Does Fan Feel and Visibility During Thermostat Interaction Affect Temperature Selection in Warm Ambient Conditions?

Higher global temperatures increase the demand for indoor air-conditioning (A/C) [1,2]. This increased A/C use not only consumes significant energy but also further warms the planet, leading to a vicious cycle. Notably, recent heat waves have increased power generation from coal-fired stations to help meet extra A/C demand [3]. Furthermore, A/C use significantly contributes to greenhouse gas emissions due to widely used refrigerants with high global warming potential [4]. The growing environmental harms of increased A/C demand motivated us to explore how to decrease its use.

Much A/C use is driven by current thermal comfort standards, i.e., temperatures are set lower than what is required to maintain health [5]. Thus, we devised and explored an intervention that could improve comfort at the time of temperature adjustment. Specifically, we examined the potential role of mechanical fans as part of the thermostat interface. Fans enhance cooling by increasing convective heat transfer. In general, fans deliver instant, targeted thermal relief, while consuming significantly less electricity than using A/C to cool an entire indoor space [6]. The current study proposes using mechanical fans, not to cool an entire space, but as an intervention while thermostat adjustments are being made.

While the role of information in increasing environmentally conscious behavior has been extensively studied, we wanted to explore product features that can increase such behavior. Other research supports that a product’s design can significantly affect environmentally conscious practices. For example, our laboratory’s past work suggests that products which present resources in discrete units rather than continuous flows can reduce resource consumption [7]. Srivastava and Shu also examined the role of product affordances on eco-conscious behavior [8]. Telenko et al. compiled 76 design-for-environment guidelines, which include, “Incorporate features that prevent or discourage waste of materials by the user” [9]. Withanage et al. incorporated behavioral observations to design for sustainable appliance use [10]. She and MacDonald examined how a product’s sustainability triggers affect pro-environmental decisions [11]. Pakravan and MacCarty integrated usage context and user behavior in designing for clean-technology adoption [12]. Adding to this body of work, the current research examined whether mechanical fans can be used to affect temperature-adjustment behavior during thermostat interaction.

Before describing the details of our study, below we summarize existing technologies and their limitations as context for our approach.

While the primary function of household thermostats is to regulate temperature according to a setpoint, energy efficiency and cost savings have become common, secondary goals [13]. Thus, thermostat technologies, e.g., smart, learning-, schedule-, occupancy-, and feedback-based strategies, have been developed to conserve energy. For example, smart thermostat technologies, such as the Nest, learn and generate a schedule based on user-set temperature adjustments [14]. However, schedule-learning functionality does not necessarily produce an energy-efficient schedule, particularly if user-specific temperature adjustments are inefficient [15]. Furthermore, Yang and Newman attributed energy savings to participants’ behavior, and not the thermostat’s automated functionalities [14]. Therefore, the energy-saving potential of schedule-learning smart thermostats is limited by users’ motivations to enact eco-conscious adjustments.

As another example, feedback-based interfaces reportedly provide users with increased usability and practical energy-saving recommendations. The ThermoCoach monitors occupancy patterns and recommends schedules to minimize energy consumption [16]. Three types of schedules are suggested to the user: (1) high comfort, (2) energy saver, and (3) super energy saver, alongside their current schedule. The user can then choose to activate one of these schedules. In a 12-week field study comparing users’ pre-intervention schedule to ThermoCoach’s recommended scheduling, the interface saved 4.7% more energy than a programable thermostat and 12.4% more energy than the Nest thermostat. Thus, a feedback-based interface can conserve energy, but still relies on the user to choose an energy-saving schedule.

Overall, the promise of energy-saving technologies has been limited by user behavior. That is, when experiencing thermal discomfort in the moment, users can override these systems and continue to set energy-inefficient temperatures.

Many frameworks exist to describe behavior-change strategies to increase sustainabilty [17,18]. One categorization method differentiates between informational and structural strategies [17]. Informational strategies aim to change individuals’ perceptions and motivations [19]. In contrast, structural strategies change the circumstances wherein decisions are made.

Pro-environmental behavior (PEB) interventions can also be considered on a range based on user versus product in control. Both extremes of this range, information-based at one end and automation/forcing at the other end, have limited effectiveness or may create new challenges [18]. Products that focus on informational strategies, e.g., by providing feedback, enable a high degree of user control [20,21]. For example, the ThermoCoach suggests energy-saving schedules [16,22], but a motivated user must activate a schedule to realize energy savings. At the other end of the range, highly automated smart thermostats can increase user convenience [23], but often at the cost of limited user control and lack of understanding around automated adjustments [14].

Interventions that aim to change beliefs, values, and motivations assume that people are rational agents who make reasoned decisions [17,24,25]. However, time, availability of information, and human computational power all significantly limit people’s ability to engage in effortful decision-making [24,26,27].

Another approach to encourage PEB involves changing people’s external circumstances [17,18]. Doing so may have downstream effects on individuals’ (1) perception of their abilities, (2) habits and behavioral patterns, (3) expectations of the self and others, and (4) attitudes, beliefs, and values [18].

Changing the physical environment to enable PEB is a valuable behavior-change strategy, as it does not rely on underlying beliefs and attitudes. For example, cigarette littering was reduced by 64% with structural interventions alone, e.g., by adding ashtrays and litter bins [28]. This significant change in littering behavior was achieved without altering attitudes on littering. As another example, a water-conservation intervention that reduced the rate of waste-water outflow was found to be effective, regardless of whether people reported practicing PEBs in general [29].

Such results support the design of interventions that do not rely on pro-environmental attitudes and intentions. More generally, external factors that can be changed to increase PEB include the: (1) physical environment, including infrastructure and facilities, (2) availability of products and product characteristics, and (3) incentives and constraints [17,18].

Withanage et al. classified wasteful behavior as energy overuse failure modes, which are attributed to high-energy-consuming habits. Such wasteful consumption can be reduced through either product design or behavior-change strategies [10].

For the current work, habitual energy-inefficient temperature setting to achieve comfort cooling can become the default choice [24], unless that choice is challenged or the decision environment is altered. An opportunity is identified by the above gap, which can be addressed through product design. That is, improved user interaction with products can increase energy-efficient behavior [9].

Specific to thermal comfort, energy-inefficient behavior can be driven by sensory stimuli from the environment, i.e., one feels hot, so adjusts the thermostat. Thus, altering the physical environment wherein temperature decisions are made is a promising behavior-change strategy.

We hypothesized that alleviating thermal discomfort during user-thermostat interaction may prompt behavior-change toward more energy-efficient temperature adjustments (i.e., higher indoor temperatures in warm ambient conditions). Therefore, we tested whether experiencing the cooling sensation of a fan, while setting the thermostat, would lead to more sustainable temperature selections, which could ultimately reduce energy consumption. Our results would inform the potential redesign of thermostats to reduce A/C use.

Many behavior-intervention studies elicit and report on participant intent [30]. For example, study participants are asked to complete statements to express intended behaviors, e.g., “Based on this, I will likely …” and “I intend to …” [31,32]. This approach is not ideal, as intention may not lead to actual behavior. Therefore, our study examined actual behavior to determine the efficacy of a fan-based experimental thermostat interface.

Our fan-based intervention interfaces were tested in-person at a university laboratory, where the temperature ranged from 24 °C to 27 °C (75.2 °F to 80.6 °F) and relative humidity (RH) ranged from 30% to 63%. The laboratory’s indoor temperatures varied due to lack of effective A/C. However, participants arrived at the laboratory from an outdoor environment, also of significantly varying conditions. For example, on a representative day, outdoor temperature varied from 20.1 °C to 33.9 °C (68.2 °F to 93.0 °F) and RH ranged from 40% to 80%.

The laboratory consisted of two adjoining rooms. Participants would interact with an interface in one room, while the research assistant prepared the next interface, out of participants’ view, in the adjoining room. Four experimental interfaces were prepared according to participants’ randomly assigned order of conditions. After interacting with each interface, participants would complete a filler task (described below). The research assistant then directed the participant to the adjoining room to interact with the next interface.

To approximate the real-world conditions of setting a thermostat, participants were asked to adjust temperature using a physical thermostat. This contrasts with specifying temperature adjustments in questionnaires [33] or through participant–researcher communication [34]. However, the physical characteristics of the study location made it infeasible to achieve the selected temperatures while the participant was undergoing the study. Thus, the thermostats did not affect the room temperature.

Most manufacturers recommend that residential thermostats be mounted 52 to 60 in. from the ground to account for natural air movement. To approximate this height, the experimental thermostats were mounted onto an existing structure that is 20 in. wide and 52 in. high. The top of this structure also enabled placement of the intervention-interface fans (∼5 in. deep) at face level.

LuxPro PSD011B pro-spec digital thermostats were selected due to their simplistic interface (Fig. 1 (left)). The large digital display only showed the current temperature, which could be adjusted with up-and-down arrow-shaped physical buttons. The thermostat had a temperature control range of 7–32 °C (44.6–89.6 °F). Two side switches and labels for fan setting (auto/on) and thermostat mode (heat/off/cool) were covered with white electrical tape so that participants only saw the temperature-adjustment buttons.

Each experimental interface used a pair of 7-in.-diameter Xuenair (N2 Night Fan HND-N2) fans, shown in Fig. 1 (right). These fans had three speeds, and a maximum power of 5.5 W, using 4000 mAh batteries that were charged using a Type-C universal serial bus (USB) interface. For each study condition, a pair of fans (in below-described configurations) was placed directly above the thermostat, atop the structure on which the thermostat was mounted.

Table 1 summarizes the four fan configurations that comprised the study conditions. These conditions involved participants adjusting the experimental thermostat interface while fans were either (1) off or directed air (2a) unobstructed, (2b) behind a ventilation grill, or (2c) behind a plexiglass case that blocked all airflow. These configurations enabled exploration of how the state of the fan (feel and/or visibility) affects temperature adjustments. In all four conditions, a white-noise generator was used to mask different levels of fan noise.

A 2 × 2 factorial design was used to examine the effect on temperature adjustments of two independent variables:

1. fan feel, i.e., feel-fan (FF) versus don't-feel-fan (DFF) and

2. fan visibility, i.e., high-visibility (HV) versus low-visibility (LV).

In the feel-fan high-visibility (FF + HV) condition, there were neither physical nor visual obstructions between the fans and participants.

In the feel-fan low-visibility (FF + LV) condition, the fans were placed behind a white ventilation grill. While this condition was intended to hide the fans from participants’ view, they could partially see them behind the ventilation grill. Therefore, this condition was denoted low-visibility instead of no-visibility.

In the don't-feel-fan high-visibility (DFF + HV) condition, the fans were placed behind a clear acrylic case so they could be seen as on, but participants could not feel the fans’ effects.

Finally, in the don't-feel-fan low-visibility (DFF + LV) condition, the fans were set at the off position. Again, because the fans were placed behind a ventilation grill, participants could partially see the fans.

Twenty-five university-student participants were recruited for the study. Two participants were removed because they were unable to follow the study instructions. Of the remaining 23 participants, 20 were enrolled in the participant pool for an introductory psychology course, and 3 were part of the general university student population. Twenty-one were between the ages of 18 and 24, and two were between the ages of 25 and 34. Eleven identified as female, 11 identified as male, and one identified as gender queer.

Since temperature preference and thermal sensation effects are highly subjective and variable, a within-subject design was used to reduce individual differences between conditions. As described above, a 2 × 2 factorial design was used to examine the effect on temperature adjustments of two independent variables: (1) fan feel, i.e., feel-fan (FF) versus don't-feel-fan (DFF) and (2) fan visibility, i.e., high-visibility (HV) versus low-visibility (LV). Participants were randomly assigned to one of the two condition orders where fans provided either delayed versus immediate thermal relief.

1. Delayed fan-effect: DFF + LV, FF + LV, DFF + HV, and FF + HV.

2. Immediate fan-effect: FF + HV, FF + LV, DFF + LV, and DFF + HV.

The two orders were used to explore whether there was an effect of exposing participants to different sequences of conditions. The study took approximately 45 min to complete. The main factor in completion time was language competency. Due to in-person COVID-19 protocols, participants were required to wear face masks.

Participants were prompted to think about their thermal comfort before interacting with the thermostat interface. They first set an initial baseline temperature, which was compared with temperatures they set in four intervention conditions, ordered as described above. This procedure was devised to approximate a naturalistic temperature-adjustment environment.

Time was not specifically allocated for participants to acclimate to the study environment. Acclimation was not prioritized in order to mimic a target user who may interact with their thermostat in a state of mild-to-moderate thermal discomfort, i.e., after recently arriving from a warm outdoor environment. Additionally, the effect of each intervention interface on temperature adjustments could be more realistically measured during thermal discomfort. However, participants spent about 5 min receiving initial instructions before interacting with the intervention interface and setting a baseline temperature. Participants were then given a basic tutorial on how to operate the study thermostat. Overall, participants were not given time limits and could take as long as required to complete the tasks.

Participants were asked to bring a laptop or tablet to the study location. Upon arrival at the laboratory, participants were informed that they would be testing different thermostat interfaces for a product-design study. Participants then used their devices to (1) read and sign the consent form, (2) complete a basic demographic questionnaire, and (3) complete an evaluation of their thermal comfort. Next, participants set their baseline temperature before interacting with the intervention interfaces located in the adjoining laboratory room.

During baseline temperature collection, participants were asked to complete four tasks:

• Task 1: Think about what a comfortable temperature would be and set the thermostat to that temperature (the set temperature was recorded by a research assistant).

• Task 2: Answer, “What feature stood out the most about the thermostat? Did that affect how you used it?”

• Task 3: Answer, “What factors led you to select that temperature?”

• Task 4: Answer, “Why did you set the thermostat to the setting that you did?”

Participants answered the questions for tasks 2–4 on qualtrics^TM, a survey software. After the above four tasks for the baseline condition, participants completed a filler task where they counted syllables from randomly generated lists of words. A filler task was used to reduce carryover effects from one condition to the next. In the meantime, a research assistant prepared the subsequent interface configuration in the adjoining laboratory room. Next, participants proceeded to complete tasks on the four intervention interfaces in series.

After participants entered the adjoining laboratory room, they were provided instructions printed on paper. These instructions described study tasks 1 and 2 (detailed below), which they were asked to perform. The papers were collected by the research assistant upon task completion. Participants answered questions corresponding to tasks 3–5 on qualtrics^TM.

The research assistant instructed participants to stand in a 40 cm × 30 cm (15.7 in. × 11.8 in.) rectangle taped on the floor, at a typical distance where users interact with thermostats. The research assistant then asked participants to complete the following five tasks:

• Task 1: Take 30 s to think about your thermal comfort in this moment. Please write down how you are thermally feeling as you stand in the rectangle.

• Task 2: Think about what a comfortable temperature would be and set the thermostat to that temperature (the set temperature was recorded by the research assistant).

• Task 3: Answer, “What feature stood out the most about the thermostat? Did that affect how you used it?”

• Task 4: Answer, “What factors led you to select that temperature?”

• Task 5: Answer, “Why did you set the thermostat to the setting that you did?”

After each condition, participants completed a filler task of counting syllables in randomly generated lists of words. A different set of words was provided following each condition.

Pilot testing had revealed some sources of participant confusion. Many participants expected an instant change in room temperature or fan state after they set a temperature. Some participants removed the mounted thermostat and/or fans to play with them. To reduce this confusion and behavior, the following was posted next to the intervention interfaces:

Important information:

• Do not remove thermostat

• Changing the thermostat temperature does not guarantee change in room temperature

• You cannot control the fan through the thermostat

After undergoing all four conditions, participants completed assessments on:

1. Their environment, to assess perception of indoor air quality, including perceived humidity, stuffiness, smell, and draft.

2. Manipulation checks, effect perception, and feedback, to gauge participants’ understanding and experience with the intervention interfaces. Participants were also asked about the intervention interfaces’ efficacy in motivating them to set more energy-efficient temperatures.

3. Value orientations, following Steg et al. [35], that included 16 items: three on hedonic values, five on egoistic values, four on altruistic values, and four on biospheric values.

4. General study feedback, to elicit comments from participants in open-ended (free-text entry) format.

5. Energy myths, to explore whether holding the incorrect conceptual model affected temperature selection. Corresponding questions were based on Pritoni et al.’s assessment of knowledge regarding energy myths [36].

6. Post-thermal comfort using the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) 7-point thermal sensation scale [37].

Assessments 2–4 were exploratory, and responses were intended to inform future research.

Two-way analysis of variances (ANOVAs) were used to analyze the effects of fan feel, fan visibility, and the interaction between these two effects on both raw temperature selections and temperature shift. Temperature shift is defined as the difference between the temperature each participant set during a condition and the baseline temperature the participant set before any conditions. To assess differences in raw temperature selections between conditions, a repeated-measures ANOVA was conducted.

A two-way ANOVA was performed to analyze the effect of fan feel, fan visibility, and their interaction on temperature shift. Comparing conditions where participants felt the fan or not, our results show that feeling the fan significantly altered temperature selection, F(1, 88) = 5.73, p = 0.02 (Fig. 2). That is, participants selected higher temperatures when they felt the fan (1.72 °C/3.10 °F) than when they did not (0.56 °C/1.01 °F). There was no significant main effect for fan visibility (Fig. 3), nor any significant interaction between fan feel and fan visibility.

A two-way ANOVA was performed to analyze the effect of fan feel, fan visibility, and the interaction between these two effects on raw temperature selections. Comparing conditions where study participants felt the fan or not, our results show that feeling the fan significantly altered temperature selection, F(1, 88) = 8.78, p = 0.004 (Fig. 4). Our results showed a marginally significant effect on temperature selection for fan visibility, F(1, 88) = 2.98, p = 0.09 (Fig. 5). There was no significant interaction between fan feel and fan visibility.

Specifically, participants’ mean raw temperature selections for the two conditions when they felt the fan (23.9 °C/75.0 °F) were higher than for the two conditions when they did not (22.5 °C/72.5 °F). With respect to visibility, participants’ mean raw temperature settings were marginally higher when fan visibility was high (23.5 °C/74.3 °F) than when it was low (22.8 °C/73.0 °F).

A repeated-measures ANOVA was conducted to assess differences in raw temperature selections between conditions. Mauchly’s test indicated a violation of the sphericity assumption, χ²(9) = 26.05, p = 0.002. Since sphericity is violated (ɛ = 0.653), Greenhouse–Geisser corrected results are reported. Temperature selection was significantly affected by intervention interface, F(2.61, 57.43) = 9.18, p < 0.001 (Fig. 6). Post-hoc analysis with a Bonferroni adjustment revealed that temperature selections were significantly higher in the feel-fan high-visibility (FF + HV) condition than baseline (p = <0.001, 95% confidence interval (CI) = [−3.42, −1.35]). Temperature selections were also significantly higher in the feel-fan high-visibility (FF + HV) condition than all other conditions, i.e., feel-fan low-visibility (FF + LV) (p < 0.001, 95% CI = [−2.05, −.75]), don't-feel-fan high-visibility (DFF + HV) (p < 0.001, 95% CI = [−2.76, −1.08]), and don't-feel-fan low-visibility (DFF + LV) (p < 0.001, 95% CI = [−3.04, −.97]).

The feel-fan high-visibility (FF + HV) condition had an average temperature setting of 24.4 °C/76.0 °F, which was significantly higher than the average baseline setting (22.0 °C/71.7 °F). The average temperature selections for the feel-fan high-visibility (FF + HV) condition were also significantly higher than the feel-fan low-visibility (FF + LV) condition (23.2 °C/73.7 °F), don't-feel-fan high-visibility (DFF + HV) condition (22.7 °C/72.8 °F), and don't-feel-fan low-visibility (DFF + LV) condition (22.4 °C/72.4).

Finally, there was no significant effect of order, i.e., the order of conditions that participants underwent did not affect their temperature-setting behavior.

Results are discussed before turning to design implications.

This study supports that feeling a fan can encourage energy-efficient temperature adjustments during warm-weather thermostat interaction. When comparing temperature shifts (between conditions and baseline), participants selected relatively higher temperatures in the two feel-fan conditions (1.72 °C/3.10 °F) than in the two don't-feel-fan conditions (0.56 °C/1.01 °F). In terms of mean raw temperature selections, participants set higher temperatures when they felt the fan (23.9 °C/75.0 °F) than when they did not feel the fan (22.5 °C/72.5 °F), a difference of 1.4 °C/2.5 °F.

Increased air speed is an effective intervention towards prompting energy-efficient temperature adjustments. Specifically, low-powered fans alone afforded thermal relief during user-thermostat interaction. The effect of air movement on expanding user thermal comfort is empirically supported. For example, Zhai et al. examined the effect of air movement using floor fans on (1) thermal comfort, (2) thermal sensation, and (3) perceived air quality (PAQ) while varying temperature and RH [38]. The authors concluded that air movement can maintain thermal comfort up to 30 °C at 60% RH, and PAQ can be maintained up to 30 °C at 80% RH. Thus, mechanical fans provide an energy-efficient means toward thermal comfort in warm and humid environments by expanding thermal tolerance.

The current-study participants’ qualitative responses also support the effect of increased air speed on whole-body thermal perception. Specifically, when asked to reflect on their thermal comfort, 20 of 46 responses in both feel-fan conditions indicated improved thermal comfort and/or satisfaction. The fan’s cooling effect is summarized by one participant’s response as, “Very much enjoying the breeze. Breeze hits larger area of my body, cooling it. Not as concerned about warm ambient temperature as before [when not interacting with interface].” Therefore, as expected, fans alleviated the thermal discomfort associated with warmer-than-neutral indoor conditions.

In contrast, a minority of participants reported thermal discomfort due to the fan-created draft. This undesired cooling also affected temperature-adjustment behavior. Seven of 46 responses in both feel-fan conditions included references to thermal discomfort. One participant response clarified how this may affect behavior as, “I feel cooler since there is a fan. It is a little bit too cold for me, so I might set the temperature higher. I had to pull down my sleeves because it is a bit too cool.” For participants already feeling thermally neutral, sensitivity to draft can lead to thermal dissatisfaction [39], as evident in a minority of current-study participants. More simply stated, some participants who were not hot to begin with may have felt overcooled by the fan. Supporting this finding, other researchers have noted that thermal discomfort due to a fan’s cooling effect can be attributed to feeling thermally neutral before fan interaction [40,41].

Fan visibility, specifically high versus low, was manipulated to understand its effect on temperature-adjustment behaviors. In the two high-visibility conditions, the fan cue was salient, i.e., the stimulus was salient to the perceiver [42–45]. Temperature adjustments were marginally affected by fan visibility. That is, participants set slightly higher temperatures in the two high-visibility conditions (23.5 °C/74.3 °F) than in the two low-visibility conditions (22.8 °C/73.0 °F). However, this marginal relationship was not preserved when examining mean temperature shifts. That is, there was no significant difference between participant-set temperatures in the high-visibility conditions minus baseline temperatures, and participant-set temperatures in the low-visibility conditions minus baseline temperatures. This marginal visibility effect is consistent with recent behavioral priming meta-analyses where most effects are small [46]. Thus, priming participants with high-fan visibility alone may not be sufficient to cue energy-efficient temperature adjustments in warm ambient temperatures.

Temperature selection was affected by the intervention interfaces, which combined fan feel and visibility parameters. Analyzing mean raw temperatures revealed a significant relationship between temperature adjustments and intervention interface. Participants’ mean raw temperatures were highest in the feel-fan high-visibility (FF + HV) condition (24.4 °C/76.0 °F), versus baseline (22.0 °C/71.7 °F), feel-fan low-visibility (FF + LV) condition (23.2 °C/73.7 °F), don't-feel-fan high-visibility (DFF + HV) condition (22.7 °C/72.8 °F), and don't-feel-fan low-visibility (DFF + LV) condition (22.4 °C/72.4 °F). These results support that increased air speed during thermostat interaction encourages users toward energy-efficient temperature adjustments.

A theme that emerged in the open-ended responses was visibility-induced frustration for the don't-feel-fan high-visibility (DFF + HV) condition. If high fan visibility alone primes people to feel cool, this condition would prompt higher temperature selection. Our results showed a marginal effect of visibility. Again, average temperature selections in the don't-feel-fan high-visibility (DFF + HV) condition (22.7 °C/72.8 °F) were lower (worse) than in the feel-fan high-visibility (FF + HV) condition (24.4 °C/76.0 °F) and feel-fan low-visibility (FF + LV) condition (23.2 °C/73.7 °F). Nine participants reported that high-fan visibility caused frustration (or a similar reaction). Seven of the nine participants also commented that visibility-induced frustration led to lower temperature selection. This effect is articulated by a participant as, “The promise of a breeze locked behind glass frustrated me, prompting me to select a lower temperature in hopes of getting relief from the heat.” Furthermore, one participant explained how the don't-feel-fan high-visibility (DFF + HV) interface affected their behavior as, “When I can see the fan clearly but cannot feel the air from the fan, I feel kind of frustrated (even more frustrated than when the fans are behind white bars when I cannot see it clearly) so I decreased the temperature …” Thus, despite quantitative analysis showing little effect of fan visibility alone on temperature selection, the lack of fan feel in the don't-feel-fan high-visibility (DFF + HV) condition frustrated some users. Therefore, fan visibility without fan feel should be avoided in actual interventions and designs.

This study also administered five questions to gauge knowledge regarding energy myths based on Pritoni et al.’s assessment of household behavior [36]. Current-study participants were divided on their understanding of heat delivery. About 52% of participants erroneously, and consistently, responded “True” to three questions explaining heat delivery through the gas-pedal metaphor. That is, participants believed that selecting significantly higher or lower temperatures would heat or cool them quicker, respectively. Overall, our study supports Pritoni et al.’s evaluation that household occupants hold misconceptions about thermostat control and energy use [36]. Thus, an incorrect conceptual model of heat delivery may be another barrier to energy-efficient temperature setting.

In hindsight, a study limitation is that fans were on the same settings in the feel-fan high-visibility (FF + HV) and feel-fan low-visibility (FF + LV) conditions; however, the grill cover in the low-visibility (LV) condition partially blocked air flow, reducing the fan’s effect. This difference in felt air speed between conditions (see Table 2) may have affected study results. Specifically, participants’ mean raw temperature setting was 1.25 °C/2.25 °F higher in the feel-fan high-visibility (FF + HV) condition than in the feel-fan low-visibility (FF + LV) condition. Because of the unintended difference in felt air speed, it is difficult to conclude whether it was this difference or the difference in fan visibility that led to the change in temperature selections. Future iterations should address this difference in felt air speed to more precisely study fan feel versus fan visibility on temperature selection. Furthermore, given a few participants’ thermal discomfort (i.e., they reported feeling cold) in the feel-fan (FF) conditions, reduced fan speed may achieve the same effect with less thermal discomfort.

Another study limitation is that participants’ behavior was measured over a period of minutes. To better understand the intervention’s effect, thermal selections over longer time scales (for example, throughout the day) and more varied conditions remain to be tested. In addition, the intervention was intended to provide brief thermal relief during user-thermostat interaction. Not tested, however, was how the fan’s absence post-interaction affects thermal comfort and acceptance.

The small and relatively homogeneous sample of university students makes it difficult to generalize the findings. The oldest two of 23 participants indicated they were in the 25–34-years-old age range. Future iterations should test the intervention’s efficacy on a more heterogeneous set of participants.

Finally, indoor temperatures during the study ranged from 24–27 °C (75.2–80.6 °F). While these ranges represent normal temperature fluctuations during the summer months at the study location, different indoor temperatures may reduce the ability to compare results between participants. Future studies should maintain control of the indoor temperature while participants are undergoing the experiment.

Implementing an energy-conserving intervention must consume less energy than the energy that the intervention is intended to conserve. Electromechanical fans alone deliver instant, targeted thermal relief, and generally consume much less energy than using A/C to cool an entire space [6]. Additionally, fans may be utilized to widen indoor temperature bands, further contributing to energy savings [6].

Of note, while current-study fans were placed directly above the thermostat to observe the effects of both fan feel and fan visibility, other configurations may also be effective. Fans integrated with the thermostat may consume less energy, as these fans can be turned on when user-thermostat interaction begins. Since fan visibility did not prove to be important, other fan locations are possible. For example, a standalone fan placed anywhere, provided it could be felt while setting the thermostat, may achieve comparable results without requiring a reconfigured thermostat.

Increasing global energy demand for A/C emphasizes the urgency associated with developing behavior-change interventions to reduce its consumption. The current study offers a potential strategy to address energy-inefficient temperature selection. Reducing energy use in heating, ventilation and air-conditioning via energy-conscious user behavior can lead to significant savings in resource consumption [47,48].

The current work examined how temperature decisions can be modified by altering the decision environment. Rather than trying to change people’s beliefs and values around energy conservation, modifying the environment (in which energy-consumption decisions are made) may be more effective. Specifically, this study demonstrated the effectiveness of using a fan during thermostat interaction to encourage energy-efficient temperature selection in warm ambient conditions. Of note, fan feel is more important than fan visibility. In fact, fan visibility without fan feel should be avoided.

Participants made temperature adjustments on actual thermostats, rather than responding to researcher questions, thus reducing the often-reported disconnect between intention and action [49]. Furthermore, awareness of current ambient room temperature mimics real-world conditions present during thermostat interaction. Combined, these study parameters led to a more realistic experimental scenario that is more likely to generalize outside the laboratory setting.

This work offers a behavior-change intervention to encourage energy consciousness toward a more sustainable future. While the study results are promising, more work is required to understand the intervention’s longer-term effects. Future work should also test the intervention in more naturalistic settings.

One of the United Nations’ sustainable development goals is to take urgent action to address climate change and its effects [50]. Climate-change-driven planet warming threatens human health as critical thermal limits of the body could be surpassed [51]. Ironically, overuse of A/C hastens overall planet warming, a problem the current work sought to address.

The authors gratefully acknowledge the support of the Social Sciences and Humanities Research Council of Canada (SSHRC), the Collaborative Specialization in Psychology, Psychiatry and Engineering (PsychEng), and the XSeed funding program at the University of Toronto.

There are no conflicts of interest.

This research was approved by the University of Toronto Institutional Review Board (Protocol No. 35708).

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.