According to the categorisation-based model of feature substitution, when people infer the energy consumption of an appliance from a feature such as its size, they categorise the appliance in terms of how typical it is of a large appliance. The degree of typicality in the large appliance category is mapped to a similar degree of typicality in the high-energy-consuming appliances category. That mapping of categories is the process of feature substitution and it provides the person with an answer to the difficult question about how much energy the appliance consumes.
This categorisation-based model of energy judgements is provided to give context to the research reported in the rest of this thesis (though testing the model is beyond the scope of the thesis). It attempts to fill the gaps in the theory of fea-ture substitution by describing the psychological processes underlying energy judgements. The model corresponds with the concept of representativeness dis-cussed as part of feature substitution by Kahneman and Frederick (2002), the categorisation-based models described by Schuitema and Steg (2005) and van den Broek (2016), and the categorisation theories assumed in other studies of energy judgements (Baird & Brier, 1981; Kempton & Montgomery, 1982). By using cat-egorisations, people can retrieve knowledge they already have about appliances and infer information that they do not already have.
Chapter 3
Exploring Perceptions of Energy Consumption (Study 1)
3.1 Introduction
Judging the amount of energy consumed by individual household appliances is difficult so householders try to simplify the judgement by using heuristics. In heuristic judgements, instead of weighing up all the relevant information before making a judgement, the person instead ignores most of the information and makes their judgement based on only a small number of pieces of information, or cues. A small amount of research in the energy perceptions literature has shown that people might use cues based on features of the appliance and its usage by the household to infer how much energy the appliance consumes. The most com-monly investigated cue so far is the size of the appliance, though other cues have also been identified using a range of quantitative and qualitative methods. Each study has identified only a very small number of cues and often different cues from other studies. The study in this chapter takes a qualitative approach to un-derstanding householders’ perceptions of energy consumption and to identify the cues they use and how they use them.
3.1.1 Judging energy consumption with feature substitution
As argued in Chapter 2, providing householders with more information about their energy consumption does not help them to make judgements about their
energy consumption because they cannot organise and make use of so much in-formation in their judgements. If a householder were to take a “rational” ap-proach (Payne et al., 1993; Shah & Oppenheimer, 2008) when comparing the rela-tive amounts of energy consumed by a washing-machine and a kettle, for exam-ple, they must consider and weight a number of factors:
1. Select a time-frame (a starting point and an end point) during which the energy consumed is to be estimated. Ensure that the time-frame is relevant to both appliances. For example, 24 hours might be fine for a kettle that is boiled three times each day but not relevant to a washing machine that is run only once a week.
2. Select a unit of measurement, which might be determined by the purpose of the estimation. For example, kilowatt hours if the aim is to reduce energy consumption for its own sake; pounds sterling if the aim is to reduce energy bills; grammes of carbon if the aim is to reduce carbon footprint.
3. Consider the multiple factors that might contribute to consumption for each appliance. For example, for the washing machine: the number of loads of washing in the selected time period; which wash and spin programmes are selected; the energy intensity required by the washing-machine on those wash and spin programmes, how often the wash and spin programmes are run. For the kettle: the amount of water in the kettle; the energy intensity of the kettle; the effectiveness of the insulation on the kettle; how often the kettle is used.
4. Weigh up how much each of the factors influences the energy consumption of each appliance.
5. Based on the weighted information for each appliance, estimate the energy consumption.
6. Make a judgement about which appliance consumes the most energy based on the previous steps.
There are far more relevant factors to consider in Step 3 than listed and it is highly unlikely that most householders could access the values of them all, estimate relative weightings for each of them, and then (keeping all this infor-mation in memory throughout the judgement) integrate it all together to decide
which appliance consumes more energy. Instead, people save effort by ignoring the majority of the relevant information to simplify the judgement (Gigerenzer &
Gaissmaier, 2011; Shah & Oppenheimer, 2008).
The heuristic process of feature substitution (Kahneman & Frederick, 2002) involves selecting a single cue at a time and substituting an energy consumption estimate with an estimate of the heuristic cue. For example, when comparing the relative energy consumption of the washing machine and kettle, the householder could compare their relative sizes, substitute energy consumption with size, and decide that the washing machine (as the larger appliance) consumes more energy than the kettle. With just a single cue, based on a single feature of the appliances, the householder can make a reasonable attempt at the judgement with signifi-cantly less effort than if they had tried to make a fully “rational” judgement.
3.1.2 Identifying heuristic energy consumption cues
A small amount of previous research in the literature has found a small number of heuristic energy consumption cues using a range of qualitative and quantita-tive methods. The studies have also assumed a heuristic process of (or similar to) feature substitution. The most commonly investigated cue has been the size of the appliance, which has been identified in two quantitative studies (Baird &
Brier, 1981; Schuitema & Steg, 2005) and one qualitative study (Chisik, 2011).
Other cues have been identified but each by different studies and only a small number of cues by each study. Table 2.1 in Chapter 2 summarises all the cues by the studies in which they were identified1. The aim of the study reported in this chapter was to explore the cues that householders can use to make heuristic energy judgements using feature substitution.
The mostly quantitative studies in the literature selected the cues to test based on theory and reasoning by the researcher, which limited them to identifying only those cues that the researcher could propose in advance. Baird and Brier (1981, Exp. 2 & 3) tested whether the size of the appliance was used as a cue based on an initial card sort task (Baird & Brier, 1981, Exp. 1). In the card sort, when partic-ipants were asked to sort a large set of appliances into groups according to their energy consumption, the investigators observed that participants had appeared
1The interviews in the study reported in this chapter were conducted at around the same time as van den Broek (2016) conducted her focus groups and without knowledge of her study.
Discussion of the links to her study are made later in relation to the findings of the present study.
to group the appliances mostly by function then by their relative sizes. Schuitema and Steg (2005), citing Baird and Brier (1981)’s findings, tested the size of the ap-pliance, as well as its visibility in the home, and the amount of noise it makes on the basis that all three are easily observable and so take little effort to perceive.
Schley and DeKay (2015) based their testing for the use of cognitive accessibility on the widely-known heuristic theory of availability. They argued that the more householders use and think about an appliance, the more “available” the appli-ance is in memory, the easier it is to recall and so is perceived to consume more energy.
The qualitative studies by Chisik (2011) and Kempton and Montgomery (1982) took more open-ended approaches but identified only a few cues each. Chisik (2011) asked participants to draw pictures of the appliances that consumed the most electricity in their homes and observed that size, frequency of use, and duration of use were perceived as proxies for the energy consumption of the appliances. Kempton and Montgomery (1982) conducted ethnographic inter-views about energy consumption and observed a small number of cues, including time switched on and visibility. The more open-ended methods of these studies showed promise for identifying cues from the perspective of householders but the varied results of the two studies suggests that neither identified a compre-hensive set of cues.
The quantitative studies in the literature selected and tested cues based on the-ory and simple exploratthe-ory studies, while the qualitative studies enabled house-holders to give their perspectives and to potentially identify other cues not de-duced by researchers designing quantitative studies. The semi-structured inter-view approach in the present study was designed to ask open-ended and non-leading questions in order to access participants’ perceptions without interfering with them as much as possible (Morgan, Fischhoff, Bostrom, & Atman, 2001).
This approach was intended to identify a larger set of cues, ideally verifying some of those already identified in the literature and reporting new ones. As well as aiming to identify a set of cues, a qualitative approach would provide greater in-sight into how people use the cues and any other strategies that they might use in order to infer energy consumption. In particular, it would help ascertain whether people use just a single cue to make their judgements as Schuitema and Steg (2005) concluded or whether they use multiple cues as Baird and Brier (1981) sug-gested. While Schuitema and Steg (2005) concluded that their participants were
using only a single cue to infer energy consumption, it is not possible to know from their study whether participants might have used additional cues that were just not measured in the study. Baird and Brier (1981) concluded that their par-ticipants were using another cue in addition to size but they could not identify what it was from their study.
3.1.3 Interviews, card sorts, and thinking aloud
The study reported in this chapter used semi-structured interviews with simple card sorts to encourage and prompt discussion about participants’ judgements of energy consumption and how they made their inferences. Card sort methods are based on the assumption that people categorise everything that they experience to help them make sense of it. Instead of considering everything that they en-counter as unique, people try to reduce the cognitive complexity of the world by grouping, or categorising, things according to similarity by some criteria (Rosch, 1978). Card sorts are, therefore, commonly used to obtain insights into how peo-ple structure their understandings of the world (Barnett, 2004). The process of conducting card sorts is assumed to reflect cognition so it was appropriate to use a card sort to prompt discussions that would help access the underlying cognitive processes. These assumptions correspond with the realist perspective from which this research was conducted: the data obtained from the study were assumed to provide insight into participants’ perceptions of reality.
To help the researcher’s understanding of the cognitive processes being demon-strated by the card sort tasks, participants were asked to think aloud about what they were doing. Thinking aloud during tasks such as card sorts is generally ac-cepted to provide valuable insight into the focus of attention of the participant at any given time during the task (Lucas & Ball, 2005)—in this case, the sequence of considerations the participant was making whilst judging the relative amounts of energy consumed by the appliances on the cards. This was especially beneficial for this study in which it was the process of judging and comparing the energy consumption of the appliances that was of interest.
Including the card sort tasks was aimed to help with participants’ engage-ment in the interviews. Having the participants generate their own set of cards labelled with appliances from their everyday lives ensured that participants were judging and discussing appliances with which they were familiar. Other benefits
of including card sort tasks were that participants generally find card sort tasks to be enjoyable and that sorting cards can be especially helpful for people who do not have strong skills at verbalising their thoughts and ideas (Barnett, 2004).
3.1.4 Research questions
The research questions addressed in this study were:
How do people try to infer the energy consumption of individual household appliances? In particular, what sources of information (cues) do they use to infer the energy consumption of household appliances?