Effort-based decision making in
Virtual Reality
The effect of task realism on human effort-based decision
making
Abstract
The current study is one of the first to explore the changes in human effort-based decision making when participants were confronted with different intensities of experimental realism. The author hypothesizes an increase of willingness to exert effort for a certain reward when more physiologically arousing environments are present. A sample of 51 students were asked to perform a physically effortful task in three different conditions. The conditions consisted of a baseline condition, similar to standard neuropsychological experiments using just basic figures, a condition with enhanced reward realism, and a condition with enhanced effort realism. An electrocardiogram was used to measure heart rate during the task, this way differences in arousal between the conditions could be measured. Every participant received all three conditions, measurements of heart rate, heart rate variability, and effort were compared between them. No significant results were found; the hypotheses could therefore not be confirmed. However, because of the complicated design for the first study in this area of knowledge there are arguably multiple errors in reliability and validity. Because of this it is advisable to conduct further research on the subject to make more definite statements.
Bachelor Thesis
University of Amsterdam Author: Hidde Pielage Student Number: 10778977
Introduction
The study of humans is one with many obstacles, it is nearly impossible to accurately measure psychological aspects of human subjects. During collection and analysis of data from humans there are many pitfalls to avoid for the sake of useful results. One of these problems comes from the many confounding variables to account for when studying the species in its natural habitat. This led modern-day neuropsychological studies to mostly avoiding natural environments and therefor reduce the number of confounding variables. However, by assessing participants in a less natural environment like a lab, their responses also become less ecologically valid or ‘natural’. This ecological validity and experimental control tradeoff is a problem in most modern neuropsychological research (de Kort, Kooijman & Schuurmans, 2003). Since ‘daily life’ is full of confounding factors and lab environments are a bad representation of ‘daily life’, researchers have to make a choice. They either choose high experimental control at the expense of ecological validity, or high ecological validity at the expense of experimental control. A possible solution is the use of virtual reality (VR) technology in empirical neuropsychological studies. VR could enhance ecological validity, while also maintaining high experimental control. The current study focuses on the implication of VR technology in empirical neuropsychological studies by researching if an increase in ecological validity of the environment will also increase the amount effort participants put in a task due to higher states of arousal.
VR is the term to describe a form of simulation that, through sensory input, creates a feeling of physical presence of the user in a virtual environment. The principle of stereoscopy forms the base of modern VR technology. Stereoscopy was discovered in the late 1860’s, allowing people to see ‘three dimensional’ pictures for the first time. Stereoscopy creates an illusion of depth by presenting slightly different perspectives of the same view to both eyes, this technique mimics the offset in perspective that the distance between our eyes naturally produces. This finding is the foundation of VR as we know it now. In the 1990’s VR technology was highly anticipated, but the products failed to live up to the expectations of the users. However, recent technological advancements have paved a path for VR technology to rise
once again, and this time with more success. Today, big companies like Facebook, Google, and Valve are reaching out to the consumers to make VR come to life. The coming of VR technology has attracted the attention of many researchers for both clinical and research applications. Research in the field of anxiety disorders like acrophobia, social anxiety, arachnophobia and their treatments (Anderson, et al., 2013; Garcia-Palacios, Hoffman, Carlin, Furness, & Botella, 2002; Krijn, et al., 2004) has already been conducted using VR technology. In addition, pain distraction (Malloy & Milling, 2010) and attention (Cho, et al., 2002) research also had its share of VR research. Consequently, using VR to create virtual environments (VE’s) as a research tool could have many advantages. Since every aspect of the VE has to be programmed, every participant would operate in a highly controlled environment. This way studies would not have to cut on experimental control. However, the main advantage of using VR as a research tool is the increase in ecological validity (Bohil, Alicea & Biocca, 2011; de Kort, Kooijman & Schuurmans, 2003). IJsselsteijn, de Ridder, Freeman and Avons (2000) argue that this is due to the concept of behavioral realism. This concept entails that when a display looks more like the intended environment, the observer’s responses will be a better approximation of those he would have exhibited in the intended environment. Therefor an increase in realism of the VE will also increase the ‘realism’ of the observer’s responses. Fittingly, Loomis, Blascovich and Beall (1999) state that the “ultimate representational system” would have to be “an experience indistinguishable from normal reality” (p. 557).
When working with virtual reality it is important to take some aspects about subjective experiences into account. Sensory input is manipulated to create a more convincing VE for the user, this manipulation can be described by two important terms: immersion and presence. Immersion refers to the amount of sensory information that the VR technology provides. For instance, immersion can be increased by taking three dimensional sound into account, as well as stereoscopy. The psychological equivalent of immersion is presence, a concept encompassing the subjective feeling of being present in the virtual environment (Bohil, et al., 2011). These core concepts of VR can also induce problems with the use of VR technology in empirical studies. When immersion is breached in the slightest way, it can create major artifacts or confounds in the collected data. This happens for example when the display of the VE is slightly delayed in contrast to head movement, which may result in simulation sickness (Akizuki, et al.,
2005). The mismatch of different sensory inputs will also cause a disruption of the sense of presence. However, these errors can be resolved by the improvement of VR technology.
Common (neuro-) psychological tasks have another shortcoming which could be resolved with VE’s. These tasks mainly consist of simple figures or pictures, which are not very response evoking. The participant’s effort to perform is based solely on the motivation induced by the experimenter’s instructions, which is not a strong or ecologically valid motivator. This way participant motivation will likely be low and not representative for a real situation. However, it is arguable that an immersive VE will motivate the participant by evoking natural responses to certain stimuli, and thus increasing the effort on the task. The high ecological validity of VR studies lay the premises for this claim, arguing that participants’ responses will be a closer representation of natural responses to an environment (Bohil, et al., 2011; de Kort, et al., 2003). Research studying the effects of VR on arousal also contribute to this hypothesis, in multiple studies an effect of increased physiological arousal is found in VR-users (Calvert & Tan, 1994; Macedonio, Parsons, Digiuseppe, Weiderhold & Rizzo, 2007). Beauducel, Brocke and Lueu (2006) describe effort and arousal to be closely linked by means of the same neural system. A note of importance is that they do not imply a linear connection between the two. In addition, Grigorovici (2001) notes that, irrespective of little knowledge about presence and arousal, there is evidence indicating an increase of physiological arousal when partaking in more presence-inducing media. Specifically, evidence for an increase in heart rate (HR) is found when the feeling of presence is increased. There is, however, little knowledge about the role of arousal in relation to effort. This paper will focus on this relationship using an effort-based decision making (EBDM) paradigm in combination with HR and heart rate variability (HRV) measurements for arousal. HRV is found to be a strong correlate of norepinephrine levels and thus arousal (Malik, 1996). Due to more sympathetic influence from the autonomic nervous system, physiological arousal is enhanced and HRV, measured on the time domain, is reduced (van Ravenswaaij-Arts, Kollee, Hopman, Stoelinga, & van Geijn, 1993).
EBDM is based on the cost-benefit analyses animals make when confronted with a decision. At the foundation of this paradigm lies the law of least work, this law states that animals have a preference for the least effortful choice alternative when forced to make a choice (Schouppe, Demanet, Boehler, Ridderinkhof & Notebaert 2014). This is accompanied by the concept of effort discounting: the reduction
in net value of a reward when more effort is required to obtain it (Prévost, Pessiglione, Météreau, Cléry-Melin & Dreher 2010). These concepts make organisms strive to maximize reward, whilst minimizing the effort to obtain the reward, also known as the optimal foraging theory (Kurniawan, Guitart-Masip & Dolan, 2011; Stevens, Rosati, Ross & Hauser, 2005). The optimal foraging theory results in effort-based decision making of organisms, in which a decision is made by weighting the perceived effort expenditure and the perceived reward value. When amplifying the effort needed for a certain reward, it is expected for an organism at some point to no longer invest the amplified effort for that reward. At this point the reward is no longer worth the effort. This concept can be made measurable by in- and decreasing the effort necessary for a certain reward and pinpointing a border between choosing and not choosing for a certain effort-reward weighting. The border between choosing and not choosing is called the point of indifference (POI). This point can also be found by lowering the reward for a fixed amount of effort, at some point the reward is low enough for the organism to no longer want to invest the same effort. This reward manipulation will be used to assess the POI in the current study.
Three conditions will be used to assess the POI, one baseline, one with improved reward realism, and one with improved effort realism. Because of the proposed more response evoking character of realistic environments, an increase of POI in contrast to the baseline is expected to occur in the realistic conditions. Findings by Grigorovici (2001) add to the expectation that arousal also plays a role in this exchange, increased environmental realism through virtual reality is found to increase heart rate.
This study will thus research the effect of increased task realism on the amount of effort participants are willing to exert for a certain reward. By measuring the HR whilst participants perform the experiment, the study also takes the HR, effort, and immersion exchange into account. Due to the expectation of higher states of arousal in more realistic environments, higher HR and lower HRV are expected to be found in the realistic effort and realistic reward conditions in contrast to the baseline. The expected reduction in HRV is based on the finding of smaller differences in time between heartbeats when arousal increases. These differences are called inter beat intervals (IBI’s). Higher influence of the sympathetic nervous system is found to stabilize heart rate and thus lowering HRV (Malik, 1996).
Method
Participants
A total of 51 participants were recruited to take part in the experiment, 10 of them were excluded from the POI analysis due to ceiling effects in measurement. Another participant also got excluded from the main analysis due to nausea, and therefore not completing the experiment. Of the remaining 40 participants 42.5% was female (N=17) and the mean age was 22.48 (3.15). For the ECG analysis all participants were reassessed. 24 participants were excluded because their data was missing or severely impaired by muscle noise and sweating. Of the remaining 27 participants of the ECG analysis 33.3% was female (N=9) and the mean age was 23.52 (3.51). Participants were recruited through contacts of the author and colleagues, social media, and pamphlets. They did not receive a fixed payment for their participation. However, a variable amount of money was awarded to one randomly drawn person, the amount was determined by that person’s choices on the actual experiment. Furthermore, some participants received participation points which first year Psychology students at the University of Amsterdam are required to collect.
Simulation
A simulation was created for the experiment using Unreal Engine 4 (Epic Games, 2012), a game development tool used in the creation of a wide variety of games. The simulation contained three different conditions where both the realism of the effort and the realism of the reward were manipulated. Every trial required the participant to make a weighted choice between the amount of required effort and the reward coupled with it. All participants completed all three conditions in a counterbalanced order, with each condition consisting of 13 trials in VR. Each trial participants were instructed to power a minecart over a track by making a pumping motion. At the beginning of each trial participants were given a choice between a high effort (HE) and high reward (HRe) route or a low effort (LE) and low reward (LR) route. Color-coding in presentation of the tracks informed participants about the amount of effort a route would
require; green sections of the track required no pumping input, orange sections of the track required medium effort and red sections of the track required high effort pumping.
In the baseline condition the different route options and coin rewards were displayed on two different virtual screens, one on the left side of the virtual room and one on the right. The coins were displayed abstractly on these screens as stacked orange bars. After a choice was made participants were shown a power bar and their track progress. The powerbar displayed if the amount of pumping by the participant was sufficient for the cart to move. The powerbar would slowly lower or ‘empty’ without pumping and increase or ‘fill’ with every pumping motion made. More pumping would result in a faster ‘filling’ of the powerbar. When the power bar was green their progression would go at full speed, when orange it would move slowly and when red or empty the cart would not move at all. This way the necessary effort of the participant to finish a trial was depicted. However, before they could start, participants first had to wait for the start sign in the virtual environment. Participants were only able to track their progress on a static bar on the right virtual screen, no cart was visibly moving. This baseline condition mimics a standard neuropsychological computer task without making it fundamentally differ from the other conditions in terms of the presence of a VR head-mounted display or not. Figure 1 shows an example of a trial in the baseline condition.
In the realistic reward condition, the same virtual environment was used. But now rewards were realistically represented as stacks of golden coins. These stacks were placed on the left and right side of a desk in front of the participant, no longer on the virtual screens. After making a choice the coins of the chosen route would fly in front of the participant and into a chest on the desk. The condition did not differ from the baseline condition anywhere else. An example of this condition is shown in Figure 2.
Figure 2. A screenshot from a realistic reward condition trial before making a choice.
In the realistic effort condition, participants found themselves in a mine cart inside the same virtual room as the other two conditions. Two screens displayed the different route options and the abstract reward, just like in the baseline condition. After selecting a route, a large door would open and the minecart either moved left or right based on the participant's choice. After the start sign the participants could move the cart forward by pumping, this time the pump was also present inside the cart in the virtual environment. Some sections of the outside tracks were overgrown with either grass (medium effort) or shrubs (large effort). A green powerbar would move the cart at full speed, an orange powerbar would move it at slowly and red did not move the cart at all. An example of this trial is shown in Figure 3.
Figure 3. A screenshot from a realistic effort condition trial after making a choice.
Every new trial the reward had to be modified to be able to calculate the POI, this was done by a scripted reward modifier. To approximate the POI, each HE choice had to reduce the difference between track rewards, and each LE choice had to increase the difference. This way every trial there is a better approximation of the POI. Figure 4 shows three examples of this process. For each trial the reward for both tracks was calculated by subtracting the total effort values from the tracks and multiplying this value with the reward modifier. The effort values were based on zero points for green, two points for yellow and four points for red. These values were added by ten for the high effort track and subtracted by ten for the low effort track. The value of the reward modifier ranges between zero, which means no difference between rewards, and 1.25. Each trial starts with a reward modifier of 0.625, the maximum reward modifier divided by two. When a HE choice was made a value was subtracted from the reward modifier and for each LE choice a value was added to the reward modifier. The added or subtracted value increased for each consecutive choice of the same effort type. This value was 0.05 for the first, 0.1 for the second, and 0.2 for the third and higher consecutive choices. With each switch in effort type choice this value would drop to 0.02, the next consecutive choice would bring it up to 0.05 again, and so on. The minimum value of the reward modifier had been set to zero to prevent the LE choice from having a higher value than the HE choice. The maximum value of the reward modifier had been set to 1.25 to avoid scores higher than 20 coins.
Figure 4. Reward modifier values per trial from three different example conditions stabilizing
around either a high, medium or low point of indifference.
The simulation was projected through the Oculus Rift, a VR head-mounted display currently in production by Facebook. For this experiment the DK2 Oculus Rift (Oculus VR, 2014) was used, a second generation developers kit.
The experiment used a custom-made input device. The goal of the device was to make participants invest considerable effort and enhance their immersion by mimicking the VE. Because of this a bicycle pump was used as the basis of the device. Operating the pump resembles the pumping motion of the pump within the VE and the (air) resistance made the motion require considerable effort. This motion was digitally registered by a computer mouse. An aluminum strip was attached to the handle of the bicycle-pump and thus followed the motion of the pump. On the strip the mouse (Logitech G300) was loosely placed so that the aluminum strip would pass beneath it. This way the mouse would register the movement of the pump handle and could also function as a way to register the participant’s choice of track with their index fingers. The pump was attached on a large MDF board which was fixed in place.
Measurements
POI values were calculated for all participants in all three conditions. Assigning POI values has demonstrated to be a reliable method for measuring individual differences in subjective effort perception (Westbrook, Kester, & Braver, 2013). A POI is a point where an individual no longer expresses a preference for any choice option. At this point a person would be just as likely to pick a HE/HRe option as a LE/LR choice. In this experiment the modifier value was used to represent the POI, and thus could take on a value between zero and 1.25. To determine the POI value for each condition the average reward modifier value of the last four trials is calculated. A low POI is achieved when participants need relatively little reward to choose for a HE option and a high POI is achieved when participants need a relatively high reward to choose for a HE option. When the POI is equal to zero the subject chooses the HE/HRe option without making a weighted decision between reward and effort. Differences in POI values between conditions would represent differences in perceived effort, the representation form of effort and reward could account for these differences. This would also imply that, when effort representation is held constant, differences in POI would be due to differences in reward representation.
HR and HRV were measured using an electrocardiograph. Three electrodes were placed just above both pectorals of the participant and on the left side of the abdomen, slightly below the ribcase. These electrodes were connected to an amplifier set to sample at 1000 Hertz. The signal was send to a laptop and filtered from AC using a 50Hz notch filter, this way electrical current noise from power sources near the electrodes was filtered out by removing data with a 50Hz frequency from the frequency domain. The outputted data was imported to MATLAB (version 6.1) to cut timed epochs between stimulus onset and track choice, and to calculate participant’s means for each condition. Those means were analyzed using SPSS (version 21). The HR data was measured in IBI’s, every IBI represents the time in milliseconds between two r-tops. From these IBI’s the HRV was calculated for each epoch by subtracting every IBI from the next, dividing the square of this number by the amount of IBI’s per epoch and finally taking the square root of the sum of these values.
To measure the presence of participants in the virtual environments a modified version of the Igroup presence questionnaire (IPQ) was administered directly after the experiment. The IPQ contained 13 items assessing the feeling of ‘being inside of a virtual environment’. A higher score on the IPQ implies
greater feelings of presence, or ‘being there’. The scores were taken from a five point likert scaling, varying from not at all to very much, or similar scoring. The IPQ score was standardized by averaging individual answers, this way there was a minimum score of one and a maximum score of five. The Dutch version of the IPQ was used and modified so that all three conditions had to be assessed by the participants. Some extra items were added to gain knowledge about the subjective feeling of differences between conditions and possible strategies for decision making during the experiment. Example questions from the original IPQ are: ‘in the computer generated world I had a sense of being there’, and ‘I was completely captivated by the virtual world’.
Procedure
Before the experiment participants were shortly instructed about the simulation, an important remark made here was for the participants to let their track choice be guided by an honest consideration of effort and reward. Participants were also told to stop the experiment and call for the experimenter whenever he or she would feel unwell. After the instruction the electrodes were placed on the participant's body in a way the wires would pass beneath the clothes to prevent them from being in the way of the pumping. Next the oculus rift was placed on the participant's head and three practice trials were started, one trial for each condition. During this practice the experimenter would monitor the participant and give additional instructions when necessary. After the three trials the simulation was started, and the experimenter told the participants they would be unable to monitor the participant's progress to minimize social influences. A headphone was placed on the participant's head and they were instructed to make their first choice. The order in which conditions were presented was counterbalanced. When the participants finished the experiment they were asked to fill out the IPQ and were thanked for their participation.
Results
Prior to the main analysis a manipulation check was executed to check if the realistic reward and effort conditions actually increased the feeling of presence. A repeated measure analysis of variance (ANOVA) was used to check if participants scored significantly higher on the IPQ when asked about the more realistic reward and effort conditions. Five participants did not complete the IPQ and were therefore excluded from this analysis. Due to possible confusion in formulation of question four it has been removed from the analysis, the question asked participants if they did not feel present in the virtual environment. The answers on this question were inconsistent with the other answers of multiple participants, this could be due to confusion because of the negative nature of the question. For example, a participant answered ‘very much’ when asked if they felt immersed in the virtual environment, but also answered ‘very much’ when asked if they did not feel present in the virtual world. Because multiple participants had these contradictory answers on question four in contrast to other questions it was deleted from the analysis. Furthermore, in this analysis the assumption of sphericity has not been met, so a Huyn-Feldt correction was used. Because of the relatively large sample and the central limit theorem the assumption of normality has been considered met. The mean IPQ scores of the analysis can be found in Table 1.
Table 1. Mean IPQ score, POI, HR and HRV with Standard Deviations (between brackets) of all
three Conditions. Baseline Realistic reward Realistic effort IPQ score 2.89 (.35) 3.01 (.38) 3.21 (.37) Mean POI .87 (.27) 1.25 (.37) 1.23 (.40) IBI (ms) 663.94 (103.98) 677.09 (200.31) 637.64 (96.86) HRV (ms) 65.64 (58.52) 86.98 (112.61) 80.62 (116.19)
There was a significant effect of presence found between the different conditions F(1.941, 45) = 18.688, p < .01. To check if this difference matches the hypothesis and the realistic conditions significantly differ from the baseline condition, simple contrasts were added to the analysis. There was a significant difference between the baseline condition and the realistic reward condition F(1, 45) = 5.47, p = .2 and a significant difference between the baseline condition and the realistic effort condition F(1, 45) = 32.62, p < .01. In Figure 5 it is clear the mean IPQ score increases when the environment becomes more realistic. These effects thus indicate that the manipulation to increase a feeling of presence was successful.
Figure 5. Mean IPQ scores and confidence intervals of all three conditions.
The main analysis to test if there is a difference of POI between conditions was also conducted using an ANOVA. The data met the assumption of sphericity, so no correction was used. Due to the large sample the assumption of normality was also assumed met. The mean POI’s of this analysis can be found in Table 1. There was no significant difference found between the three conditions on POI values F(2, 39) = .82, p = .45. This contradicts the expectations that the POI of the realistic conditions would differ from the baseline. The confidence intervals in Figure 6 show the lack of a difference between conditions. Simple contrasts were also added to the analysis, these showed no significant difference for realistic
reward versus baseline F(1, 39) = .17, p = .68 and neither for realistic effort versus baseline F(1, 39) = 1.93, p = .17. Due to the lack of significant difference these results contradict the expectation of higher POI’s in the more realistic environments. A lack of significant difference indicates the participant's choice behavior was barely influenced by the difference in realism between conditions.
Figure 6. Mean POI and confidence intervals of all three conditions.
The IBI and HRV data was also analyzed using an ANOVA. For this analysis all excluded participants of the main analysis were taken into account and reevaluated. From this analysis 24 participants were excluded due to missing or faulty data from muscle noise and excessive sweating. The remaining 27 participants were all included in the IBI and HRV analysis. The means of the data can be found in Table 1. In both cases the assumption of sphericity was violated and the Greenhouse-Geisser correction was used to interpret the results. Due to a smaller sample, because of the many excluded participants, the assumption of normality had to be tested. A significant difference was found in the Kolmogorov-Smirnov tests of HR and HRV data for all conditions except for the HR realistic effort condition data. The p-values reach from smaller than zero to 0.046 with the exception of 0.20 for the HR realistic effort condition data. Because of these differences the assumption of normality for both tests has
been considered violated. Further study of the data revealed large differences in kurtosis and skew between conditions, because of this severe impairment of the assumption the chance of false positives is considered to be higher. Simple contrasts were executed for both analyses to check for differences between the baseline and realistic conditions. In the main ANOVA’s there was no difference found between the conditions for the IBI’s F(1.15, 26) = .77, p = .40 as well as for HRV F(1.48, 26) = 1.09, p = .33. This contradicts the hypothesis that there is a difference in HR and HRV between different kinds of presence in a virtual environment. The corresponding baseline versus reward contrast for the IBI data was also found to be non-significant F(1, 26) = .11, p = .74. The baseline versus realistic effort condition for the IBI data was found to be significant F(1, 26) = 4.63, p = .04, however due to the violation of the assumptions this result should be interpreted with great care and is likely to be a type-I error. The HRV contrasts were found to be non-significant with F(1, 26) = 1.60, p = .22 for realistic reward versus baseline and F(1, 26) = .79, p = .38 for realistic effort versus baseline. These results contradict the hypothesis of higher HR and lower HRV when participants are emerged in more realistic environments. The confidence intervals in 7a clearly show a lack of significant difference in HRV data between conditions, which contradicts the hypothesis. Figure 7b shows a decrease of IBI length and thus an increase in HR between baseline and realistic effort, which would support the hypothesis of higher HR in more realistic
environments.
Figure 7. Mean HRV (a) and HR (b) and confidence intervals of all three conditions.
B A
Discussion
The current research has examined if more realistic virtual environments increase the amount of effort people will put into a task for a certain reward. However, no significant difference was found, there was no increase in the POI, a point where there would not be a preference for either choice option. The manipulation of a feeling of presence was successful but did not seem to have effect on the POI. Additionally, the hypothesized increase of HR and decrease of HRV were not found with the exception of a significant increase of heart rate between the baseline condition and the realistic effort condition. This effect should however be interpreted with care and is likely to be a type-I error. Because of this the null hypotheses could therefore not be rejected and no effect could be supported by this experiment.
Earlier research argues that VR could greatly increase ecological validity of neurocognitive research by simulating more similar to real-life environments without losing much control (Bohil, et al., 2011; IJsselsteijn, et al., 2000; de Kort, et al., 2003). With more ecological validity participant’s reactions to that environment would be more true to realistic reactions (Loomis, et al., 1999). Other research on VR technology found an increase of arousal within these more ecologically valid environments (Calvert & Tan, 1994; Macedonio, et al., 2007). Together with knowledge of a strong connection between arousal and effort (Brocke & Lueu 2006) this lays the premises for the hypothesis of increased effort in more realistic virtual environments. However, the current research failed to support this hypothesis. It is therefore not consistent with the joined findings of earlier research on VR. The current research adds insights in effort based decision making within humans when emerged in virtual environment, and is one of the first to do so. The lack of findings invites to further research this area of knowledge with different approaches.
Analyses were also executed to analyze HR and HRV using IBI data collected by an electrocardiograph. Studies on HR and HRV found strong correlations between the two measures and states of arousal. In a state of arousal, more sympathetic control will increase HR and decrease HRV (van Ravenswaaij-Arts, et al., 1993). The current research did however not find any difference in HR and HRV. The contrast of heart rate between baseline and realistic effort was found to be significant but should be interpreted with great care. Due to the violation of both the assumption of sphericity and normality, the chance of a type-I error is greatly increased, furthermore the main analysis of the contrast did not point out
a significant difference. The results did not support the hypotheses of increased states of arousal in more realistic and immersive environments. These results could however be due to errors and imprecision in measurement.
Due to the innovative and progressive nature of the study it is one of a kind, no similar research was conducted before. This also means the design is in early stages of development and may not be a valid or reliable instrument. Some shortcomings that could have influenced the data will be summed up, these points should be reconsidered and improved when conducting future research on the subject.
A first factor of influence was the relatively low automatization of the experiment. A lot of steps required human interaction and were executed by different experimenters, which led to a series of errors. Due to experimenter influence some participants did not receive the right counter-balance, this has especially high influence on the data because of the physically tiring task they had to perform. Late conditions are biased by possible tiredness, boredom, annoyance and higher HR. This way when the counter balance is not executed correctly these factors are confounding variables in the design. Another human error comes from the forgetting and adding of steps in the process. The experiment required a relatively high amount of explanation and direction from the experimenters, even the simulation’s mechanics had to be explained. When an experimenter forgets or adds anything in the process this could greatly influence the data, so understanding of the simulation’s mechanic could be a strong confounding variable. These errors can however be resolved easily, more automation will remove these human errors by selecting the counterbalance automatically and instructing every participant the same way within the virtual environment. This way no counter balance errors can occur and instruction are standardized.
Another shortcoming of the design is the measurement of a clear point of indifference. To measure it people should seriously have to consider whether the effort is worth a certain reward, the design most likely did not require participants to make this serious decision. It becomes apparent when highlighting the high number of cases with a ceiling effect, this effect is reached when both HE and LE share the same reward. Participants that reached this ceiling effect probably did not make a weighted decision between effort and reward. Questioning possible strategies to complete trials support this hypothesis, some participants indicated they ‘choose as much money as possible, no matter the intensity’, sometimes made a HE choice ‘just to exercise a bit’, and were focused on ‘going fast’. The lack of
weighted choices could be due to too little effort required to actually be considered a chore, or the lack of an incentive reward. Either of the two may not be enough for participants to consider both and weigh them against each other. For future research on the subject it would be recommendable to increase the required effort and increase the incentive value of the reward by making it something participants would want to work for just enough.
The input device did also not suffice to accurately measure the pumping motion of the participants. The pump required a specific motion to register the movements of the handle correctly, when slightly off it could miss registering some or all of the motion and would not fill the powerbar. This is not only especially frustrating for the participants but also a bad measure of effort. Sometimes controlled and slow motions delivered more result than actual fast pumping, this way the data would not be representational for HE or LE and thus lacks validity. In future research the device should be modified or replaced and the actual registration should not be a delta value of movement like the mouse, but a fixed scale of movement. This way the movement will always be registered.
Because of the very recent development of VR another confounding factor influenced the data. The technology is still very new and consumer versions are just now beginning to appear on the market, this way very little people had a VR experience before. Because of this, participants were often amazed by the virtual environment and enjoyed the experience. Participants could have been influenced by this excitement and choose HR tracks simply because they enjoyed the simulation. Furthermore, some participants were distracted by their surroundings, looking around them and discovering the environment. This could have distracted them from making weighted choices. Because of this future research with similar uses of VR technology should consider to first offer a fun and exciting VR experience and let the user get familiar with the technology, so the will be less amazed during the actual experience.
Also, there were factors influencing and interfering the HR data. The placements of the electrodes allowed muscle noise, especially from the chest muscles to interfere with the ECG data. Even though the epochs were cut between stimulus onset and track choice where no pumping was required, muscle noise still influenced and sometimes strongly disturbed the data acquisition. Even though the algorithm was good at detecting the r-tops, sometimes it was still greatly disturbed. An example is the appearance of a flat line for a certain amount of time shortly after participants stopped pumping, this happened with
multiple participants. These disturbances were accompanied by noise due to excessive sweating. Not only did the skin conductance increase but some electrodes also fell of the body due to the sweat. Eventually this resulted in the high number of excluded participants for the HR and HRV analyses. These problems can be resolved by using a different instrument for HR measurements that is compatible with exercise. Furthermore, because of the earlier mentioned problems with counterbalancing, higher HR’s due to exercise from earlier condition could have influenced the data on later conditions.
Lastly, because of relatively new technology for making and rendering the virtual environment, no markers could be placed in the ECG data on points of interest. Because of this the data had to be cut manually which resulted in some errors of measurement. Differences in system clocks between the two systems and failure of experimenters to accurately note the start of the measurements resulted in variable amounts of imprecision of measurement. The HR data used in this experiment was therefore not very valid and should be interpreted with care. Future studies should use game engine and electrocardiogram hard- and software that can communicate with each other to resolve these issues.
The current study hypothesized an increase of arousal when environments become more realistic which would result in a higher preparedness to put effort into a task for a certain reward. However, it did not find evidence for this hypothesis and thereby failed to support it. But because the research on this topic is in its starting stages and the design is still flawed, future research would have to be conducted to make more definite statements about the subject.
References
Akizuki, H., Uno, A., Arai, K., Morioka, S., Ohyama, S., Nishiike, S., ... & Takeda, N. (2005). Effects of immersion in virtual reality on postural control.Neuroscience letters, 379(1), 23-26.
Anderson, P. L., Price, M., Edwards, S. M., Obasaju, M. A., Schmertz, S. K., Zimand, E., & Calamaras, M. R. (2013). Virtual reality exposure therapy for social anxiety disorder: A randomized controlled trial. Journal of consulting and clinical psychology, 81(5), 751.
Beauducel, A., Brocke, B., & Leue, A. (2006). Energetical bases of extraversion: Effort, arousal, EEG, and performance. International Journal of Psychophysiology, 62(2), 212-223.
Bohil, C. J., Alicea, B., & Biocca, F. A. (2011). Virtual reality in neuroscience research and therapy. Nature reviews neuroscience, 12(12), 752-762.
Calvert, S. L., & Tan, S. L. (1994). Impact of virtual reality on young adults' physiological arousal and aggressive thoughts: Interaction versus observation.Journal of applied developmental psychology,
15(1), 125-139.
Cho, B. H., Ku, J., Jang, D. P., Kim, S., Lee, Y. H., Kim, I. Y., ... & Kim, S. I. (2002). The effect of virtual reality cognitive training for attention enhancement.CyberPsychology & Behavior, 5(2), 129-137.
Epic Games. (2012). Unreal Engine 4 [Game engine]. Retrieved from https://www.unrealengine.com/
Freeman, J., Avons, S. E., Meddis, R., Pearson, D. E., & IJsselsteijn, W. A. (2000). Using behavioral realism to estimate presence: A study of the utility of postural responses to motion stimuli.
Presence, 9(2), 149-164.
Garcia-Palacios, A., Hoffman, H., Carlin, A., Furness, T. U., & Botella, C. (2002). Virtual reality in the treatment of spider phobia: a controlled study.Behaviour research and therapy, 40(9), 983-993.
Grigorovici, D. (2001). Affectively engaged: affect and arousal routes of entertainment virtual reality. In Virtual Systems and Multimedia, 2001. Proceedings. Seventh International Conference on (pp. 634-643). IEEE.
IBM Corp. Released 2012. IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp.
Krijn, M., Emmelkamp, P. M., Biemond, R., de Ligny, C. D. W., Schuemie, M. J., & van der Mast, C. A. (2004). Treatment of acrophobia in virtual reality: The role of immersion and presence.
Behaviour research and therapy, 42(2), 229-239.
de Kort, Y. A., Ijsselsteijn, W. A., Kooijman, J., & Schuurmans, Y. (2003). Virtual laboratories: Comparability of real and virtual environments for environmental psychology. Presence: Teleoperators
and virtual environments,12(4), 360-373.
Kurniawan, I., Guitart-Masip, M., & Dolan, R. (2011). Dopamine and effort-based decision making. Frontiers in neuroscience, 5.
Loomis, J. M., Blascovich, J. J., & Beall, A. C. (1999). Immersive virtual environment technology as a basic research tool in psychology. Behavior Research Methods, Instruments, & Computers, 31(4), 557-564.
Macedonio, M. F., Parsons, T. D., Digiuseppe, R. A., Weiderhold, B. A., & Rizzo, A. A. (2007). Immersiveness and physiological arousal within panoramic video-based virtual reality. Cyberpsychology
& Behavior, 10(4), 508-515.
Malik, M. (1996). Heart rate variability. Annals of Noninvasive Electrocardiology, 1(2), 151-181.
Malloy, K. M., & Milling, L. S. (2010). The effectiveness of virtual reality distraction for pain reduction: a systematic review. Clinical psychology review,30(8), 1011-1018.
MathWorks Inc. Released 2000. MATLAB, Version 6.1. Natick, MA
Prévost, C., Pessiglione, M., Météreau, E., Cléry-Melin, M. L., & Dreher, J. C. (2010). Separate valuation subsystems for delay and effort decision costs. The Journal of Neuroscience, 30(42), 14080-14090.
Schouppe, N., Demanet, J., Boehler, C. N., Ridderinkhof, K. R., & Notebaert, W. (2014). The role of the striatum in effort-based decision-making in the absence of reward. The Journal of Neuroscience,
34(6), 2148-2154.
Stevens, J. R., Rosati, A. G., Ross, K. R., & Hauser, M. D. (2005). Will travel for food: spatial discounting in two new world monkeys. Current Biology,15(20), 1855-1860.
van Ravenswaaij-Arts, C. M., Kollee, L. A., Hopman, J. C., Stoelinga, G. B., & van Geijn, H. P. (1993). Heart rate variability. Annals of internal medicine,118(6), 436-447.
Wardle, M. C., Treadway, M. T., Mayo, L. M., Zald, D. H., & de Wit, H. (2011). Amping up effort: effects of d-amphetamine on human effort-based decision-making. The Journal of Neuroscience, 31(46), 16597-16602.
Westbrook, A., Kester, D., & Braver, T. S. (2013). What is the subjective cost of cognitive effort? Load, trait, and aging effects revealed by economic preference. PLoS One, 8(7), e68210.
