1. Introduction
Buildings account for about 40% of global energy consumption and for roughly 39% of global material resource use [
1]. Efforts to increase sustainability in the building sector focus primarily on technological innovations as well as policy [
2]. However, whether these technologies are implemented also depends on how decisions are made in building management, construction, and consumption. Cognitive science has well documented that human decisions are not always rational and has observed systematic deviations from purely rational behaviour, so-called cognitive biases. However, the influence of cognitive biases on sustainability and energy-related decisions in the building sector has been sparsely studied so far.
In recent decades, there have been strong technological developments in building technology that have made better building envelopes, HVAC (heating, ventilation, and air conditioning) systems, and sustainable on-site energy generation much more accessible and affordable. This means that it is technically and economically feasible to design and construct buildings with very low energy consumption and low embodied energy. However, research shows that energy efficient and sustainable solutions and technologies are not always adopted [
3]. Wherever an aspect of sustainability such as building envelope U-value or total energy consumption is not enforced by standards, regulations, and legislation, the shift toward sustainability is driven by human decisions in the building design and operation process. Human decisions have cognitive, emotional, and social aspects. For example, the influence of social norms on energy use in buildings [
4] and of descriptive norms in engineering design [
5] and their ability to drive change has already been demonstrated. However, the findings of cognitive science and behavioural economics in the second half of the last century have made it clear that decisions are indeed not always rational.
The deviation of human behaviour from purely rational behaviour is referred to as bounded rationality (e.g., [
6,
7,
8]) and has three main sources: lack of information availability, the use of heuristics—i.e., rules of thumb—and systematic deviations from rational behaviour, so-called cognitive biases. Dozens of these biases have been documented in the scientific literature in different contexts. One example is the endowment effect, which describes the tendency to attribute a higher value to an object if it is in one’s possession than if it is not [
9,
10].
Several studies evaluated the effects of bounded rationality in the building sector. Martin and Perry [
3] (p. 312) found “…sustainable construction technology adoption to happen serendipitously…” and Christie et al. [
11] (p. 457) report that “…if the rate of adoption of energy-efficiency technologies is to increase, then appeals to rationality are unlikely to work for a large proportion of homeowners.” However, the influence and potential of cognitive biases in the building design process have been studied only sporadically. In their review on bounded rationality in engineering design, Klotz et al. [
12] (p. 226) report that “…for design, biases, and choice architecture interventions remain underexplored and disconnected across fields of practice and academic disciplines.”
However, several studies point to the potential of systematically exploring cognitive biases in the building sector. In a seminal experiment, Klotz et al. [
13] showed an anchor effect for building energy performance goals in a survey of 76 engineers: completing a series of questions regarding a “90% energy reduction” vs. questions with a “30% energy reduction” inspired more ambitious energy reduction goals in the 90% group. In another study, a higher consideration of sustainability was achieved when building design choices were framed in design software as a loss of sustainability—starting from a sustainable variant—rather than a gain [
14]. Ebeling and Lotz [
15] found an impressive default effect for electricity consumers’ decisions, where an opt-out green power option generated about ten times more actual contract purchases compared to an opt-in scenario for the same green power. Harris et al. [
16] examined various cognitive biases in monitoring-based commissioning for the building energy management process used to optimise building energy performance. The data showed that almost 30 percent of the barriers faced by their cohort were caused by diverse cognitive biases. Furthermore, Shealy and co-workers [
5] showed empirically that designers for a civil infrastructure system exposed to positive descriptive norms set 28 percent higher sustainable performance goals on average. A recent study by Hancock and colleagues [
17] with 261 professional infrastructure designers investigated the effect on future discounting by using a present- and a future-frame. Their work shows that the use of a future-frame for a sustainability request proposal promotes the construction of significantly longer targeted useful life to the community, significantly longer design life, and the willingness of participants to accept a significantly higher number of years for return on investment. In a large survey, Blasch and Daminato [
18] found an association between a measure of the strength of the status-quo bias of participants and the age of their appliances and their level of consumption of energy services (dishwasher, washing machine, etc.). In a review paper, Delgado and Shealy discuss the potential of using cognitive biases to improve energy efficiency in facilities via choice architecture, presenting several suitable cognitive biases [
19].
These studies show potential for improvements in sustainability through cognitive biases. There are some limitations, however, as previous research has only considered some of the dozens of cognitive biases described in the psychological literature, and empirical evidence for cognitive biases in the field of building energy decisions is still scarce. Furthermore, some parts of the design process have not been studied at all, such as the iterative evaluation of the planned building design by building professionals prior to its construction [
12]—which is actually critical to achieving sustainable design. The building design process is crucial, as it is the part of the whole construction process where most decisions with long-term implications are made. However, cognitive biases can be taken accounted for in the design process by deliberately introducing changes in the decision environment. This process is called “choice architecture” and is applied frequently already in everyday life, for example in the form of calorie information on food packages which can give a certain framing to the consumption. The ethical considerations of applying choice architecture for sustainability or more general “green nudges” were considered in detail by Schubert [
20], who concluded that they can be an effective and ethical complement to traditional incentive-based measures if they are organised in a transparent way, i.e., so that people can in principle “unmask the manipulation”.
Considering the known cognitive biases, we identified four cognitive biases described in the literature that may be relevant to building sustainability:
The
framing effect describes that different formulations of a message—without changing the content—can change a viewer’s preference [
21]. For example, attitudes towards a method may depend on whether it is said: “There is an 80% chance that no error will occur” compared to “There is a 20% chance that an error will occur”. Even when the information conveyed is the same, decision makers tend to behave more risk-seeking when a positive frame such as the first is used, compared to more risk-averse such as in the second frame.
The
anchor effect describes how a decision maker’s judgement can be anchored by information seen before the decision is made. For example, in one of Tversky and Kahneman’s original experiments, participants were presented with a supposedly random number, which however would only take the value of 10 or 65. When subsequently asked to estimate the number of African countries in the United Nations, median estimates were 25% and 45% for the participants who saw the numbers 10 and 65, respectively [
22].
The
default effect describes the increased likelihood that an option will be chosen when it is presented as the default response, which will be chosen when the decision maker does not make any changes [
23,
24]. For example, Johnson and Goldstein showed that the percentage of those agreeing to organ donation can be increased by assigning participants to an opt-out condition (action to decline organ donation) compared to an opt-in condition (action to accept organ donation).
The
decoy effect describes that the preference of a decision maker can be nudged between two options by constructing a decoy that is asymmetrically dominated [
25]. An option (decoy) is asymmetrically dominated if it is inferior compared to one option (target), while it is partially inferior and partially superior for the other option (opponent). The inclusion of a decoy increases the proportion of decision makers who choose the target over the opponent. In marketing, the goal is usually to maximise revenue. In the context of building sustainability, however, the decoy effect can also be interesting for the promotion of environmentally friendly products.
These four cognitive biases can be directed at three different levels of energy-related decision-making: (a) general attitude towards building sustainability (framing effect for greenhouse gas emissions from buildings), (b) design decisions at the level of planners, home owners and managers (anchor effect for solar panels and default effect for building design), and (c) consumer decisions of building users (decoy effects for electrical appliances). Based on the described findings on cognitive biases and choice architecture, the goal of the present study was to find effects of bounded rationality in the building context that could be used to increase sustainability in this area. Specifically, we looked at the following research questions, each considering one of the mentioned cognitive biases:
- 1.
Does the framing effect increase awareness for greenhouse gas emissions from buildings through appropriate framing?
- 2.
Does the anchor effect increase decision makers’ willingness to pay more for solar systems when a higher anchor is introduced for past prices?
- 3.
Does the default effect increase the willingness to pay more for an energy-efficient home by offering the energy-efficient options as default?
- 4.
Does the decoy effect increase the proportion of decision makers who choose a more energy efficient target option in the purchase of an appliance (refrigerator, vacuum cleaner) when an asymmetrically dominated decoy is introduced?
In order to decrease the knowledge gap on cognitive biases in the building sector as well as to identify specific decision contexts that could be optimised towards sustainability, we conducted a behavioural study based on different decision scenarios to answer the above research questions using online surveys. We identified three cognitive biases, namely the framing, default, and the decoy effects that can be utilised to create more sustainable decisions in and around buildings. The decision contexts that we studied here were selected in such a way that they can be transferred to real-life decision situations where they can actually increase sustainability in the building sector.
In the following, we will first describe the general methods, then the methods, results, and discussions of the four research questions individually, followed by a general discussion and the conclusions.
2. General Methods
To answer our research questions, we designed a survey using the online tool LimeSurvey [
26]. To enable high performance and quickly available survey data, the survey was offered on the survey platform Prolific [
27].
2.1. Ethical Considerations
We consulted with a member of the Empa ethics committee for the proper implementation of the consent form. To ensure fair collaboration with the participants and to counteract the sometimes poor worker protection of crowd-sourcing platforms, we sought the most ethical platform possible, paid workers above the minimum wage, anonymized all data, and had participants read and accept the consent form. The online provider Prolific [
27] guarantees worker protection, requires minimum wages, and offers general accessibility to participants as well as cross-stratified representative samples (gender, age, ethnicity) from the UK and US, and niche recruitment. Prolific has a satisfactory reputation for ensuring high quality survey participants [
28,
29] while protecting their participants.
2.2. Participants
Having selected Prolific for participant recruitment due to the described ethical considerations, the geographic location of the participants was limited to the UK and the US. We decided to recruit participants who were all current UK residents as Prolific has the highest reach in the UK (March 2020: 37,791 UK participants active in the 90 days prior to study start) and, as the UK is similar in terms of climate, culture, and building industry to other Middle and Northern European countries. This makes the studied decision contexts of this study relevant for a number of countries in Europe and likely to Northern America. The effects of cognitive biases are in all likelihood more general than this context and may be applied globally by accounting for cultural and industry differences. We paid participants £h for completing the study, which was above the national minimum wage in the UK at the time of the study (spring 2020). In addition, participants were profiled on Prolific as follows: at least 20 previous study participations on Prolific, acceptance rate of previously submitted surveys on Prolific ≥90%, and no participation in our previous study. To achieve gender balance, we offered participation in the survey to an equal number of women and men. In total, 300 participants completed the survey on Prolific. Four participants read the description but did not open the survey and were therefore not included for analysis. Participants were randomly assigned to two groups with different treatments, groups A and B. Three attention questions were included, which were easy to answer for anyone who actually reads the given texts. Due to insufficient performance in these questions, the answers of 25 participants had to be removed, resulting in a total number of 271 responses (136 group A, 135 group B). Unless otherwise stated, all further analyses were conducted with the remaining participants.
The average survey completion time was 15 min (
min). Prolific provides certain demographic information of its participants directly through their profile (age, gender, student status), while we used additional questions to collect other information (gross household income, homeowner status, highest education, ethnicity, environmental rating). Participants had a mean age of 34 years (
). Of the 271 participants, 137 were women and 134 were men; 64 were students. The median of the gross household income was in the range from £30,000 to £39,999, 110 participants were homeowners, as their highest degree the majority of participants (99) had a Bachelor’s degree (BA/BSc/other), and the majority of participants (229) described themselves as white. For more details, additional tables and figures are given in
Appendix A.
In addition, participants were assigned points for their answers in three questions rating their environmental attitudes, with more points corresponding to a more environmental-friendly attitude. Out of a total of 15 points, the average environmental attitude was 12.89 (). Furthermore, three cognitive reflection test (CRT) questions were asked. The distribution of correct CRT questions was: 112 zero correct CRT questions, 66 one correct CRT questions, 52 two correct CRT questions, and 41 three correct CRT questions.
2.3. Procedures
Participants were active survey takers on Prolific, where they could browse the survey descriptions of different surveys from different institutions and select the ones that they were interested in. For each survey available, Prolific presents a description page including payment, duration of study, remaining study slots, and a text from the researcher. We described the research team to the participants and that we were interested in their opinions on issues related to energy in buildings and environmental issues. We emphasized that the survey must be given full attention and that attention would be controlled by corresponding questions. Such questions are a standard procedure at Prolific and the question texts were approved by Prolific staff. Participants were also informed that they would be given more time than necessary so that they would not have to rush through the survey. Lastly, contact information was given. By executing the survey, participants consented to our use of the anonymized data collected for publication in a scientific journal. After a brief CAPTCHA test and providing their Prolific ID for participation, we asked our demographic questions described above. The CRT questions were placed at the end of the survey so as not to exhaust participants before the main section. The remainder of the survey consisted of questions about cognitive biases related to building energy decisions, which are explained in the next section. With the exception of the order of the demographic, environmental assessment, and CRT questions, all questions were randomly ordered. That is, the order of all question groups and questions within question groups was randomized. For questions with different answer choices, the order of these was also randomized. The survey questions can be found in
Appendix B. It should be noted that there were new pages for questions that were unrelated and participants could not return to a question they had already answered.
2.4. Data Analysis/Data Collection
Data analysis was performed using the statistical programming language R [
30]. Responses were either Likert scale, multiple-choice, importance rating scales, or numerical inputs. Standard statistical hypothesis tests were used to compare participants from group A and B (two-sample
t-test, two-proportion
z-test, Wilcoxon rank sum test). We used an alpha level of 0.05 for all statistical tests.
In addition to the test results, we present effect sizes for interpretation. For a two-sample t-test, we use Cohen’s , where for are the group means and is the pooled standard deviation. For the two-proportion z-test, we use Cohen’s , where for are the group proportions. For the Wilcoxon rank sum test, we use Pearson’s , with sample size n and the W-statistic is standardized to z.
In the following sections, we present the methods, results, and discussion of the four cognitive biases studied: framing, anchoring, default, and decoy effect.
7. General Discussion
With this study, we reduced the gap in empirical research on cognitive biases in energy-related decisions in the context of buildings, as addressed for example by Klotz and colleagues [
12]. As summarized in
Table 3, the collected data support three out of four of our alternative hypotheses with small to medium effect sizes on the existence of cognitive biases in energy-related decisions in buildings at three different levels of decision-making. The results indicate that cognitive biases can shift decisions towards more sustainable alternatives and increase the focus on energy-related issues in buildings as well as the willingness to spend more money on more sustainable alternatives. Building on initial work on cognitive biases in the building process by Klotz and colleagues [
13] as well as Shealy and co-workers [
14] who reported single cognitive biases, namely an anchor and a framing effect, we have shown three cognitive biases: a framing effect that can increase awareness of greenhouse gas emissions, a default effect leading to a more sustainable building design and a decoy effect shifting consumer choices towards more energy efficient devices. These cognitive biases can be applied in real-life scenarios, such as the building design process, in order to favour decisions with more sustainable outcomes.
However, the experimental approach chosen here has two main drawbacks: first, the decision scenarios were hypothetical and no money had to be spent by the participants. Second, the participants were not building professionals or all building owners who are responsible for these type of decisions.
Regarding the first issue, in reality, sustainable choices are often a trade-off between price or convenience and level of sustainability, where decision makers usually need to sacrifice either money or comfort for the positive feeling of protecting the environment. In a hypothetical experiment, however, participants get that feeling for free. This could lead to less time and effort being spent on really weighing up the options, which introduces a bias. Furthermore, participants may be more willing to act sustainably—complying to a general social bias of being environmentally friendly—since it comes at no real cost. However, while the experiments and the amounts participants were willing to pay do not reflect reality one-to-one, they do show that cognitive biases have an effect in these types of scenarios. Since cognitive biases are systematic deviations from purely rational behaviour, it is likely that our findings are at least partially transferable to decisions with real financial consequences, possibly with weaker effects. The fact that cognitive biases can be realized in financially significant contexts is also shown by the broad application in marketing that is already taking place.
The second limitation of this study is concerned with how well the findings can be transferred to the decision-making of experts who work in the field. This limitation only affects research questions 2 and 3 on the anchor and default effect, respectively, since research question 1 is covering the general awareness with respect to GGE from buildings and question 4 is a consumer decision, which is relevant to the general population since most people own a vacuum cleaner and probably many a fridge. For research questions 2 and 3, it needs to be considered that about 40% of the participants in our study were in fact home owners, for whom these questions may be a real-life scenario. Finally, since the propensity for cognitive biases is a trait common to humans, our sample of mixed background does provide a basis for conclusions towards the target group. For example, even experienced engineers, scientists, or managers can be influenced by the framing effect [
33] and emission trading research has shown that practitioners, i.e., professional traders, seem to show significantly stronger endowment effects compared to students and that only practitioners show status quo bias [
34].
From the above discussion, it can be concluded that the cognitive biases demonstrated here in an online study with a general population would likely also be effective in a group of experts who are faced with these types of decision scenarios in their professional activity. It should be noted that the effect sizes observed here were small to medium. That means that the potential for behaviour change through these biases is limited in its size. Furthermore, it needs to be assessed how well our findings translate into real-life decision scenarios. A more general limitation of the use of cognitive biases towards increased sustainability is their limited behavioural effectiveness: Schubert points out [
20] that green nudges are highly context-dependent and may only have temporary effects. Hence, they should be seen as a complement to traditional policy making, based for example on incentives or laws, however one that comes at significantly lower costs to implement. Thus, the examined biases are of interest to be used in real-life scenarios by modifying existing decision contexts in order to promote more sustainable decisions, i.e., nudge people to “greener” outcomes.
The question remains if applying choice architecture for sustainability is ethically desirable since it is at odds with preventing the occurrence of cognitive biases in order to increase transparency and autonomy in decision-making. We need to consider, however, that we are probably limited in the extent to which we can prevent cognitive biases at all. Shealy et al. [
14] (p. 3), for example, concluded “Whether intentionally designed or not, there is no neutral framework to present information.” Eliminating one can introduce another cognitive bias (e.g., trying to create a neutral frame for one group of people may introduce an intervention for another group, etc.). On the other hand, some observers might interpret the active application of choice architecture as sinister or evil, whereas sustainability is probably generally considered heroic or good. This would amount to “using evil for something good.” However, staying out of it and being as neutral as possible (e.g., avoid “nudges for sustainable buildings”) could tip the overall balance into the wrong direction, as choice architecture is already widely used by other parties, e.g., in marketing, mostly to maximise sales, retain customers, or promote products (which are not necessarily environmentally friendly). Thus, maybe there is a silver lining in actively using cognitive biases to promote sustainability—a goal which in many countries has been democratically decided upon—instead of going with the flow of monetary interest groups. Furthermore, since the struggle for sustainability is a collective affair, a centralized approach through choice architecture might be more appropriate than a decentralized one where the decision context for the individual would need to be as neutral as possible. This is because the individual may lack the expertise and incentive to choose an appropriate sustainable option and may be better served by a “soft guidance” in the form of choice architecture.
Thus, the systematic study of cognitive biases and its subsequent input for choice architecture of the building design process may lead to a significant decrease in building energy use, embodied energy, and carbon footprint. Furthermore, such an adaptation of the design process could also be utilised to create “nudges” towards the use of more sustainable materials (e.g., materials without harmful components) or the adherence to the principles of circular economy (design for disassembly, possibility to reuse materials). While some of these new design pathways might initially be a bit more expensive, in the long run, overall costs are likely to decrease as sustainable designs, systems, and materials become more commonplace and as negative consequences of a non-sustainable approaches are factored in (e.g., carbon penalty, etc.). Finally, choice architecture in the building design process is a much cheaper measure to increase sustainability compared to interventions based on regulation or incentives.
8. Conclusions
As the influence of bounded rationality on sustainable energy decisions in buildings is largely unknown so far, the empirical evidence presented here can stimulate answers to questions such as how cognitive biases prevent the adoption of available technology, how this can be remedied and, on the other hand, how cognitive biases can be used in the design process in order to favour sustainable options. In this work, we found evidence of three cognitive biases—framing, default, and decoy effect—that may nudge the results of decision-making processes relevant to the energy use in buildings towards increased sustainability. Our data of approximately 270 participants from the UK show that one can increase the willingness to pay more for an energy-efficient home by offering it as default quote (medium effect), raise awareness of greenhouse gas emissions in buildings by framing (small effects), and increase the proportion of decision-makers who choose a more energy efficient target option by the use of a decoy (small effects). On the other hand, we were not able to support the hypothesis that we can increase the willingness to pay more for a solar system by anchoring the decision-makers with past prices. These results promote the identification of pathways to an optimised building design process using choice architecture, which may help to create more energy-efficient and sustainable buildings. They can also help to “leverage” the introduction and uptake of novel sustainable building technology by removing cognitive barriers. Since cognitive biases have scarcely been studied in buildings so far, our findings significantly increase the available “toolbox” for such an endeavour.
On a more theoretical level, this study further deconstructs the tacit assumption that building professionals and building owners behave rationally, simply optimising the financial outcome within given boundary conditions: homo faber is not the homo economicus.
Our findings constitute a basis for further studies with building professionals or homeowners entailing real financial consequences. It is important to assess how our findings translate to decision processes in the field, if effect sizes will be different, and if the decision scenarios need to be adapted for these contexts. Ideally, this will be done in collaboration with architectural offices or building companies. On the other hand, this study indicates that choice architecture theory could also be applied to other developments in the built environment. For example, there are many behavioural and decision barriers in implementing the principles of circular economy for buildings, or one can focus on studying the promotion of CO-neutral materials.
In summary, we have shown that cognitive biases in the building context can be used to support the exploitation of already available and economically feasible technological solutions by encouraging people to make more sustainable choices.