Experimental procedure
The research procedures reported in this paper are in compliance with the Swiss Federal Act on Research involving Human Beings (et al.) and Nanyang Technological University’s policy on Research Involving Human Beings. They were approved and certified by the ETH Zurich Ethics Commission. EK 2018-194 (January 18, 2019); reference no. IRB-2019-04-205, 23 May 2019.
This study has been pre-registered in order to allow for data analysis.https://osf.io/q5hnk/).
The experiment was performed in the courtyard of National Institute of Education, Nanyang Technological University. It took place between June and December 2019. The sample population includes staff, students, visitors and staff from the University. Posters were placed on campus advertising the study to recruit participants. Eligibility criteria included the following: a minimum age of 21 to 55 years, a physical condition that allows for walking in outdoor environments, and no medical conditions that prevent you from long-term walking in outdoor spaces.
A 1.5 hour time slot was reserved for each experimental session. Participants arrived at a pre-described instruction spot, which was located outdoors and protected from direct sun. After reading the information sheet and giving their informed consent, participants were asked to complete the pre-experiment questionnaire (can be found at pre-registration). It contained questions about the participant’s socio-demographic characteristics, attitude towards Singapore and lifestyle. Each participant was given a physiological wearable sensor (wristband Empatica E4) for monitoring their physiological status. Data not reported in this paper. After reading a short story, the participant was instructed on how to conduct the experiment. Once the participant had confirmed their readiness, an action camera was placed on her/his chest to record the participants’ decisions and environmental events. Start and End of each trial.
The participant was directed towards the beginning of the experiment, and then informed again about the procedure. The participant was required to make choices. These choices were made in a choice booklet (see Appendix B for the choice set booklets). Trial 0 was used to explore the environment. The participant was asked walk around the lawn until they reached the target. The following trials (trials 0 to 13) asked participants to reach their target using the routes indicated by the arrows in a booklet. The origin of the current trial was the target of the previous trial. The participant was asked visually to identify the target and the path options in each environment at each point. Next, the participant was asked if he/she preferred to make decisions based upon his/her preferences. There was no correct or wrong choice. The participant was informed that he/she wasn’t being tested for speed of completion. Participants were given a water bottle to prevent dehydration. They were asked to use it at their discretion. The experimenter left the participant to complete the tests and was watching from a distance, without giving any additional instructions. Participants were asked if they needed any assistance by standing still and raising a hand. Participants who needed intervention from an experimenter due to environmental conditions (rain), confused paths, or other reasons were removed from the analysis reported here. The researcher met the participant after the final trial was completed and took him back to the instruction site, where sensors had been removed. The researcher asked the participant to complete the post-participation survey. This included questions about their motivations, feelings, perceptions, and acceptance of the chosen path. After completing the experimental procedures, participants were debriefed. They were then compensated with 20 Singapore Dollars in cash. To minimize bias in participants’ behavior, neither the recruitment nor the instruction materials contained an explicit formulation of this research question. Instead, the study’s purpose was stated as follows: To investigate navigational attributes or features of outdoor ambulation within a variety of Singapore environments. We will also be focusing on the most influential environmental factors.
The experiment involved 74 participants. 4 participants had missing data or could fail to complete the experiment due rain. 9 participants took unspecified paths or had navigational problems that required intervention by the experimenter. 2 participants self-corrected their incorrect paths, but they were still excluded from this study’s analysis.
Data processing
Raw datasets resulting in experiments include video taken on the participant’s chest with the camera mounted, physiological signals from the Empatica E4, responses from pre- and post-experimental surveys, and microclimate data from two Kestrel 5400 weather stations that were placed in the sun and shade. The data was taken from video recordings and used in the current paper.
Student research assistants processed the video-recordings of each participant according to a protocol. They entered all events from the video into an Excel spreadsheet with a predefined structure. The times on the wristband, experimenters smartphone and video were synchronized using the synchronization events from the video with cameras. Participants coded the following events:
-
1.
Decision event: Start by a participant in a particular trial.
-
2.
End of trial: Participant steps on the target during the current trial.
-
3.
Sun presence: An alteration of the sun in one state to the other. The following states are:
-
(a)
Full sun (sharp shadows can be seen on the ground);
-
(b)
Cloudy sun (soft shadows visible on the ground)
-
(c)
There is no sun (sun behind the clouds and there are no shadows visible on the ground).
-
(a)
-
4.
Sun exposure event: Alteration of sun exposure from one state to the next. The following states are:
-
(a)
No shade (participant walks on a surface that is directly exposed to sunlight).
-
(b)
Shade from trees (participant walks on the ground covered by the shadow cast the tree).
-
(c)
Building shade (participant walks across the surface covered in shadow by the building).
-
(a)
-
5.
Water intake event: It appears at recording that the participant is drinking water.
The following attributes are recorded for each event:
-
1.
Event code
-
2.
Time of the event
-
3.
XY coordinates of the approximate location for an event can be accessed by clicking in the realistic space-sun position model (described in next section);
-
4.
Only for decision events: An indicator of whether option A was chosen by the participant.
All trial events, including the end of trials, were cross-coded and verified for agreement by two student research assistants. Data coding disputes (events differing in decision labels, in sun presence, or in start or ending time by more than five seconds) were solved by a third party (experimenter).
In the current study, events data was used to provide information on participants’ decisions and the presence of the sunlight at the moment of decision (determining if decision is treated as treatment one). The timing information of decision events was used to calculate the sun-shade composition for the path options. This was done by adjusting the sun position in a model described in the next section.
Events that diverge from the standard experimental procedure (e.g. Intervention of participant or experimenter making a shortcut) or potentially ambiguous situations (e.g. Uncertainty regarding the sun’s presence were recorded by data coders in a notes file. The experimenter reviewed the notes file and made decisions about how to treat the participants data (e.g. Rejection from the analysis
Calculation of sun-shade composition for the path options
The 3D model was imported into Unity 3D and visually validated to reproduce the shading of the walking routes. (See Appendix B for a comparison between video shots and reproduction in the model).
As the walking surfaces were covered by polygons, all the paths were included in the 3D model. The paths that run along the building’s 6 m width were divided into 5 strips, each measuring 1.2 meters in width. Each path option was assigned 5 polygons (path strips). The time information from event files was used to adjust sun position in the model when calculating the sun-shade composition for each of the path options during a particular trial. The rays that cover each polygon of a particular path option (on an grid of 0.1) are then calculated. (times)The rays (0.1 m) were directed towards the sun. The intersection of each ray with a tree or building was detected. The fractions of rays hitting nothing, hitting a tree, and hitting a building were then considered to be the fractions sun, shade, and shade on a specific path option polygon. The intersection of the rays and the tree was determined as their intersection with the convex shell around the tree crone. Therefore, the Unity 3D tree shades and those used in calculation of the sun-shade composition of each path option may differ slightly. Each path option’s polygon (strip), with the lowest fraction of sun was taken to be representative of its overall sun-shade content. Building shade that was less than 15% (i.e. Participants were not allowed to consider paths that had less than 0.9m of the wide paths between the buildings. Path options with such shading patterns were considered insufficient and deemed unsuitable for consideration.
The sum of the lengths for each option was used to calculate the length. These were measured using a laser distance meter. One researcher operated the meter, while the other held a mark marking the spot where the laser was shot. A length of each segment of path was determined by taking three repeat measurements. Also, the distance between each tree’s selected anchor point was measured. 3D tropical plants were placed at the appropriate locations in 3D models. The measurements of each tree were adjusted to closely match shading recorded by the chest mounted action camera during the experiment in the two seasons. Refer to Supplementary materials to see a comparison of the 3D model and camera shots.
The length of sunlit stretch, tree shade, and building shade along the option was calculated by multiplying length of path by fraction of each component (calculations are described in the paragraph).
Hierarchical model for the choices
The following cost function is defined for the path option:
$$begin{aligned} c^{(A)}_{ji} = beta _{j}[a^{text {sun}}_{ji} + (1-rho )a^{text {tree}}_{ji}]+ a^{text {shade}}_{ji} + rho a^{text {tree}}_{ji}, end{aligned}$$
(1)
Where (a^{text {sun}}_{ji}), (a^{text {tree}}_{ji})And (a^{text {shade}}_{ji})These are the metric distances in sunlight, tree shade and building shade for each option of path ATrial iParticipant presented j. (beta _j > 0)This is the participant-specific distance-inflating coefficient (cost fact) for walking under the sun. (rho in [0, 1])The parameter of shade intensity (relief), associated with tree shade that is common to all participants. Consider a similar definition for the cost per option B ((c^{(B)}_{ji})), the difference in the option costs is:
$$begin{aligned} Delta c_{ji} = c^{(A)}_{ji} – c^{(B)}_{ji}. end{aligned}$$
(2)
The likelihood of choosing the path option A (p(y_{ji}=1))It is modelled using logistic function, which is widely used in dichotomous selection models37:
$$begin{aligned} p(y_{ji}=1|Delta c_{ji};beta _j, rho , tau _k) = frac