Environmental Control Issues in a New Energy-Efficient Building

Abstract

The availability of adjustable lighting controls, with options other than “on” and “off,” have been associated with increased energy savings. In the present study, we describe the way customizable, networked lighting controls are actually used in a recently built research facility that received a formal designation for sustainable design (LEED Gold). In addition to usage patterns, we explored occupants’ knowledge of control operation as well as the expectancies of nonoccupants. We also identified a variety of features that, if redesigned in accordance with HF/E principles, might lead to reduced lighting usage.

Keywords

environmental design work space design control interface design lighting sustainability USGBC 2009 LEED Guideline EQc6.1

Even in LEED-qualified buildings, lighting controls can be confusing.

Digitally networked environmental systems have increased people’s options for controlling their interior space. Such innovations support, for example, passive lighting adjustments driven by daylight, and by occupancy sensors and new interaction modalities, such as gesture-based controls (Spaulding & Holt, 2012). In the present article, we consider some of the human factors/ergonomics (HF/E) design challenges posed by highly flexible environmental control systems, especially as they relate to the goal of decreased lighting usage by designers, building managers, and/or users.

Our approach began with a usability inspection of the system, focusing on control consistency, adequacy of feedback, and the visibility and discoverability of functions. We then observed the lighting usage, particularly the use of options other than “all on” and “all off” (i.e., dimming and sectional lighting). Then through semistructured interviews, we probed occupants’ usage strategies and knowledge about the controls. Finally, we tested nonoccupants’ expectancies of control mappings to determine if they matched the actual mappings. Even though we focused on lighting controls, we believe the HF/E problems identified through our usability inspections and postoccupancy research may generalize to other environmental control systems.

The Marksbury Building

When the Davis Marksbury Building (DMB; Figure 1) opened in 2011, it was the first building on the University of Kentucky campus in Lexington to receive a Gold certification in Leadership in Energy and Environmental Design (LEED) from the U.S. Green Building Council (USGBC). The certification was based on a variety of features, including the building’s use of solar energy and the presence of a shower for commuters who cycled or walked to work. One of the features that contributed to the LEED designation was having “a high level of lighting system control” (2009 LEED Credit EQc6.1; USGBC, 2013).

Figure 1.

The Davis Marksbury Building (DMB), showcasing its many windows for natural light and daylight harvesting.

DMB houses faculty, staff, and graduate students. Much of the research occurring in DMB focuses on the development of innovative visualization technologies that involve the use of large-format, multiprojector displays, making lighting control especially important. The building’s open floor plan with many team-oriented work spaces was designed to facilitate configurability and multidisciplinary collaboration. The functional flexibility of the spaces makes it challenging to design lighting control panels that are appropriate for every work space configuration, even within a single room, across time. Additionally, the lighting control design may need to overcome occupants’ perceptions of lack of control in these more open, public environments (McCarrey, Peterson, Edwards, & von Kulmiz, 1974).

Lighting System Description

In DMB, overhead electric lighting adjusts based on input from three sources: daylight sensors, vacancy sensors, and control panels. All three are networked and mediated through a central computer server with access restricted to campus facilities managers. The server will adjust lighting based on the amount of daylight (“daylight harvesting”), and it will turn off lights if the vacancy sensors do not detect movement. Occupants must interact with control panels to turn lights on. When a button is pressed on the control panel, a request is sent to the server, which in turn sends information to a particular set of lights to create a setting (“lighting scene”) within the constraints set by the daylight sensors. Scenes define which lights turn on and how bright they are. Thus, each button on a control panel can be programmed to activate a single light or a collection of lights at a particular intensity level.

Although scene customization and button remapping are possible, in practice, occupants of DMB are limited to the settings selected by campus building facilities managers. More features are described in Table 1, and an example work space and its lighting controls are described in Table 2.

Table 1.

Review of Usability Inspection of the Lighting System in the Davis Marksbury Building

Lighting System Feature	Design Heuristic	System Usability Problems	Suggested Usability Solutions
Many control panels of differing layout and functionality (see Figure 2)	Consistency	One work space controlled by panels with differing layouts (inconsistency) Similar layouts function differently in different work spaces (false consistency)	Map buttons to lighting usage goals for each area (e.g., instead of topmost button being “all on” for some spaces but not others, have it always be the scene using the least amount of light required for the task most frequently performed in that space)
Button press creates a particular setting (“lighting scene”) Buttons backlit for visibility in low light	Visibility	Limited visibility of functions and status No labels (although buttons can be purchased with engraved labels) No indication of overall system being controlled (i.e., fans, HVAC, electrical outlet)	Labeling strategies for partial scenes (e.g., labeled lab and cages in our example in Table 2) to draw attention to an otherwise unexplored control Haptic indicators (perhaps inexpensive adhesive “bump dots” for energy-efficient settings)
Mild audible and haptic “click” with a button press Pinpoint LED present on the buttons of some control panels Communication with server creates variable delays (up to 2 s) between the button press and the onset of the scene	Feedback	Delay in scene onset may cause users to believe that a button is inactive Uncertainty about which button actually activated the scene if multiple buttons pressed in quick succession (perhaps contributing to “superstitious” behavior)	Limit delay to less than 100 msec (Seow, 2008), or short enough to prevent additional button presses between control activation and lighting response Make other buttons inactive during the delay Provide immediate visual feedback to button press (e.g., immediate increase in the pinpoint LED on the control during the delay)
Scenes vary in terms of which lights are on and their intensity	Discoverability	Weak S-R compatibility of mapping control layout to the room layout Exacerbated by uncertainty of cause-effect relationship from the delay	Redesign shapes of buttons and control panels to communicate interactions redundantly (location and shape)

Note. HVAC = heating, ventilating, and air conditioning; S-R = stimulus-response.

Table 2.

Example of Control Panel Functions and Nonoccupant Expectancy Data

PHOTOGRAPH OF SCENE	DIAGRAM OF SCENE	DIAGRAM OF CONTROL PANEL (with expectancy data)
Five lighting scenes of a lab space similar to other spaces throughout DMB in size, function, lighting, and other environmental controls. The photo to the right shows the five-button control panel used for this and other spaces. On this control panel, each button has a pinpoint LED on the left side of the button and backlighting to aid in low-light visibility. When the button is pressed, the pinpoint LED will turn on or off, indicated by a mild audible and haptic “click.” After a delay, the lights will then change to the predetermined scene, pictured in the first column and diagrammed in the second column. The third column shows diagrams of the control panels with the results of an online survey of nonoccupant users, with participants’ expected control choices for each lighting scene superimposed on the diagram of the control panel. The dashed line indicates the correct button number corresponding to the scene of that row.

Note. DMB = Davis Marksbury Building.

Figure 2.

A sample of the many lighting control panel designs located throughout the building.

Lighting System Evaluation

Usability inspection

An initial usability inspection of the lighting controls throughout DMB was motivated by problems we encountered when using the control panels in our own laboratory space. We then inspected the lighting controls in other rooms and noted issues in the interactions. Our inspections took into consideration well-established usability principles relevant to environmental control systems, such as the importance of maintaining stimulus-response compatibility (e.g., Fitts & Seeger, 1953; Fitts & Deininger, 1954; Proctor & Vu, 2006) and conforming to user expectancies and stereotypes (Peacock & Schlegel, 2004; Lewis, 1986; and Jack, 1985) as well as common design heuristics.

Common design heuristics – such as consistency, visibility, feedback, and discoverability – become more important as lighting control designs shift from isolated switches and dimmer knobs to more complex environmental control systems, such as those we inspected in DMB. In fact, in describing these heuristics, Norman (1988) often used lighting controls as examples. Table 1 summarizes the specific problems we encountered based on these heuristics, along with suggested solutions that may lead to greater usability of controls similar to those in DMB. Note that some solutions may address more than one problem.

Observations of lighting usage

Although our inspections of DMB control panels clearly pointed out a number of violations of basic design heuristics, we also wanted to understand how DMB occupants actually use their lights. System usage data were collected manually by recording the lighting status of each occupied room – primarily offices and lab spaces – at the beginning, middle, and end of the 8:00 a.m. to 6:00 p.m. workday. Observations were made over a 5-day span in September 2011, approximately 6 months after DMB was occupied. There were no special events requiring unusual light usage taking place during the week of our observations (e.g., building tours or symposia). Of the 187 instances recorded, 110 instances (59%) had the lights “all on” and 64 instances (34%) had the lights “all off.” Partial lighting scenes were activated by occupants in the remaining 13 instances (7%).

These data indicate that at least some portion of those who might want to use the partial lighting scenes had discovered this functionality and chose to use it during the week of observations. The 7% utilization may represent a ceiling on the use of these settings, reflecting constraints imposed by task-specific lighting demands or simply by user preferences. However, it seems plausible that in at least some of the 59% of cases in which the lights were “all on,” one of the partial settings would have been more appropriate.

For example, some lab areas have sectional scenes that provide lighting to the lab itself and to the “cage” storage areas at the entrance to the lab (see Figure 3 and Table 2). An “all on” choice means that both lab and cage lighting are on, a situation that is, in our experience, rarely necessary; the “lab on” lighting scene would be adequate in most cases. However, of the 15 observed instances of these labs with cages, 10 (67%) of them had “all on,” 4 (27%) of them had “all dim,” and 1 (7%) had only “lab on.” Unfortunately, at the time of our observations, we did not also obtain information on actual cage use by other labs to know whether we can generalize from our own limited need for cage lighting.

Figure 3.

A view from inside a lab space in the Davis Marksbury Building with two different light settings, “all on” (left) and “lab on” (right). The lights at the end are largely to illuminate the cages across from the lab space and minimally affect the ambient light in the lab space. These images also illustrate one of the projection displays found throughout the building while also showing the open, group-oriented floor plan.

Occupant interviews

Although it was beneficial to observe actual lighting usage by occupants, it was also important to ask occupants directly about their understanding and use of the lighting system. Semistructured interviews were conducted at 10 months postoccupancy with 30 individuals who work in DMB. These interviews were conducted with individuals working in the same spaces that were the focus of our observational research; however, because some observations were made in common areas or shared work spaces, the individuals interviewed may have differed from those who were responsible for the light settings at the time of the observation.

The sample included graduate students, professional staff, and faculty, who were asked about their general knowledge of the lighting controls and how they used them. Respondents were allowed to view, but not interact with, the lighting controls during the interview. Twenty-five respondents (83%) knew that their controls had options other than “on” and “off.” Of these, 16 participants (53%) reported using the partial lighting scenes, of which 6 (20%) articulated a particular strategy: using low settings regularly (2 occupants, 7%) and adjusting settings depending on the context, such as outdoor brightness or presence of other people in the work space (4 occupants, 13%).

More detailed interviews were conducted with an additional 10 building occupants to evaluate their understanding of the specific functions of the controls in their work spaces. They were asked to describe the function of each button on the control panels in their work spaces (n = 16; some evaluated multiple panel designs depending on their work space). Of the 98 buttons about which occupants were questioned, 38 (39%) of the functions were accurately predicted. Of the 18 “all on”/“all off” buttons, 12 (67%) were accurately predicted by occupants.

Those who made the inaccurate predictions expected all buttons on the panel to control specific spatial sections, an interaction design seen in some control panels in the building. For the remaining 80 buttons (i.e., those for partial lighting scenes), occupants accurately predicted 25 (31%) of the functions. Occupants were more accurate when identifying the dimming functions (19 out of 39; 49%) than the spatial-section functions (6 out of 41; 15%). When they were asked to rate their confidence in their responses on a 5-point scale, the participants’ responses mirrored their performance, with a two-tailed paired-samples t test showing higher confidence, t(15) = 2.795, p = .014, d = 0.52, in their “all on”/“all off” responses (M = 3.25, SD = 1.46) than in their responses for the remaining buttons (M = 2.51, SD = 1.35).

The interview responses about usage of the partial lighting controls seem to contradict the data obtained from observations. Responses indicating knowledge of settings other than “all on”/“all off” are not surprising given that most panels included four or more buttons, a design that implies greater functionality than simply “on”/“off.” The higher percentage of individuals reporting use of the partial lighting controls compared with actual observed usage may reflect the longer-term perspective taken by the respondents (e.g., giving a yes response for infrequent usage). However, evidence from observations and interviews converged when we focused on relative knowledge and usage of the different scenes, with greater knowledge and usage of “all on”/“all off,” followed by “all dim” and, finally, the sectional scenes.

The most surprising finding, however, was that respondents failed to accurately identify the “all on”/“all off” buttons one third of the time. One explanation for this finding may relate to violations of expectancies of the spatial control mappings (explored further later); another is the delayed feedback from control actions (up to 2-s delay in lighting response) that may make it difficult to learn about individual control functions and may even lead to superstitious behaviors. One person, for example, reported that it was necessary to push all the buttons from bottom to top in quick succession to turn the lights on.

Nonoccupant expectancies

To see if the most typical default control mapping for lighting scenes in DMB lab spaces matched the expectancies of users unfamiliar with the DMB controls, 48 participants completed an online survey using Amazon Mechanical Turk. “MTurk” has been used to replicate results of in-person studies of expected control mappings (Sublette, Carswell, & Seidelman, 2012) as well as visualization designs (Heer & Bostock, 2010), so we used it as an inexpensive and convenient means of collecting data.

All participants were over the age of 18 (M = 34, SD = 11.59), accessed the questionnaire in the United States, and were paid for completing the survey. To reduce the likelihood of haphazard or random responding that could compromise data collected from a large, anonymous, and remote population of this sort, we sampled only participants from those members of the pool who had completed at least 95% of their previous tasks in a manner that was satisfactory to those posting MTurk work requests. As a further check for evidence of random responding, chi-square analyses were conducted, and we found that responses were significantly different from chance, χ2(4) > 20.94, significant at a = .05.

Participants were given a brief overview of DMB, including a photo of the work space with all the lights on (Table 2) and a diagram of the work space and the control panel. They were then shown diagrams of each of the possible lighting scenes and were asked to select which button on a five-button control panel should be pressed to activate each scene, including “all on,” “all off,” “all dim,” one section (the lab) on at the brightest setting, and another section (the cage) on at the brightest setting (Table 2). The presentation order of the scenes was randomized for each participant. Following the survey, participants were asked about their previous experience with similar systems. Ten participants (21%) reported having encountered such panels, and of these participants, none reported using one more frequently than once a month.

Forty of the 48 participants (83%) correctly paired two or fewer scenes with buttons. Only 1 (2%) accurately specified the function of all five buttons. As shown in Table 2, for each scene, there was a reliable preference for one of the five buttons for each control function. However, participants showed the strongest agreement for buttons that should result in the “all on” and “all off” scenes, which they believed should be controlled by the top and bottom buttons, respectively. This finding is not particularly surprising, as these mappings adhere to conventional U.S. stereotypes (Lewis, 1986). Even so, only 19 participants (40%) identified both the “all on” and “all off” scenes correctly.

When looking at participants’ expected mapping of the partial lighting scenes (e.g., only one section on or all lights dimmed), 6 participants (13%) identified at least two of the three scenes correctly (and only 1 participant identified all three). Responses appear to conform to a so-called gradation trend whereby participants believed that when moving from top to bottom, all the lights would be dimmed before breaking into the sections; 15 participants (31%) responded in this fashion. These results seem to indicate that even without changing the scenes themselves, the mappings of controls to scenes could be changed to make them agree with the expectations of more participants. However, we judge this solution is still far from ideal.

Flexible Environmental Controls Revisited

The current case study has focused on the usability characteristics and actual use of control designs for flexible lighting scenes in a LEED-certified workplace. Through usability inspections, we identified interaction design problems relating to consistency, feedback, visibility of status and function, and discoverability of functions. In addition, the expectations of nonoccupant users are violated in the current design, specifically for control of specialized partial lighting scenes (i.e., those other than “all on” and “all off”). Actual system users, when probed regarding their knowledge of these mappings, also showed discrepancies between actual and expected control functions. Surprisingly, “all on” and “all off” buttons were identified correctly only two thirds of the time. Knowledge of how to control more specialized partial lighting scenes – especially the sectional lighting scenes – was lower still.

Despite the usability problems we identified, we discovered that the more specialized partial lighting scenes were used 7% of the time in work spaces sampled three times daily during a 1-week period. From a sustainability and energy conservation perspective, the use of these scenes is important if they replace more energy-demanding settings (e.g., “all on”). We cannot say whether, in the present case, these more specialized scenes would be used more (therefore decreasing energy usage) if these lighting functions were easier to discover and activate. However, it certainly seems worthwhile to pursue these modifications given the many ways in which the usability of the current controls could be improved (Table 1).

Inexpensive postinstallation modifications, such as labeling (verbal or haptic) or changing the scenes to better match expected mappings, can and should be used for the sake of improved usability in its own right. Because “being green” is often associated with extra effort on the part of product users, leveraging usability to increase energy-conscious behaviors could be a welcome change of pace.

Conclusions and Broader Impact

Previous research (DiLouie, 2004) has shown reduced energy usage as a result of introducing controls that include both intermediate settings and controls for individual occupants, a design choice for which LEED Credit EQc6.1 is currently awarded. Although the USGBC requires having a “high level” of controllability, we argue that this may not translate into actual energy savings because of a variety of moderating factors, including occupants’ preferences and the task appropriateness of more energy-efficient light settings. Another potential moderator is the ease with which occupants can discover and use energy-efficient functionality.

Determining the impact of environmental control usability on actual energy usage demands the results of well-controlled field experiments that directly manipulate control panel design. It may prove to be the case that a focus on control usability is a practical way for HF/E professionals to contribute to sustainability initiatives. In the meantime, those individuals selecting control systems should also be made aware of the possible problems introduced by digitally networked lighting controls that may mitigate such systems’ potential benefits. Common design heuristics can and should be applied to such systems.

Footnotes

The authors have published a proceedings paper on this topic (Lee, Carswell, Seidelman, & Sublette, 2013), although it has been extensively rewritten and additional data have been collected for this manuscript. We thank Jay Brand for his insights on environmental design in group work spaces, and Rosanna Smith for her work in creating and directing a companion video (). Finally, we acknowledge Cathy Emery for her insights and the constructive comments of the reviewers.

Michael P. Lee is pursuing a PhD in experimental psychology at the University of Kentucky. He received his MS in experimental psychology and a certificate in human–technology interaction in 2014. His research interests include user-centered design in consumer products and how perceptual judgments of aesthetics and perceptual illusions relate to suit design.

C. Melody Carswell is a professor of psychology at the University of Kentucky, where she is also an associate director of the Center for Visualization and Virtual Environments. Her research interests include display design, usability assessment, and metacognitive influences on the use of cognitive tools.

Will Seidelman is a human factors and usability specialist at GE Appliances and a doctoral candidate in experimental psychology and human–technology interaction at the University of Kentucky. His research focuses on haptic data visualization, visual attention, and the impact of haptics on usability assessments.

Michelle Sublette is a doctoral candidate in experimental psychology at the University of Kentucky. She received her MA in experimental psychology from New Mexico State University in 2001 and has worked as a human factors engineer focusing on Web applications and consumer electronics. Her research interests include metacognition and users’ abilities to predict product ease of use.

References

Agarwal

Balaji

Gupta

Lyles

Wei

Weng

(2010). Occupancy-driven energy management for smart building automation. In Proceedings of the 2nd ACM Workshop on Embedded Sensing Systems for Energy-Efficiency in Building (pp. 1–6). New York, NY: Association for Computing Machinery.

Davis

M. C.

Leach

D. J.

Clegg

C. W.

(2011). The physical environment of the office: Contemporary and emerging issues. International Review of Industrial and Organizational Psychology, 26, 193–235.

DiLouie

(2004). Personal control: Boosting productivity, energy savings. Rosslyn, VA: Lighting Controls Association. Retrieved from http://lightingcontrolsassociation.org/personal-control-boosting-productivity-energy-savings/

Fitts

P. M.

Deininger

R. L.

(1954). S-R compatibility: Correspondence among paired elements within stimulus and response codes. Journal of Experimental Psychology, 48, 483–492.

Fitts

P. M.

Seeger

C. M.

(1953). S-R compatibility: Spatial characteristics of stimulus and response codes. Journal of Experimental Psychology, 46, 199–210.

Hedge

(2013). The importance of ergonomics in green design. In Proceedings of the Human Factors and Ergonomics Society 57th Annual Meeting (pp. 1061–1065). Santa Monica, CA: Human Factors and Ergonomics Society.

Heer

Bostock

(2010). Crowdsourcing graphical perception: Using Mechanical Turk to assess visualization design. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 203–212). New York, NY: Association for Computing Machinery.

Jack

D. D.

(1985). Rocker switch tactile coding and direction of motion stereotypes. In Proceedings of the Human Factors Society 29th Annual Meeting (pp. 437–441). Santa Monica, CA: Human Factors and Ergonomics Society.

Lee

Carswell

C. M.

Seidelman

Sublette

(2013). Green expectations: The story of a customizable lighting control panel designed to reduce energy use. In Proceedings of the Human Factors and Ergonomics Society 57th Annual Meeting (pp. 1353–1357). Santa Monica, CA: Human Factors and Ergonomics Society.

10.

Lewis

J. R.

(1986). Power switches: Some user expectations and preferences. In Proceedings of the Human Factors Society 30th Annual Meeting (pp. 895–899). Santa Monica, CA: Human Factors and Ergonomics Society.

11.

McCarrey

M. W.

Peterson

Edwards

von Kulmiz

(1974). Landscape office attitudes: Reflections of perceived degree of control over transactions with the environment. Journal of Applied Psychology, 59, 401–403.

12.

Norman

D. A.

(1988). The design of everyday things. New York, NY: Basic Books.

13.

O’Neil

M. J.

(1994). Work space adjustability, storage, and enclosure as predictors of employee reactions and performance. Environment and Behavior, 26, 504–526.

14.

Peacock

Schlegel

(2004). Expectancy and compatibility. Ergonomics in Design, 12(2), 4–31.

15.

Proctor

R. W.

K. P. L.

(2006). Stimulus-response compatibility principles: Data, theory, and application. Boca Raton, FL: CRC Press.

16.

Sellers

B. C.

Fiore

S. M.

(2013). Sustainable and user-centered: Applying human factors solution towards improving the effectiveness of “green” buildings. In Proceedings of the Human Factors and Ergonomics Society 57th Annual Meeting (pp. 550–554). Santa Monica, CA: Human Factors and Ergonomics Society.

17.

Seow

S. C.

(2008). Designing and engineering time: The psychology of time perception in software. Boston, MA: Addison-Wesley Professional.

18.

Spaulding

J. M.

Holt

(2012). Exploration of gesture-based control for tunable solid state lighting applications. In Proceedings of the Human Factors and Ergonomics Society 56th Annual Meeting (pp. 1902–1906). Santa Monica, CA: Human Factors and Ergonomics Society.

19.

Sublette

M. A.

Carswell

C. M.

Seidelman

(2012). A first look at identifying strategies people use in making predictions about task demand. In Proceedings of the Human Factors and Ergonomics Society 56th Annual Meeting (pp. 1867–1871). Santa Monica, CA: Human Factors and Ergonomics Society.

20.

U.S. Green Building Council. (2013). Controllability of systems: Lighting. Retrieved from http://www.usgbc.org/node/1731102?view=language