Manipulating Emotional Arousal with External Stimuli: Study on Background Lighting, Study notes of Experimental Design

Download Manipulating Emotional Arousal with External Stimuli: Study on Background Lighting and more Study notes Experimental Design in PDF only on Docsity! Master thesis MANIPULATING EMOTIONAL AROUSAL USING EXTERNAL SECONDARY STIMULI August 16th 2019 Jacob Brok-Jørgensen Engineering Psychology, Group 1089a Electronic Systems, Aalborg University 19gr1089a@es.aau.dk Contents 1 Introduction 1 1.1 Physiological factors of emotion . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.2 Theories of emotion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 James-Lang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Cannon-Bard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Two-Factor Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Appraisal Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3 Research Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2 Method 7 2.1 Background Stimuli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Main Stimuli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.4 Experimental design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3 Results 15 4 Analysis 17 5 Discussion 21 5.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 5.2 The experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 5.3 Unintended factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 6 Conclusion 25 Bibliography 27 Appendices 29 A Timeseries data 31 B Electronic Appendix 37 C Electronic Appendix 39 D Electronic Appendix 41 E Electronic Appendix 43 Introduction 1 Any stimuli has the potential to change our emotional state, almost everyone has experienced this when listning to music, watching a movie, or reading a book. Stimuli outside of our selves have a huge capacity to influence how we feel. Our understanding of the way emotional changes take place has changed over time, some theories assert that emotions are less discrete categories but rather a product of physiological and cognitive processes, responding to a stimuli. This idea that emotions are essentially a product of several different factors is teased all the way back in Wundt and Judd’s Outlines of psychology (Pages 358-368). Wundt identified three main factors, Quality, Intensity and Form of occurrence, as the building blocks of any emotion, and used these to describe the differences between the more well- know categorical emotions(fear, anger, joy, exitment, etc.). Joy and rage for instance exhibit widely different affective qualities but, when it comes to intensity and form of occurrence, manifest in quite similar fashion. Wundt also details that there is an inherent linguistic limitation when trying to describe emotion. This is exemplified by the fact that anger, while remaining clearly identifiable as such can have a variety of forms of occurrence, just as surprise can have both a pleasant and unpleasant quality. This idea of a multidimensional model describing emotion has changed over time, with Russell (1980) proposing a circumplex model, shown in Figure 1.1 , based on multiple analysis methods like principal component analysis and multidimensional scaling, resulting in a model with two main dimensions interpreted as valence and arousal. These describe the pleasantness and energy of an emotion, for instance terror is very unpleasant and has high energy, while contentment is pleasant and has low energy. 1 Group 1089a 1. Introduction James-Lang According to the James-Lang theory, the physiological arousal respons contains all the information needed to produce an emotion and no cognition/consideration of the initiating event is necessary for a person to experience and emotion, the sequence of the theory is shown in Figure 1.3. Figure 1.3. Shows the basic flow of an emotion response according to the James-Lang Theory of emotion, inspired by KhanAcademy (2019). An event leads to physiological arousal, which is then interpreted, leading to an emotion. In order to show this an experiment was conducted by Ax, A. F. (1953). The physiological differentiation between fear and anger in humans. Psychosomatic medicine, 15 (5), 433–442 in which participants were either angered or frightened, and a clear difference was pressent in the physiological response despite the experience of either experiment was quite similar in many respects. Cannon-Bard This theory is a response to certain inadequacies of the James-Lang theory. Through Animal experiments Cannon (1927) has show that severing connections between the brain and various parts of the sympathetic ANS resulted in no change to the animals ability to show anger and rage in a behavioural manner. This despite not showing any of the usual physiological symptoms of anger and rage. Figure 1.4. Shows the basic flow of an emotion response according to the Cannon-Bard Theory of emotion, inspired by KhanAcademy (2019). An event leads to physiological arousal and an emotion simultaneously. An event leads to a simultaneous physiological response and emotional experience, as such the emotional context of the experience is in no way influenced by any physiological changes. Two-Factor Theory Schachter and Singer (1962) created the Two-Factor Theory to account for evidence supporting both the James-Lang and Cannon-Bard theory of emotion. By inducing a physiological arousal response through injections of epinephrin, and using social stimuli to provoke anger and elation in subjects, Schachter was able to show that when a person is not able to explain physiological arousal these feed into what ever emotion an event 4 1.3. Research Question Aalborg University might elicit. However, when an explanation for the physiological symptoms is pressent the emotional response is unaffected. Figure 1.5. Shows the basic flow of an emotion response according to the Two-Factor Theory of emotion, inspired by KhanAcademy (2019). An event leads to physiological arousal, which is then interpreted in the context of the event, leading to an emotion. An event leads to physiological arousal, this arousal response is then interpreted in the context of the event, taking into account past experiences, prior mental state and the larger context of the event. Only after processing all elements can a person actually feel the emotion. Appraisal Theory The goals of Appraisal Theory were to give a more comprehensive explanation of the processes that produce emotion, in order to account for several unaddressed weaknesses with alternative theories. Figure 1.6. Shows the basic flow of an emotion response according to the Two-Factor Theory of emotion, inspired by KhanAcademy (2019). An event is appraised, and labeled, based on this label physiological arousal and an emotion is felt. The event is appraised and receives a rudimentary emotional label, based on this label a physiological arousal response is induced and an emotion is felt. Scherer, Schorr, and Johnstone (2001) lists several weaknesses with alternative theories. 1) Many theories rely heavily on dimensional models of emotion which do not account for evidence supporting the idea of distinct emotions. 2) A single event can evoke different emotions in different people, 3) just like different events can evoke the same emotion. 4) Emotions are more often than not appropriate to a given event like the calming nature of sadness following unavoidable tragedy, where the very active nature of anger or rage would have been a wast of energy. 1.3 Research Question While the scientific community has yet to agree on a single definitive theory or model to describe emotion, it seems clear that, among other things, a persons physiological arousal is central and of great importance to the process. Prolonged exposure to stimuli, or stressors, can lead to chronic stress, this in turn can have detrimental effects on a persons ability to solve problems, and can elevate the risk of acquiring a wide range of ailments, ranging from depressions to cancer. 5 Group 1089a 1. Introduction There are countless ways to influence a persons emotional state. Several of these methods are complex and demand a certain amount of attention, while others, like various environmental changes can be less demanding of a persons attention. As a first step in creating a system to potentially help people alleviate stress the following research question is asked. What are the effects of a subtile background stimuli on a persons physiological arousal, as a responce to a main stimuli, and how does this effect performance? The goal with this question is not to solve stress for everyone but the investigate the viability of a tool to help people who suffer from chronic stress by artificially lowering the physiological response to a stressor using a subtil background stimuli. 6 2.3. Measurement Aalborg University Time C ha lle ng e Le ve l Figure 2.2. An illustration of a fictive game of Tetris in terms of challenge level over time The version of Tetris used for the test is found at n-blox.com. The game is run in full screen, with the sound off, and an adblocker to remove sidebar advertisements. A screenshot from the game is shown of Figure 2.3 Figure 2.3. The version of Tetris used for the test, found at n-blox.com 2.3 Measurement Finally, a method for acquiring a measure of physiological arousal throughout the experiment is needed. This method needs to provide a continuous measure over the entire duration of the experiment, as it is not known where the effect of a background stimuli could be most prominent. The method also needs to run passively, meaning it should require little to no active input from the subject or facilitator, this is to avoid interrupting the participant while they are actively engaged with the main task. Many self report methods (e.g. Self Assessment Manikin(SAM)) require a high level of self insight as well as familiarity with the scales and terminology which is not alway pressent with participants. The method should to be as little intrusive as possible. that is to say, it should require as few sensors and electrodes as possible, as to minimise the risk of influencing the data by for example intimidating the subject. Finally, the subject needs to be able to actually engage with the main task of playing Tetris while having their physiological arousal measured. 9 Group 1089a 2. Method For this experiment the best suited solution is using galvanic skin response(GSR) electrodes to measure skin conductivity, and use that as an estimate of physiological arousal. In praxis that means using a Shimmer GSR+ with two reusable snap electrodes for the making the actual measurements, combined with the ConsensysBasic software used for logging and exporting the GSR data. For the best results Shimmer (2018) recommends attaching the electrodes to the index- and middle-finger, on the palmar surface of the finger. However, through pilot tests it was found that the reusable velcro straps which are used to attach the electrodes tended to rub against each other causing noise in the GSR data. To avoid this the electrodes were placed on the palmar surfaces of the index- and ring-finger. This way the velcro straps can never rub against each other on accident. 2.4 Experimental design Setup The subject sits in a chair in-front of a monitor, mouse and keyboard. The shimmer GSR+ is docked next to the monitor and equipped with two electrodes with reusable velcro mounting straps. The monitor, mouse, keyboard and shimmer dock is connected to a laptop placed next to the monitor, of which the entire experiment is run. The Tetris game is shown on the monitor while the laptop runs the ConsensysBasic software, controlling the Shimmer GSR+. The GSR electrodes are attached to the subjects left hand and they control the game with their right hand exclusively, to avoid any noise or interference on the GSR reading. Gameplay and sound is recorded with the Windows 10 Game Bar screen recording tool, to allow synchronisation between the GSR timeseries data and video of the Tetris gameplay. Three Philips Hue lightbulbs are used, one is placed behind the monitor, the other two are hung from the ceiling behind the subject, one in each corner, about 50 cm from the ceiling and either wall. A sketch of the experimental setup is shown on Figure 2.4. 10 2.4. Experimental design Aalborg University Figure 2.4. Sketch of the experimental setup. The three red dots represent the locations of the three Philips Hue Lightbulbs. The blue dot shows the location of the subject and the green dot the location of the facilitator. The placement of the lightbulbs is shown on Figure 2.5. Figure 2.6 is an image of the entire experimental setup with the subject sitting in the chair on the right and the facilitator in the chair on the right. Figure 2.5. Left: The placement of the two Philips Hue lightbulbs hanging from the celling. Right: The placement of the lightbulb place behind the monitor. 11 Results 3 During the execution of the experiment the GSR and video recordings for one subject was disrupted, so an extra subject was added to replace the corrupted data. Only after the completion of the experiments was it discovered that another three recordings were corrupted leaving 15 complete datasets. All 15 subjects are male, aged 21-26 (mean=22.9) all recruited from Aalborg University Fredrik Bajers Vej 7. Each dataset consists of timeseries data of skin conductance measured in µS at 10 hz each with a duration of approximately 15-20 minutes. 6 datapoints were identified as outliers resulting from the automatic range setting of the Shimmer GSR+, and these have been removed from the dataset, leaving a total of 151,026 data points spread over the 15 subjects. An example of the timeseries data is visualised in Figure 3.1, where the skin conductance measured in µS is plotted as a function of time. A clear difference between the different lighting settings can be seen for subject 13, just from looking at the visualisation. Here the third run with the Energize lighting mode is clearly higher than the other two modes. Figure 3.1. Subject 13 timeseries data The complete data files are located in Appendix B , and are all visualised similarly to Figure 3.1 in Appendix A. Additionally each plot is separated into three sections, in accordance with the different modes of lighting that were used. 15 Aalborg University 6.8 7.0 7.2 7.4 B right_baseline B right C oncentrate_baseline C oncentrate E nergize_baseline E nergize Lighting C on du ct an ce Figure 4.4. Plot of means, now including base line data Lighting count mean sd median IQR Bright baseline 8995 6.71 4.95 5.86 8.16 Bright 42977 6.97 5.23 5.77 8.87 Concentrate baseline 8996 7.26 6.39 3.84 9.52 Concentrate 41420 7.36 6.68 3.96 9.04 Energize baseline 8996 6.83 5.34 3.83 8.48 Energize 39642 6.78 5.68 3.92 7.95 Table 4.2. A table containing a summary of the Skin conductance data for the three lighting modes, now including base line data Again an H0 hypothesis is defined, There is no difference in skin conductance between baseline and experimental under any mode of lighting, and another repeated measures ANOVA is performed. Mauchly’s Test for sphericity indicates that the assumption has been violated, p<.05. A Greenhouse-Geisser sphericity corrected test (ε = 0.362) shows no significant difference between baseline and experimental skin conductivity, F(1.809, 25.333) = 0.280, p = 0.737 and ges = 0.001. Finally subjects’ performance in the game under each mode of lighting can be investigated. A reasonable measure of performance is created by consolidating the number of lines the subject has completed in a given run a the total playtime of that run, in accordance with Equation 4.1 Performance score = Number of completed lines Total run duration × 10 (4.1) Using this performance score, a new plot of means, Figure 4.5, and summary, Table 4.3, is created. These show no apparent difference in performance under the different modes of lighting. 19 Group 1089a 4. Analysis 1.1 1.2 1.3 Bright Concentrate Energize Lighting P er fo rm an ce Figure 4.5. Plot of means of the performance scores. Lighting count mean sd median IQR Bright 15 1.19 0.471 1.15 0.688 Concentrate 15 1.15 0.418 1.21 0.492 Energize 15 1.16 0.394 1.21 0.410 Table 4.3. A table containing a summary of the Skin conductance data for the three lighting modes One last anova is performed with the H0 hypothesis defined as ’There is no difference in performance score under the different modes of lighting.’. Mauchly’s Test for Sphericity confirms the assumption with p = 0.464. The test show no significant difference in performance under the different modes of lighting, F(2, 28) = 0.0.114, p = 0.893 and ges = 0.002. 20 Discussion 5 In this section the results and findings of the analysis are discussed in a larger context. The experimental method is also discussed, along with some unintended factors which in hindsight might have influenced the results. 5.1 Results In the analysis three different facits of the results are explored, the overall level of skin conductivity, the change in skin conductivity, using a baseline, and the overall performance during the different runs. The highest mean and median values of overall skin conductivity conflict, with the highest mean value belonging to Concentrate while Bright has the highest median value. The differences between the different means and medians is also quite small, sugesting that the difference is just a result of random variance. This is supported by the repeated measures ANOVA which finds no significant difference in skin conductivity across the different modes of lighting. This sentiment is mirror in the analysis of both the baseline and performance measures. with the differences being very small and the ANOVA’s finding no significant difference in skin conductivity or performance. Individual differences in normal skin conductivity, means that there is a huge amount af variation in the overall skin conductivity during the experiment. This could be solved by normalising the data prior to the statistical analysis, or by utilising a larger sample size. Kang et al. (2019) points to the contradictory findings on the effects of lighting on physiological arousal, which might help explain why the lighting, in this experiment, has had little to no effect on the subjects skin conductivity. During the analysis, a measure of performance was established as a function of time spend playing and the total number of lines cleared. As the speed of the game slowly accelerates over time, the speed at which the player is able to place new pieces also accelerates. This together with the fact that participants were able to pick their own starting level could lead to the measure of performance being poorly balanced, and might have favoured players who started below or above their skill level. The game is also based on some amount of randomness and as such the performance score is somehow dependant on how lucky the player was in a given run. 21 Conclusion 6 This thesis attempts to investigate the effects of a subtile background stimuli, on a physiological arousal response to a main stimuli, in accordance with the research question detailed in Section 1.3 (Research Question). In pursuit of this, 15 male students at Aalborg University has played the game Tetris undet three different modes of lighting, while having their skin conductivity measured using a Shimmer GSR+ sensor. The lighting was supplied using three Philips Hue v1.0 White and Color Ambiance lightbulbs, and three different standard lighting modes from the Philips Hue companion application were used, Bright, Concentrate and Energize. In a repeated measures one-way ANOVA, no significant difference was found in the over all skin conductivity between any of the three modes of lighting, p = 0.778. A rudimentary baseline was established for each subject under each mode of lighting, and still no significant difference was found in the level of skin conductivity, p = 0.737. Finally, a basic measure of performance was established, as a function of the time played and the number lines cleared, again a repeated measures one-way ANOVA found no significant difference in the performance between the different modes of lighting, p = 0.893. Via the experiments conducted in the thesis, it was not possible to show what, if any effect a subtile background stimuli has on the physiological arousal response to a main stimuli. 25 Appendices 29 Timeseries data A Timeseries data for each subject. The vertical grey lines inclose the first run with Lighting = Concentrate, the blue lines, Lighting = Energize and the orange, Lighting = Bright Figure A.1. Subject 1 timeseries data Figure A.2. Subject 2 timeseries data 31 Group 1089a A. Timeseries data Figure A.9. Subject 9 timeseries data Figure A.10. Subject 10 timeseries data Figure A.11. Subject 11 timeseries data 34 Aalborg University Figure A.12. Subject 12 timeseries data Figure A.13. Subject 13 timeseries data Figure A.14. Subject 14 timeseries data 35 Group 1089a A. Timeseries data Figure A.15. Subject 15 timeseries data 36 Electronic Appendix C Timestamps for synchronising GSR data contained in csv-file at path: Hand-in/syncpoints.csv 39 Electronic Appendix D Matlab script used to rearrange the GSR data into a file for further data analysis, as well as for plotting the timeseries data for fro all 15 subjects. The Matlab script imports the datafiles available in Appendix B. The script is located at: Hand-in/matlabScript.m 41