6. EXPERIMENTS AND RESULTS
6.5 T HE A UDITORIUM E XPERIMENT
6.5.5 Conclusions from the Auditorium Scenario Experiment
This test aims to find if people in general, when inside an auditorium, tend to use the nearest exit or the same exit they use to get inside. Another important aspect is related to whether people exit selection depends on the relative location where they are sitting, particularly the ones further away from the entry they use.
Looking deeper to the results and comparing them with the help of the graphical plots (see Figure 6.3). The statistical analysis shows that, although apparently both the questionnaire and the SG have similar results, there is no direct association between them. And when analysing using other parameters, such as gender and fire expertise, no significant association is found either. This fact leads to conclude that using questionnaires or virtual environments might bias the results. Another possible explanation is that subjects tend to change their answers to the same question or scenario, leading to unpredictability of human behaviour and decision-making process.
Figure 6.3 - Graphical plot of the results for the Auditorium (questionnaire and SG)
But when considering globally the aggregate values, the results are quite similar, and a trend can be made from the results. In fact, subjects prefer the nearest exit (the worse case is 61%). And the SG has better results in this comparison with questionnaire when the player sits at top rows (77% versus 61% when get in using the top entry; 82% versus 65% when get is using the bottom entry). The other interesting aspect is that the bottom exit is chosen oftener than the one on the top (over 90% when sitting near the bottom; 57% of the subjects playing the game and 68% of the respondents when they are sitting at the top row).
These tendencies might help emergency planners and safety experts when designing evacuation strategies for spaces similar to this one. They can be used as well to improve evacuation simulators by incorporating intelligent agents mimicking behaviours elicited from real persons.
6.6 Summary
The experiments carried out and presented in this chapter allow some interesting conclusions to be drawn. 0% 20% 40% 60% 80% 100% 1 - Top entry / Top row 2 - Top entry / Bottom row 3- Bottom entry / bottom row 4 - Bottom entry / Top row 61% 4% 0% 65% 39% 96% 100% 35%
Auditorium scenario - Questionnaire
Top exit Bottom exit
0% 20% 40% 60% 80% 100% 1 - Top entry / Top row 2 - Top entry / Bottom row 3- Bottom entry / bottom row 4 - Bottom entry / Top row 77% 7% 5% 82% 23% 93% 95% 18% Auditorium scenario - Serious Game
First, the experimental setup designed for human behaviour elicitation using questionnaires and the SG concept, an instantiation of the SPEED methodology, has resulted in a collection of data that can be further used for improving existing pedestrian simulators. The main goal was to implement a set of experiments using SG in the specific domain of evacuation scenarios. These scenarios include correct alarm identification; the way-finding process when exiting from an unknown place, facing some adverse situations such as smoke, fire and people running on the opposite direction of the emergency signage; and at last which exit to choose when having to leave rapidly an auditorium: the nearest one or the same way used to get in. This preliminary set of experiments, designed with the help of an expert panel, and improved using a methodology based on the Delphi method, can be further expanded to include other scenarios and experiments.
Despite some differences found, among the responses using either the questionnaire or the SG, some tendencies can be found, such as the tendency that pedestrians show when preferring to go right than left, when no indication on the direction to take is available. The reaction of subjects when facing smoke or fire can give fire researchers valuable information. Whereas fire is impassable, smoke may or may not be. Even though the emergency signage points towards the smoke or fire direction, evacuees have to decide whether they follow that indication or go elsewhere. More interesting is the situation when people are running on the opposite direction of the emergency signs. And this methodology can be used for further test other scenarios and hypothesis.
Another conclusion from the analysis of the experiments carried out, is that subjects behave differently when answering a questionnaire or playing a game. The statistical analysis carried out did not find strong evidences of association between the results collected using the questionnaire and the SG. The immersion nature of the SG affects the perception of the scenario, and this might be one of the explanations for these results. It became clear that human behaviour is not constant and varies according to other aspects that these experiments were not able to show. Although aspects such as gender, fire training, and even psychological dimensions were explored, using statistical tests, the associations found are scarce and insufficient to support a sound theory. Due to the limited scope of the experiments designed, no further analyses were possible at this stage. However, these results are not as deceptive as they might seem at a first glance. They illustrate that the human decision making process is very complex and not as straightforward as expected in order to be used by pedestrian simulators. The randomness related to humans reactions and behaviours leads to the need of further research looking for other ways to elicit human behaviour.
The comparison of the results between the questionnaire and the SG, demonstrates that, at least for the experiments and sample presented, they should be taken cautiously because the shifts in the responses is too high and suggest that subjects do not always proceed in the same way, for there is a random aspect that still needs to be quantified.
Nevertheless, the confirmation of the unpredictability of humans’ response in such limited scenarios with a small number of possibilities of choice (binary: going up or down, left or right), can shed light upon the cautions that researchers must have when using pedestrian simulators to mimic human behaviours in scenarios, such as the ones presented. Some tendencies in the scenarios described previously may be used to improve existing simulators, since some real data was acquired and can be used for steering intelligent agents during their path towards an exit. Combining the elicited data with the PDA concept, as described thoroughly in chapter 4 (see Section 4.3), as conceived in the SPEED methodology, an iterative process can be implemented, using subjects performing the experiments to improve the collected data. This simulation technique, known as “human-in-the-loop”, is the basis to implement intelligent agents using the PDA concept, and the core of the SPEED.
These aspects are of paramount importance for both fire safety engineers and researchers. Forecasting how people will behave during the evacuation process might help stakeholders to take preventive actions in order to avoid risky situations, such as herding behaviours leading to clogging exits. The elicited knowledge on human behaviour can help architects and engineers at the design of safer buildings, by using the tendencies found to improve the layout of buildings. Other scenarios can me envisaged to meet particular situations, following the methodological approach described herein. A collection of scenarios typically present in evacuation situations can be then used for elicitation of behaviours, to further improve the behavioural data that is so much needed in this field of expertise.
Meanwhile, the games built for the aforementioned purpose can be used as well for training and education, an important aspect that is of great importance. Ultimately, the aim of this research is to find ways and strategies leading to safer behaviours that can save lives in the future.