Tuesday, May 22, 2018

The Smart Pop-Up Wine Shop demo

This Wednesday and Thursday I’m very excited to be running a demo of my final PhD project, something I’ve spent the past few months designing, building and testing. The project is a Smart Pop-Up Wine Shop, a shop that learns from its customers’ behaviour and recommends wines based on their popularity.

Smart Pop-Up Wine Shop Invite by gonzillaaa

The focus of my research is the design of smart products and services and how these may change people’s behaviour. As computers become ever so small and disappear inside walls, tables and many everyday objects the way we interact with information changes radically. Like footprints in the snow our encounters with smart objects and spaces leave a permanent mark in the information substrate. From these traces smart artefacts can learn how, when and where they are used, and can react to that information.

Using collaborative filtering, a branch of artificial intelligence frequently used by online retailers, the Smart Pop-Up Wine Shop recommends the most popular wines as well as favoured purchase combinations by learning from its customers. Lights integrated in the shop’s furniture indicate the popularity of each wine using a colour scale. Additionally visitors can scan each bottle to see what others bought in combination with it, creating a real world version of the famous “users who bought this also bought” system popularised by Amazon.

Po-Up Wine Shop

After a 5 week trial with over 100 participants conducted in February 2011 at the Pervasive Interaction Labs at The Open University, the Smart Pop-Up Wine Shop will be installed for a demonstration in central London, at Arup’s Phase2 Gallery on number 8 Fitzroy Street.

I’ll be personally attending the shop from 10 am till 7pm. If you’re interested in Interaction Design, Ubicomp technologies and services, Recommender systems please come by, would love to have a chat.


My Carbon and Energy Hack Weekend Project: Realtime Ambient Energy Display for nº10 Downing Street

The weekend before last was Rewired State’s Carbon and Energy Hack Weekend I previously mentioned the event and my plans for it. I did not get the chance to spend both days that weekend working on my project, I decided to retake the project this weekend and put the finishing touches to it.

My aim was to develop a prototype for an ambient information display by connecting it to a “real-time” energy data feed. I also wanted to explore the different ways in which we can present the data on the device. The ambient display is initially intended for personal or home use, my fist task for the event was to find a suitable data feed.

Screen shot 2010-11-08 at 10.35.31.png

The guys at Carbon Culture provided me with an interesting resource. They supply, amongst other feeds, a real-time energy consumption data stream coming from nº10 Downing Street at 30 minute intervals. I realise nº10 surpasses a typical household both in terms of size and energy consumption, I assume however The Camerons, like anyone else, can benefit from an increased awareness of their energy consumption behaviour.

I’m using the device to display the live feed in three ways:

A trend graph that shows the last 8 hours of data.

A parallel graph that compares two readings, for instance now vs. yesterday at the same time, or now vs. an hour ago. The comparison graph is coloured green or red depending on the direction of change in the data.

Finally, an arrow that represents the direction of the fluctuation in the data feed. In other words if energy consumption increases in comparison to an hour before the arrow points upwards in red, on the other hand if consumption decreases it points downwards in green.

In my opinion both the comparison graph and the arrow showing the fluctuation in data work much better than the trend graph. Both representations are much more concerned with the “now” and they suit the realtime nature of the data. The trend graph requires context to be understood, if we only take a day worth of data to provide that context a single spike in the distribution changes the shape of the graph dramatically from one hour to the next, which can be confusing. Counting only with 8 columns to provide context means we have a reduced resolution to give insights on the data. A potential solution would be to use a larger data set as context, a month let’s say, to plot the proportion of a daily distribution.

A feature of ambient displays, and this display in particular, is that it does not show values of data but maps fluctuation in the proportions of that data to colour and shapes. As a result relative representations like the comparison graph and arrow are more immediate and work much better.

The incoming data feed for Downing Street in kW h for the last 8 hours looks something like this:

[59.833999999973457, 62.183333333348862, 60.283333333325572, 
59.197435897454852, 59.26666666669189, 64.550000000017462, 63.999122807028463, 64.388888888875954]

Since we have a very low resolution 8×8 display the data is re-scaled to fit a 0-8 range.

[1.8324886226952222, 4.9049101796373273, 2.4201197604444991, 1.0, 1.0905389221639117, 8.0, 
7.2795713835399347, 7.7893013971657767]

Finally since we cannot display 1.83 pixels the data is rounded to its nearest integer in order to be displayed.

[2, 5, 2, 1, 1, 8, 7, 8]

Here is a clip of that sequence displayed:

Place-stat displaying live energy data feed from nº10 Downing Street from Gonzillaaa on Vimeo.

A script running on my laptop pulls data from Downing street at regular time intervals, it processes and shapes the data to be displayed, and sends it wirelessly to the device via Bluetooth. For those interested in seeing how this works have a look at the code.

Why use an ambient display?

  • As I mentioned before, none of us have a mental model of what 1kW h looks like, we might understand the formal abstraction of what it represents but there isn’t a single image of how it looks like.

    Those who work with energy and carbon data know energy in watt hours is the multiplication of power in watts and time in hours. So 1kW h is a unit of energy equal to 1000 watt hours. But that does not mean much to most people.

  • Since we don’t have a mental model to clearly describe our energy consumption we think of energy in relation to other things. We consume more or less than yesterday, object A consumes more that object B. If you look at many of the devices that fall in the category of smart meters with exceptions like Wattson and the Energy Joule they mostly rely on the presentation of data in kW h on alphanumeric displays. I believe there is ample room for exploration.
  • Finally our goal is to improve our energy consumption behaviour. Behaviour change technologies are frequently informed by the work of BJFogg and his team at the Stanford Persuasive Technology Lab.

    According to Fogg’s Behaviour Model a crucial aspect to enabling change in behaviour is to create “Triggers” and place them in the path of motivated people. Ambient computing enables us to embed information in the places we inhabit, making it “glanceable”, easily accessible and provide people with the least path of resistance to action.

What’s next?

I think there is a lot of room for the development of a flexible and low cost display system that enables us to access information in a spatially relevant way. As sensing energy consumption from appliances and other devices becomes common place, the tightly coupled sensing and displaying systems that are commonplace today will give way to other possibilities.

I am keen to test my assumptions with this work and develop the project further. I want to develop the hardware and software as an easy to use self assembly kit people can buy and use in their homes, offices, workplaces etc. Interested? Want to help? Get in touch.

Thanks to the Rewired State Carbon and Energy weekend organising team and the guys at Carbon Culture for their in-situ API support.


Work plan for Carbon and Energy Hack Weekend

This weekend is Rewired State: Carbon and Energy Hack Weekend an event organised by 10:10, The Guardian and Rewired State with the aim of putting people together under one roof to develop applications around Energy and Carbon. The event is formatted so people can either arrive with a project to develop, or can lend a had to others, here is my proposed project.

People install energy monitors in their homes to become more aware of their consumption, and to improve based on the information the meter provides. To lower energy consumption the key information people want from their meter is whether they are doing better or worse, compared to yesterday, their neighbours, or a national average, for instance. The majority of devices in the smart meter space tend to provide energy consumption data in kW h, sometimes mixed with a trends graph. There is a mismatch between users’ understanding of their energy usage, which is informal and relational, kettle vs. toaster or today vs. yesterday, and the continuous numeric data stream provided and presented by meters. I believe end users can benefit form a different approach, an ambient display which presents information as relationships, to encourage behaviour change.

I have an early prototype for an ambient display that presents fluctuating sources of data such as energy, and it can show trends as well as comparisons, see a picture below, also here, here, and you can read more about the project here. I would like to take the opportunity of this weekend to take that prototype a little further, working with others who can provide input into the project. I have a functional prototype of the device, I want to create a webservice that provides data streams, and need to design a serial protocol that feeds the data to the device.


What I’m I hoping from the experience of working with others:

  • Data: I would like to get individual household energy consumption data, hopefully something along the lines of hourly summaries. I don’t need to pinpoint individual addresses, so data can be anonymised. I want to prototype a realistic use scenario for the device use in a home environment. I want to display both trends of use for a day and comparisons between data. For instance the same space at different points in time or comparisons with other spaces of similar characteristics (m^2) & location.
  • I would like to build a prototype for a web service in which users can select what data to display. I’m looking to build it in django sprinkled with jquery. I have worked with django before (a while ago) I would really like someone’s input in designing the models and thinking about the structure of the service. Help with the implementation is obviously also welcome, particularly with the jquery side.
  • I want to design a lightweight protocol to send data from the service to the device. This can be a very simple ascii protocol if we run out of time, or something a bit more intelligent. I have already written an arduino library that takes care of the plotting, the protocol only needs to get the data and plot it. I would love to talk to someone with c++ experience in particular for embedded devices/arduino as a sanity check, to get input into that side of the project. Serial communications currently happen via USB, but I can very quickly move to either Xbee or Bluetooth.

I am very interested in discussing additional aspects of the project with others. In particular, energy experts, statisticians, social scientist, product designers, marketers, potential partners, investors. I am also really looking forward to see what others are doing, I’m happy to provide input where I can. I have been involved in a number of projects visualising energy data in the past.

Do you want to make the Internet of things a reality? join me! ping me on twitter @Gonzillaaa


John Tukey, the Bit and Box plot


While working with John von Neumann on early computer designs, John Tukey introduced the word “bit” as a contraction of “binary digit”. The term “bit” was first used in an article by Claude Shannon in 1948.

The term “software”, which Paul Niquette claims he coined in 1953, was first used in print by Tukey in a 1958 article in American Mathematical Monthly, and thus some attribute the term to him;

Tukey also introduced the Box plot in his 1977 book,“Exploratory Data Analysis”.



The Cognitive Bias Video Song

Here is an excerpt of the brilliant “Cognitive Bias Son”

I’m Biased because I knew it all along.

Hindsight Bias! I knew it all along!

I’m Bias because I put you in a category in which you may or may not belong.

Representativeness Bias: don’t stereotype!

I’m bias because of a small detail that throws off the big picture of the thing.

Anchoring Bias: See the forest for the trees!

I’m biased toward the first example that comes to mind.

Availability Bias: the first thing that comes to mind!

I’m Biased because I’ll only listen to what I agree with.

Confirmation Bias: you’re narrow-minded if you are like this!

I’m Biased because I take credit for success, but not blame for failure.

Self-Serving Bias: my success and your failure!

I’m Biased when I remember things the way I would have expected them to be.

Expectancy Bias: false memories are shaped by these!

I’m biased because I think my opinion now was my opinion then.

Self-consistency Bias: but you felt different way back when!

via Nudge


Place-stat* Ambient signage system at Pervasive2010

Yesterday I spent the day presenting a demo of a flexible ambient signage system I’ve been working on called Place-stat* at Pervasive 2010. I had an amazing response from everyone at the conference, the feedback to the idea and its execution was beyond my expectations. I received many requests from people to buy it commercially, use it on their own research, photographs and a place to follow up on the project. I’m jotting down some notes on the initial rationale for the design for those who did not attend the demo, and the future development of the project. This also sets the direction for something that will occupy a good part of my time in the months to come.

Slide used for demo presentation at Pervasive2010

What is an Ambient signage system? Ambient computing sits at the centre of an area of research called ubiquitous or pervasive computing. Ambient computing proposes devices that can present information in a way that sits at the periphery of our attention. This sits in contrast with the current desktop, laptop and mobile devices that we use to access information which require dedicated attention. Ambient devices are embedded in our surroundings and we can choose when and how to engage with them by simply glancing at them. There are two projects that clearly exemplify Ambient information systems: Live wire by Natalie Jeremijenko a real-time local network traffic indicator in the form of a wire hanging form the ceiling – Live wire wiggles proportionally to the amount of traffic on the net; And Ambient Orb a glass lamp that uses colour to show weather forecasts, trends in the market, or the traffic designed and sold by Ambient devices a spin off setup by David Rose and others from the MIT Media Lab. Why is this relevant? How does it matter? As we interact with an ever increasing mass of information exploring ways in which we can better access it becomes more relevant.

One property that traditionally characterises ambient information displays is the one to one relationship that exists between the display and the information presented. In other words, often a single type of information i.e a stock value is mapped into a single display element; for instance a colour, motion or sound range. As a result of this close coupling between display and data, ambient displays seldom have the flexibility to communicate effectively more than one set of information, neither do they have the expressiveness to show more than one perspective on a set of information. The proposition of Place-stat* is quite simple, we combine a colour display that allows us to present both abstract and figurative representation of information with an exchangeable physical outer shell that contextualises the information presented.

Images of demo display by pseudonomad

The applications are countless: energy awareness, environmental monitoring, realtime analytics, notifications of data coming from the cloud, email, social network activity, trends from log data… Wouldn’t you like to have a flexible display you can connect to sources of information and notify you unobtrusively for your personal use? Is it really necessary to check email/facebook/twitter as many times as we sometimes do? Wouldn’t it be better to concentrate on what we’re doing and just lift our eyes to see if there’s anything new and continue working? or enjoying life. For groups it can be used as a team collaboration and notification tool, as a group activity or target alert system. It can be used at home, in workplaces, public spaces…


We are starting trials exploring its use as an energy display at Arup. In the coming weeks, we will finalise development work and test the device in real world scenarios. We will then incorporate the feedback gathered into the next generation where a larger number of devices will be manufactured. From then on we will study the viability of its development as a commercial venture. If you are interested in the developments of the project you can of course follow us on twitter.

Garcia-Perate G, Conroy Dalton R, Dalton NS, Wilson D. Place-stat* Ambient signage system. In: Pervasive Computing Adjunct Proceedings 8th international Conference. Helsinki, Finland: 2010. (pdf)

You can find more info over at emtech primer and images at Flickr here and here


The internet real state, server space comparison


Great treemap infographic illustrating the amount of servers owned by the network’s biggest players.

via intac


@timoreilly talks about data services ecosystem, real-time data and government data


The History of Location Technology


via mashable


Augmented reality, aligning the past with the present.


High Street, Pinner, Middlesex; archive shot, 50s
Photograph: Adam Leach/leachy.com


University of Wisconsin Green Bay campus; archive shot, early 70s
Photograph: Todd Sanders

A great image set on the guardian that illustrates very nicelly the essence of augmented reality, overlapping the past to the present.

see. read. write. do.

A research blog about interaction, design research, urban informatics, ambient computing, visualisation, emerging technologes and their impact on the built environment.

About me

This is a blog by Gonzalo Garcia-Perate a PhD researcher at The Bartlett, looking at adaptive ambient information in urban spaces.

Get in touch