Participatory Data Analysis: A New Method for Investigating Human Energy Practices Gerd Kortuem, Jacky Bourgeois, Janet van der Linden, Blaine Price The Open University Milton Keynes, UK gerd.kortuem@open.ac.uk, jacky.bourgeois@open.ac.uk, j.vanderlinden@open.ac.uk, blaine.price@open.ac.uk Abstract — This paper presents a novel data-driven method to and energy companies were dominated by energy and money, investigate the interdependence between technology design and transactions in the smart grid are dominated by information human energy practices. The method – called Participatory Data exchanges including - among others - real-time consumption – makes use of fine-grained energy data collected via smart and generation data, price signals and demand load shedding meters and smart plugs, and behaviour visualisation during home visits to spark self-reflection among householders. signals. As the relationship between energy and people is becoming I. INTRODUCTION more complex a new challenge is emerging for designers of The relationship between people and energy is changing. energy systems, HCI researchers and social scientists who are Many years ago, when energy prices were low, people were interested in understanding the interdependence between content to act as passive consumers of energy with only a faint technology design and human energy practices: what methods understanding of the relationship between their behavior and do we use to investigate behaviour change, and changes in the money they had to spend for energy monthly or half-yearly. attitudes and self-image? Observing and understanding Today, this situation has changed dramatically for three behaviour change in a real world context, such a home or a reasons: 1) energy prices have skyrocketed which has forced large organisation, is difficult. The home is a highly contextual consumers to pay attention to costs of energy 2) increasing environment steered by everyday life, habits and implicit rules. awareness of climate change has led people to question the To understand how people accept new concepts and impact their action have on the environment; 3) alternative technologies in their domestic life, we need to get people sustainable energy technologies allow people to generate their thinking and talking about it. There is a birth of established own energy at home using for example, solar PV or ground methods and methodologies (e.g. ethnography, technology heat pumps. Together these changes are transforming people’s probes [2]), all of which are useful in different ways. attitudes towards energy and lead to a widespread change of The increasing amount of fine-grained data about energy domestic energy practices. consumption and generation data – collected via smart meters Digital technology is playing a key role in mediating the and smart plugs – makes it now possible to add data analytics relationship between people and energy. On the one hand, and data visualisation methods to the methodological tool people increasingly use comparison websites such as uSwitch chest. However, the difficulty is how to combine more and GoCompare to seek out the cheapest energy tariffs and traditional human focused methods with new data-driven switch suppliers (although most people tend not to switch methods. In this short paper we highlight a novel method suppliers often). On the other hand, plenty of studies have which we call participatory data analysis. The key novelty of shown that energy display and similar energy feedback this method is the use of energy behaviour visualisations during technologies can facilitate behaviour change with the goal of home visits to spark self-reflection among householders. reducing energy consumption (some of the people, some of the II. PARTICIPATORY DATA ANALYSIS time) [1]. The smart grid represents the next wave of the digital Participatory data analysis is a method to understand human technology revolution in the energy sector and is likely to have energy practices by enabling people to reflect on their own a transformative impact on the relationship between people and behaviour. These reflections in turn provide insights into energy. The smart grid enables two-way communication factors that influence people’s behaviour such as attitudes, self- between generators, consumers and those that do both, and image, and motivations as well as social conventions and turns the (soon to be) “smart home” into an intelligent end- norms. To understand the motivation and purpose of this new point of the electricity grid, thereby paving the way for method we will describe participatory data analysis in the widespread adoption of innovative schemes such as dynamic context in which we first developed and applied it. demand response and peer-to-peer energy. The changing relationship between people and the energy system is depicted in Figures 1 and 2. While formerly transactions between people they were generating and wanted to consume as much of it as possible. We learned that they manually shift some of their Energy loads, like the washing machine or the dishwasher by “chasing the sunshine”, that is, looking out of the window and switching Energy Energy bill Energy on when it is sunny. System Consumer The aim of our study was to investigate more precisely how Money household members were carrying out this process of manually shifting their appliances. What were their struggles and constraints when aiming to maximize their self-consumption? Figure 1. Old-style Relationship between People and How good were they at manually doing this, and what scope Energy System was there for further improvement? B. Energy and Behavior Data Energy Over the course of six months we collected fine-grained Consumption + generation data data about appliance use and energy consumption. Each household was equipped with three smart meters to measure: Price signals (i) imported electricity from the grid (the typical fiscal meter), (ii) generated electricity from the solar panels and (iii) the Demand shedding signals exported electricity to the grid. In addition smart plugs were Smart Energy deployed to monitor the electricity consumption of individual Grid Incentives Consumer appliances every five seconds. Energy and System Personalised recommendations Generator C. Visualisations To inform the design of technological interventions we Appliance usage data conducted interviews with each household with the aim to let residents reflect on their own laundry routines and their EV usage data relations to local energy generation. These interviews were Money conducted in-home lasting between 25 and 50 minutes and at a time suitable for the participants. Energy In order to enable this process we developed customized visualizations of people’s personal electricity data. For each participant, we printed out a set of three visualizations for the Figure 2. New-style Relationship between People and most relevant summer month on A3 size paper (one of which is Smart-Grid Energy System sown in Figure 2). The visualizations were developed to give participants an overview of their washing machine loads over a A. Motivation and Purpose month. Each washing machine load was indicated as a distinct event in the week and month, and was represented as a multi- We developed and applied participatory data analysis colored dot. The y-axis indicates which time the wash was during a recent study [3,4] that explored the potential of started and the lower x-axis indicates the day of the week and interactive electricity demand-shifting – a particular form of date for the wash. The actual weather for each day – important technology-mediated behavior change where electricity for estimating energy generation from solar PV - was displayed consumption is shifted towards times that are optimal in terms at the top x-axis in the form of a “sunshine” or “cloud” symbol of cost or CO2 emissions. Specifically we focused on etc. households with residential solar electricity generation and As an example, the first visualization uses a pie chart model explored how these households can maximise self-consumption for each load, showing for each load how much electricity was of locally generated “green” energy. To limit the scope of the coming from the solar PV (lightly shaded part of circle = green study we honed in on energy self-consumption for doing the in the original printed version) and how much electricity was laundry and using washing machines and dryers. coming from the grid (dark shaded part of circle = red in the We conducted a participatory user study with 18 original printed version). The bottom of the pie chart represents households, over a period of 6 months. During this period the actual start of the load. For example, during the day on the residents carried out their normal laundry routines and we were far left this household did 3 lots of washing, one before 8 in the able to track their electricity data through a variety of meters morning (using mostly grid electricity), one around noon and smart plugs. The user study was situated within a wider (mostly electricity coming from solar energy) and one at 4 in program of research, involving some 75 households the afternoon (again with mostly grid electricity). The investigating issues around household electricity usage. The 18 participants were all very familiar with the concept of selected households had all invested in solar electricity. From importing and exporting electricity and this visualization was earlier focus groups and in-home visits we had become aware designed to draw their attention to potential opportunities to that participants had a keen interest in the amount of electricity increase their self-consumption. We deliberately gave this the chart the questions were phrased in terms of “Would it have been possible to...”. III. DISCUSSION Participatory data analysis (PDA) uses fine-grained, longitudinal data from smart meters and smart plugs to create high-level visualisations of household behaviours. By using visualisations during interviews we enabled participants to reflect on their own behaviour and ground discussions. In our experience participatory data analysis as a method has the following advantage: • PDA visualisations create a common ground for discussions between experts (researchers) and non- experts (households). • PDA grounds discussions by providing an accurate ! representation of (past) behaviours. This avoids discussions drifting off into unrealistic hypotheticals. • PDA can be used to inform the design of novel technology interventions and does not require development and deployment of prototypes. (In our case we used PDA to inform the design of a recommendation system to help people optimise self- consumption [4]). • PDA makes it possible to collaboratively explore the possible impact of technology interventions. For example, participants can be asked to create their own visualisations to represent behaviours after deployment of technology interventions. So far we have used PDA only in the context of domestic energy self-consumption. We believe that PDA has similar potential to explore behaviour and social aspects in fully smart- grid connected homes, for example for investigating behaviour ! responses to dynamic demand response approaches [5,6]. Figure 3. Behaviour Visualisations for REFERENCES Participatory Data Analysis [1] Fischer, C. (2008) Feedback on Household Electricity Consumption: A Tool for Saving Energy? Energy Efficiency 1, title 'Waste' – to be provocative (even though there is no actual no. 1 (May 6, 2008): 79-104. waste) and to make the point that participants could have [2] Bourgeois, Jacky; van der Linden, Janet; Price, Blaine and consumed more electricity coming from solar energy and thus Kortuem, Gerd (2013). Technology probes: experiences with reduced their electricity import from the grid. The objective home energy feedback. In: Methods for Studying Technology in would be to have a full green circle, which means that the the Home, 27 Apr 2013, Paris. washing machine load had been entirely powered by the solar [3] Bourgeois, Jacky; van der Linden, Janet; Kortuem, Gerd; Price, PV. This visualization thus gave a quick overview of the Blaine A.; and Rimmer, Christopher (2014). Using Participatory “green-ness” of the household's loads over the month and Data Analysis to Understand Social Constraints and helped open the discussion. Opportunities of Electricity Demand-Shifting. ICT4S 2014. The second visualization was designed to show participants [4] Bourgeois, Jacky; van der Linden, Janet; Kortuem, Gerd; Price, when would have been the “greenest” time to start the washing Blaine A.; Rimmer, Christopher (2014). Conversations with my machine and how much delay it would have implied. The Washing Machine: An in-the-wild Study of Demand Shifting bottom of Figure 2 shows an example of this shifting with Self-generated Energy. Ubicomp 2014. visualization with the actual loads as dark circles (red in the [5] Barker, S.; Mishra, A.; Irwin, D.; Shenoy, P. and Albrecht, J. original) and the best time for this load in light circles (green in (2012) “SmartCap: Flattening peak electricity demand in smart homes.” 2012. 67--75. the original). For example, towards the middle of the month this household carried out four washing loads, during the [6] Haghighi, Pari Delir, and Shonali Krishnaswamy (2012). Role of context-awareness for demand response mechanisms. In afternoon (shown as dark circle), and as light circle is indicated Information and Communication on Technology for the Fight that the morning would have been a better time for these loads, against Global Warming, 136–149. Springer, 2011. given the specific weather conditions for that day. Using this