, the average
user is unlikely to have the resources to perform all
their desired analysis on their own, and so the role of
application developers within the HAT ecosystem
becomes apparent.
This brings us to the topic of generativity.
Zittrain (Zittrain, 2008) defines generativity as “a
system’s capacity to produce unanticipated change
through unfiltered contributions from broad and
varied audiences.”
Zittrain identified five features of
a solution conforming to the generative pattern:
leverage (making a difficult job easier), adaptability
(alterable for a variety of purposes), ease of mastery
(requiring minimal training to use and extend),
accessibility (ease of obtaining and developing a
working system), and transferability (ease of
distributing updates).
The HAT platform, even at the prototype stage,
has nearly all of Zittrain’s features of a generative
solution. It has the leverage of its “human-think-
alike” (Ng, 2014) database model, as well as the
adaptability of the REST API which provides
interoperability to any HAT-ready device and
consequent combination of device data for decision
support. The database model and REST API
combine to provide ease of mastery by making it
simple to make a device HAT-ready and simple to
combine data downloaded from the HAT; the
platform’s openness provides the requisite
accessibility. Only transferability remains
unaddressed at the prototype stage, in that changes
affecting the Beautybox must be manually uploaded
to the microprocessor. Improving transferability will
be addressed in future work.
Businesses and product designers also stand to
gain unprecedented insights from the purchase of
data that has been contextualized and horizontally
integrated according to preferences defined by the
users themselves. As Chen et al. (Chen et al., 2013)
explain, “The usage context may also have a
significant impact on the product performance,
which is not considered in existing methods that
simply treat product performance as ‘constant’
across all customers and usage contexts in choice
modeling.”
7 CONCLUSIONS
In this position paper, we have described the
Automagic Box of Beauty, a prototype smart device
to demonstrate an example use case for user-
centered decision support within the Hub-of-all-
Things, a personal data store for the Internet of
Things that places ownership of personal data in the
hands of the individual (Ng, 2014). In section 1, we
explained what the HAT project is, what the
Beautybox is, what a DP0 is and the current stage of
deployment of the Beautybox, and how we intend to
use the data generated by the Beautybox. In section
2, we explained the scope of this position paper and
what this article is and is not about. In section 3, we
placed the Beautybox in its larger context as a smart
device. In section 4, we described the materials and
methods used in creating the Beautybox, walked
through a single user session with the Beautybox,
and described the resulting state of the system. In
section 5, we described the work currently in
progress to develop an application to demonstrate
TheAutomagicBoxofBeauty-APrototypicalSmartDeviceasUseCaseExampleforUser-centeredDecisionSupportvia
theHub-of-all-Things
95
the data analysis potential of this use case. In section
6, we discussed the wider implications of this work.
In this section, we will state our position about the
impact that we believe the HAT project can have.
The use case described in this article is just one
example out of a potentially limitless set of
scenarios in which the HAT can facilitate decision
support for a vast variety of individual user
preferences.
By returning personal data to the control of the
individual, and contextualizing raw data that until
now has been held in vertical silos (Ng, 2014), users,
with the help of application developers providing for
their needs, will have an opportunity for the kind of
decision support that was previously only available
at the enterprise level.
The individual user owns all their own data, can
see all their own data, is uniquely capable of
interpreting what the data tells them about their own
behaviour patterns, and will stimulate a market for
application developers to provide decision support
tools at the individual user level and for businesses
to gain unprecedented insight into usage context.
ACKNOWLEDGEMENTS
The custom circuit board for the weighing scale and
the design of the Perspex box structure, as well as
copious and invaluable expert advice, were provided
by Brian Jones.
The work, and in particular the wiring of the
small components, could not have been completed
without the assistance and advice of Andrius
Aučinas and Carlos Molina Jiménez.
REFERENCES
Chen, W., Hoyle, C., Wassenaar, H. J., 2013. A choice
modeling approach for usage context-based design. In
Decision-Based Design, Springer. London, pp.255-
285.
Dourish, P., 2004. What we talk about when we talk about
context. Personal and Ubiquitous Computing 8(1),
pp.19-30.
Ng, I., 2014. Engineering a market for personal data: the
Hub-of-all-Things. RCUK Digital Economy HAT
Project. Available online at http://hubofallthings.com/
hat-briefing-paper/ Accessed 6 November 2014.
Vermesan, O., Friess, P., Guillemin, P., Sundmaeker, H.,
Eisenhauer, M., Moessner, K., Arndt, M., Spirito, M.,
Medagliani, P., Giaffreda, R., Gusmeroli, S., Ladid,
L., Serrano, M., Hauswirth, M., Baldini, G., 2014.
Internet of Things strategic research and innovation
agenda. In Internet of Things – from research and
innovation to market deployment, River Publishers.
Aalborg, pp.7-142. Vermesan, O., Friess, P. (Eds.).
World Economic Forum, 2013. Unlocking the value of
personal data: from collection to usage.
Zittrain, J., 2008. The future of the Internet – and how to
stop it, Yale University Press. New Haven.
SENSORNETS2015-4thInternationalConferenceonSensorNetworks
96
