• No results found

Choosing the future : decision-making in virtual reality

N/A
N/A
Protected

Academic year: 2021

Share "Choosing the future : decision-making in virtual reality"

Copied!
21
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Bachelorproject

Choosing the Future: Decision-Making in Virtual Reality

By: Tomas van Duin

Student number: 10386947

Mentor: Jasper Winkel

(2)

Reflectie

Algemeen

Ik heb geprobeerd om aan twee kanten uit te lijnen, maar dat had als gevolg dat het einde van elke alinea bestond uit 2/3 woorden die verspreid werden over een hele regel, zag er niet mooi uit.

Ik heb het verslag zoals je gesuggereerd hebt naar PDF geconverteerd zoals je hebt gesuggereerd.

Inleiding

Ik ben eerst op volgorde je feedback comments af gaan werken. Toen ben ik zelf nog het stuk gaan doorlezen en ben tot de conclusie gekomen dat ik de structuur van de inleiding nog niet helemaal goed vond. Ik heb toen even in steekwoorden voor mezelf uitgeschreven wat ik allemaal wilde vertellen (zie bulletpoints hieronder), vervolgens ben ik gaan puzzelen wat de meest logische volgorde van deze punten zou zijn, en dat is geworden:

- Ecologische validiteit is een issue in psychologie

o Voorbeeld met EEfRT (hierin wordt gesproken over het real ervaren van experimentele stimuli)

- Virtual reality biedt een mogelijke oplossing voor deze issue

o Illustratie door verschillende mate van immersion van onze gebruikte taken - Decision making

o Kort uitleggen wat EBDM is o Kort uitleggen wat DD is

o Beargumenteren waarom delay discounting ook van toepassing is op delayed effort.

o Beargumenteren waarom de verschillende taken verschillende mate van perceived distance hebben

- Verwachting

o EBDM: immersion > presence > echtere ervaring > hogere PE

o DD: Immersion > perceived distance > hogere mate van DD > hogere PE bij immediate vs delayed effort.

Ik heb geprobeerd om eerst alle argumenten voor mijn onderzoek aan te voeren, en vervolgens pas aan het einde deze samen te brengen tot één verwachting over de onderzoeksresultaten. Hiervoor heb ik nog een korte beschrijving van EBDM in het

‘Decision-Making’ stuk toegevoegd. In ‘Ecological Validity’ begin ik al over EBDM. Ik zag helaas geen mogelijkheid om deze twee stukken over EBDM samen te voegen zonder het hele verslag om te gooien, dus refereer ik in het tweede stuk naar het eerste stuk.

Methode

Ik heb je kleine feedbackjes verwerkt en het een en ander dat irrelevant was geschrapt en plaatje toegevoegd om de uitleg van de taak toe te lichten.

(3)

Resultaten

Ik heb tabellen en grafieken toegevoegd. Ik heb het exploratief stuk over Liking en Trying geschrapt, aangezien dit toch geen verschilscores kon verklaren en veel ruimte in beslag nam. Ik heb in plaats daarvan een exploratieve tabel (bij tabel 1 gevoegd) en grafiek toegevoegd met het verschil tussen POI1 en POI2 tussen de drie condities. Dit zijn geen significante resultaten, dat heb ik erbij vermeld. Ik gebruik deze om in ieder geval iets van onderbouwing toe te voegen aan het argument in de discussie dat de manipulatie van effort te zwak was. Discussie

Zoals hierboven vermeld gebruik ik insignificante data om een argument te illustreren. Je hebt in je feedback aangegeven dat ik los per proefpersoon de variantie moest berekenen tussen condities om zo te onderbouwen dat er inderdaad regressie van POI naar 0 is over tijd. Ik ben even naar de methodologiewinkel geweest daarvoor, en die gaven aan dat ik het beter zo kon doen omdat het toch niet significant was, maar wel mooi om te gebruiken op de manier zoals ik gedaan heb aangezien het wel wat toevoegt aan een discussie met zonder data

onderbouwde argumenten.

Verder heb je aangegeven dat het argument dat de 2D taak onbedoeld saai was een goed argument is, en dat ik mogelijk kon voorstellen om in vervolgonderzoek 3 taken in de Oculus te doen met enkel verschillen in action fidelity, dat heb ik gedaan.

Verder heb ik geprobeerd om de discussie zo vele mogelijk tot theoretisch taalgebruik te beperken, en dit congruent te houden met het theoretisch taalgebruik in de inleiding. Verder waren er geen deelnemers met gemiddeld POI2 < 3 die geen gemiddeld POI1 < 0 hadden, je had namelijk aangegeven dat ik daar een discussiepunt van zou kunnen maken. Verder heb ik in de discussie nog een argumentatielijn toegevoegd over effort manipulatie, perceived effort, perceived reward, en de gevonden resultaten op POI. Mogelijk ben ik hier te stellig, maar ik vond dit het mooiste discussiepunt om te maken gezien ik door de data vrij tamelijk vrij was om de discussie in te vullen zoals ik wilde.

Ik ben niet helemaal blij met hoe kort ik ben over DD in de discussie, maar zo simpel is het ook, er zijn geen significante resultaten. Omdat er geen meting was van perceived distance heb ik dit uit mijn stuk moeten laten. Dit heeft het argumenteren weer verder beperkt omdat ik een variabele minder heb om mee te redeneren.

(4)

Abstract

As long as experimental psychology has existed there has been a tradeoff problem between ecological validity and experimental control. Virtual reality (VR) offers a promising solution for this tradeoff problem. This pilot study aimed at exploring the application of VR in experimental psychology by assessing the effect of more immersive paradigms on effort based decision-making (EBDM) and delay discounting (DD). Results yielded no significant effects of more immersive paradigms on EBDM and DD, however show promising effects on presence. These findings and the future of research on exploring the applicability of VR in experimental psychology will be discussed in detail.

(5)

Choosing the Future: Decision-Making in Virtual Reality

Tomas E. van Duin

Introduction

“This is your last chance. After this, there is no turning back. You take the blue pill - the story ends, you wake up in your bed and believe whatever you want to believe. You take the red pill - you stay in Wonderland and I show you how deep the rabbit-hole goes (The Matrix, 1999).”

We have all seen the movie The Matrix in which Neo discovers that the life he’s been living is not real, but a simulation. At first, even though he believes this new discovery, and knows that what he is experiencing is not real, he has a hard time not behaving like it’s real. VR is like this. Imagine that you are in a place that you, like Neo, know to be a simulation. It’s not a physical place, but an illusion created by a VR system. You know that what you see, hear and feel is not really happening, yet you find yourself thinking, feeling and behaving as if the place were real and as if it were really happening. You see a big cliff in front of you, your heart is pounding in your chest and you approach it with caution, afraid to fall down. From a cognitive point of view, you know that there is nothing there, but both consciously and unconsciously, you respond as if there is (Sanchez-Vives & Slater, 2005).

In experimental psychology we want participants to react to test stimuli as if they were real in order to acquire ecologically valid data. We however also want sufficient experimental control over our variables. To acquire experimental control simplified stimuli are used in experimental environments, leaving us guessing about the ecological validity of the results (Bohil, Alicea, & Biocca, 2011).

VR might offer a solution for this tradeoff problem. VR enables participants to behave in contextually rich and naturalistic scenarios, while granting researches high control over the variables of interest (Bohil, Alicea, & Biocca, 2011). VR has existed for a while, and has become more and more accessible to researchers. However, experimental psychology and VR have remained quite separated (Sanchez-Vives & Slater, 2005). In this article we will explore the application of VR in a combined EBDM/DD paradigm, comparing similar tasks with varying levels of immersion, assessing it’s influence on EBDM and DD.

Ecological Validity

In experimental psychology our goal is to measure naturalistic behavior. In order to make exact predictions about the causal relationships of this behavior, experimental control over all variables involved in this behavior is needed. It is impossible to control all variables in the natural environment in which the behavior of interest occurs, therefore an experimental environment which contains just those variables we are interested in is created in the

laboratory (Araujo, Davids, & Passos, 2007). The environment is an important factor that shapes behavior. Therefore, stripping the experimental environment down to just those

(6)

variables we are interested in will result in behavior qualitatively different from the

naturalistic behavior we aim to measure (Brunswik, 1956). This distinction between behavior we aim to measure and behavior that is measured has implications for the ecological validity of results obtained in psychological research: to what extent can we generalize results obtained in the lab to real life (Bohil et al., 2011)?

This ecological validity issue can be illustrated with an example from an EBDM paradigm. EBDM is about the analysis of cost-benefit valuation (Croxson, Walton, O’Reilly, Behrens, & Rushworth, 2009). The Effort-Expenditure for Rewards Task (EEfRT) (Figure 1) (Treadway, Buckholtz, Schwartzman, Lambert, & Zald, 2009) is the most used task to

simulate EBDM in humans. In this task an experimental environment is created, in which participants are to make decisions about monetary rewards. The goal here to simulate naturalistic decision-making behavior in participants and measure it.

In order to simulate such behavior, the participant must feel like the choices in the task are real choices (Araujo et al., 2007). However, the participant doesn’t feel like the choices are real, because the experimental environment doesn’t feel real. In order for the participant to experience the experimental environment as real, the participant must have the experience of presence in this environment (Witmer & Singer, 1998). Presence refers to an individual’s subjective experience of ‘being there’ in an artificial environment, while being physically present in another (Bowman, Mcmahan, & Tech, 2007).

Presence is the product of immersion. Immersion is the extent to which an artificial environment enables a participant to perceive oneself to be enveloped by, included in, and interacting with an environment that provides a continuous stream of stimuli and experiences (Witmer & Singer, 1998; Sanchez-Vives & Slater, 2005). An environment’s level of

immersion depends on whether the participant is isolated from the physical environment, has the perception of being included in this isolated environment, has a natural way to interact with the environment, and has perception of self-movement (Witmer & Singer, 1998).

When engaged in the EEfRT, the participant is sitting in a laboratory, looking at a computer screen. Therefore, he is not isolated from the physical environment. While engaged in the task, his surroundings provide cues that distract him from the task, which enables his

Figure 1. Schematic diagram of a single trial of the EEfRT. A, Trials begin with a 1s fixation cue. B,

Subjects are then presented with a 5 s choice period in which they are given information regarding the reward magnitude of the high-effort option and the probability of receiving a reward. C, The 1s “ready” screen. D, Subjects make rapid button presses to complete the chosen task and watch a virtual “bar” on the screen that fills up as they progress to their completion goal. E, Subjects receive feedback on whether they completed the task. F, Subjects receive feedback as to whether they received any money for that trial.

(7)

mind to wonder off to business in the physical environment (Witmer & Singer, 1998). Non-isolation from the physical environment also prevents the perception of self-inclusion (Witmer & Singer, 1998). Because the participant has the experience of his own body being in the physical environment, he cannot at the same time experience this as to be included in the virtual environment. This automatically prevents the perception of self-movement because he does not perceive his physical body to be moving within the environment (Witmer & Singer, 1998). The EEfRT also lacks action fidelity. Action fidelity refers to the extent to which the participant can interact with an artificial environment in a ‘natural’ way (Araujo et al., 2007), i.e. a driving simulator must have paddles and a steering wheel in order to allow the user to interact with it in a ‘natural’ way, and is essential for immersion in a virtual environment (Sanchez-Vives & Slater, 2005). In the EEfRT participants interact with the experimental environment in an unnatural way: button presses represent their choices and exerted effort, and the consequences of their actions are perceived through a remote screen. The EEfRT perfectly illustrates why most tasks in experimental psychology do not have high levels of immersion: they lack isolation, self-inclusion, the perception of self-movement, and action fidelity.

Virtual Reality

VR might offer the solution to this problem. VR is an immersive medium with that has proven to induce high levels of presence and realistic responses in users (Bohil et al., 2011). VR can be used to simulate a great variation of naturalistic environments, while maintaining experimental control over the variables in these environment (Bohil et al., 2011).

To explore the application of VR in experimental psychology we have set up an experiment in which we assess the effect of three similar decision making paradigms with varying levels of immersion on task performance. We set up the same task in three different forms: 2D, 3D and VR, with corresponding ascending levels of immersion. In the task participants have to make a choice between two symbolic rewards, each requiring the participant to exert a certain amount of effort to obtain them. To exert effort participants are ought to repeatedly move two joysticks in opposite direction along a forward backward axis.

In the 2D conditions participants only see a schematic representation of the choice alternatives, effort, and task progress on a computer screen. This condition was designed to have the same degree of immersion as the EEfRT: there is no isolation, self-inclusion, perception of self-movement or action fidelity. The 3D condition has a higher degree of immersion than the 2D condition. The 3D condition is like a video game: on the computer screen the participant sees a cart moving across a track in a three dimensional environment. This is experienced through first person view, as if one is sitting in the cart and controlling it, resulting in a low degree self-inclusion (Witmer & Singer, 1998). Movements made with the joysticks are coherent with the movements seen in the task, contributing to action fidelity (Araujo et al., 2007). The 3D condition, however, still lacks isolation and perception of self-movement. The VR condition has the highest degree of immersion. The VR condition is exactly like the 3D condition, but viewed through the Oculust Rift DK2 (Oculus VR, 2014). This is a head mounted display which visually isolates the user from the physical world. This

(8)

prevents cues from the physical environment to distract the participant from the experimental environment, increasing immersion (Witmer & Singer, 1998). The Oculus Rift tracks the participant’s head movements, displaying these in the virtual environment. The congruency between head movement in both the physical and experimental environment allow

participants to use natural movements to perceive the experimental environment, causing a high level of action fidelity (Sanchez-Vives & Slater, 2005). These head movements give rise to an egocentric point of view through which the participant experiences the experimental environment, leading to self-inclusion (Bohil et al., 2011). Because the participant’s point of view moves as the cart moves, and the participant is isolated from the physical environment, the participant has the perception of self-movement (Witmer & Singer, 1998).

Decision-Making

Decision-making is a complex process in which many factors play part (Doya, 2008). Choices do not only reflect the expected reward magnitude of choice alternatives, but also the valuation of the costs of the effort that is require to obtain a reward (Croxson et al., 2009). As described earlier, this is referred to as EBDM. Another factor in decision-making is time. Humans have the urge to discount value over time, thus perceiving the value of matters in the future as less than those same matters in the here and now. This is referred to as DD (Kirby & MarakoviĆ, 1996).

When considering economic models describing curves for both DD and EBDM, these two are almost identical (Prévost, Pessiglione, Météreau, Cléry-Melin, & Dreher, 2010). Neuroscientific research, however, shows that there are distinct brain mechanisms involved in EBDM and DD (Prévost et al., 2010). Considering that from a neuroscientific point of view these are two completely separate phenomena, investigating the effect of immersive

paradigms on both EBDM and DD seemed to be a valuable expidition.

Most DD paradigms involve discounting the value of monetary rewards over time. However, humans also discount the value of effort over time, even more strongly so than monetary rewards (Soman, 2004). According to the temporal construal theory (Trope & Liberman, 2003) this is because effort, compared to money, has more abstract aspects that are easily discounted. Effort for example requires the performance of an actual deed. This deed takes time, requires certain actions to be performed, these actions require energy, etc. All these abstract characteristics of effort that determine its value are easily forgotten when perceived it in the future. Therefore humans underestimate the value of effort in the future, resulting in DD of effort, and delayed effort being perceived as less than immediate effort (Soman et al., 2005).

To investigate whether participants experience higher perceived effort (PE) in more immersive paradigms we set up an experiments in which we compared decision-making behavior between three paradigms with differing levels of immersion. Decision-making behavior is based on PE and perceived reward (PR). We assumed PR to be equal between paradigms, and therefore that any difference in decision-making behavior would be due to a difference in PE. We expected more immersive paradigms to induce higher levels of presence in participants. We expected higher levels of presence to result in a ‘realer’ experience of

(9)

effort, resulting in higher PE, and therefore would express itself in participants choosing low effort choice alternative more often. We took separate measures of presence and PE to verify whether changes in decision-making behavior would indeed be due to changes in presence and PE.

To investigate whether participant show higher rates of DD in more immersive

paradigms we compared the effect of delayed effort on decision-making behavior to the effect of immediate effort on decision-making behavior. We expected delayed effort to induce DD in participants, which would lead to lower PE, which would express itself in participants choosing high effort choice alternatives more often. We expected immediate effort to have the opposite effect, which would lead to higher PE, which would lead to participants choosing low effort choice alternatives more often. We expected these effects to be stronger in more immersive paradigms.

Method

Participants

The sample consisted of 39 participants, of which 20 were recruited through the university website and 19 were recruited through verbal advertisement and posters. Participants were offered 20 Euros or 2 university research credits for participating.

Materials

All tasks were performed on a standard university lab computer. The 2D and 3D task were displayed on a 1680*1050 resolution monitor. The VR condition was displayed through the Oculus Rift DK2. An Xbox One controller (Microsoft, 2014) was used to operate the task. All tasks were custom built in Unreal Engine 4 (Epic Games, 2005).

Presence was assessed with the Igroup Presence Questionnaire (IPQ) (Shubert, Friedmann & Regenbrecht, 2001). PE was assessed with the Borg’s rating of perceived exertion and pain scales questionnaire (Borg, 1998). Listening span was assessed with a Dutch translation of the Listening span test (Vos, Gunter, Kolk & Mulder, 2001), which participants listened to through headphones. Other questionnaires in this experiment were conducted by colleagues. These were not relevant for this particular study, and therefore will not be discussed. These questionnaires were the Locus of Control Questionnaire, Temporal Experience of Pleasure Questionnaire, Social-Economical Status Questionnaire, SQUASH Questionnaire, Immersive Tendencies Questionnaire, and Game Experience Questionnaire.

(10)

Procedure

We always tested two subjects simultaneously during two one hour blocks in adjacent rooms. Both blocks started off with five minutes of starting up, reading the information sheet, and signing the informed consent. The test block proceeded with five minutes of task

instructions, practicing, and time to ask questions. This was followed by 45 minutes of tasks (15 minutes per condition). The order in which the three tasks were performed was

counterbalanced to prevent sequence effects. The last part of this block consisted of filling in the IPQ. The questionnaire block proceeded with 25 minutes of listening span task, and was followed by approximately 30 minutes of questionnaires. Each of two participant followed the blocks in a different order, automatically counterbalancing the order of the procedure.

Task

At the start of each trials, two choice alternatives were offered to the participant

(Figure 2): a high effort option and a low effort option. Choosing the high effort option would result in a high reward, choosing the lower effort option would result in a low reward. Reward was represented as a number. Effort was represented as blocks (Figure 2). These blocks represent track parts. Green blocks represent low effort track parts, orange blocks medium effort track parts, and red block high effort track parts. Medium blocks require twice as much effort as low blocks, and high blocks require twice as much effort as medium blocks. The longer a block was, the faster the cart moved across it. This way all blocks required the same amount of time to pass. This was done to keep the time duration between choice alternatives equal, in order to prevent time from being a confounding factor in decision-making.

Figure 2. In task choice alternatives, taken from 3D condition. On the left is the high effort option, on the right

(11)

Trials were designed by psuedorandomly generating two strings of five items per trial, with each having the value 1, 2 or 4. These two strings represented the two choice alternatives for each trial. The strings were constrained so that their sum was always unequal to the

corresponding counterpart.

The difference in reward between the two choice alternatives was calculated by an automatically updating valuation function (Figure 3), based on the difference in effort between each track and the participant’s choice on each track: reward/effort modifier * difference effort current trial * difference effort next trial (Tversky & Kahneman, 1992). The outcome of this function was added to 1000 for the high reward, and subtracted from 1000 for the low reward.

Phase 1: In this phase the goal was to create a staircase procedure in which

participants, depending on their level of PE, would chose exclusively either the high reward or the low effort option until the point where they would choose the option in the opposite direction. This phase lasted for 10 trials. For the first five trials the reward/effort modifier decreased from 10 to 2 in steps of 2 to fine tune participants’ responses.

Phase 2: In this phase the goal was for participants to have reached a point where the reward per unit effort made the participant feel indifferent about the choice alternative: the point of indifference (POI). This would be the point where participant’s choices would alternate between the high and low effort option, and was our measure of decision-making behavior. The average reward per unit effort in this phase was taken to calculate POI1.

Phase 3: To compare the effect of immediate effort and delayed effort on POI

participants were randomly assigned to two groups: the High Late (HL) group and High Early

(12)

(HE) group. In this phase, which lasted five trials, the HL group received choice alternatives in which the high effort option required effort to be exerted with a delay (late) and the low effort option required effort to be exerted immediately (early). The HE group received choice alternatives in which the high effort option required effort to be exerted immediately, and the low effort option required effort to be exerted with a delay. In the task ‘late’ effort was centered on the second half of the track, and ‘early’ effort was centered on the first half of the track. This phase was aimed at creating a difference in effort delay between groups, which would result in a decrease of POI for the HL group, and an increase of POI for the HE group. The average reward per unit effort in this phase was taken to calculate POI2.

There were three conditions 2D, 3D and VR, representing ascending levels of immersion. In the VR and 3D condition participants navigated through a virtual world in a train cart across a track. Participants were asked to make fists and place these on the joysticks of the Xbox One controller. To make a choice, both joysticks were to be moved to the

direction of the desired choice alternative. After participants had made their choice they were ought to move the two joysticks in opposite direction across a forward backward axis to exert effort to propel the train cart. A non-recording webcam was aimed at the participants’ hands. The participants were told that their performance would be monitored through this webcam in order to stimulate proper execution of the task. The participants saw a small power bar in their field of view. This power bar represented the exerted effort. They were instructed to always keep the power bar at least half full. When they crossed effortful track parts, which were represented as bushes and grass, the power bar drained faster, thus requiring the participants to exert more effort to keep the power bar above the desired level (Figure 4).

The 2D condition consisted only of schematic representations. The mechanics were be the same. The visual feedback of task progress was reduced to a blue progress bar filling up over a schematic representation of the track (Figure 5).

(13)

Results

Analysis

From the total sample of 39 participants 17 participants were excluded. Two

participants were excluded for missing data due to technical issues, 15 for qualifying as non-responsive (average POI1 < 3) reducing the final sample to 22 participants. Participants were randomly divided between the HL group (n = 12) and the HE group (n = 10).

An average POI lower than 3 implies that these participants neglected effort and made their choices based solely on reward, qualifying them as non-responsive to our effort

manipulation. The effort manipulation on our responsive participants turned out to be insufficient, resulting in weak data. Our small sample contributed to the data being even weaker. These two factors together contributed to the data not meeting the assumptions of normality and homogeneity needed to perform proper data analysis. This was a pilot study with educational purposes. In order to gain maximum understanding of the data we decided to ignore these assumptions and run data analysis as planned. Therefore, these assumptions will not be mentioned in the results.

Effort Based Decision-Making

A mixed design was used to analyze the effect of condition on POI, while controlling for counterbalancing order. Mauchly’s test indicated that the assumption of sphericity had been violated, X ² (2) = 31.25, p < .001, therefore Huynh-Feldt statistics were used. In line with expectations there was no interaction effect between condition and counterbalancing order, F(10, 32) = 1.36, p = .242 . Contrary to expectations there was no main effect for condition on POI, F(2, 32) = .044, p = .957. (Table 1; Figure 6).

Figure 5. Trial progress taken from 2D condition. In the middle is a schematic representation of the track. This

(14)

Table 1.

Average scores on POI1, POI2 presence and PE and standard deviations (between brackets) for the 2D, 3D and VR condition.

Figure 6. Average scores on POI1 for the three conditions with standard error of the mean, computed over

within-subject normalized data.

Before analyzing covariance of presence and PE in the effect of condition on POI, a manipulation check was done to assess whether there was an effect of condition on both presence and PE. A repeated measures design was used to analyze the effect of condition on presence. Mauchly’s test indicated that the assumption of sphericity had been met, X ² (2) = 5.28, p = .071. There was a main effect for condition on presence, F(2, 42) = 157.56, p < .001, (Figure 7). In line with expectations planned comparisons showed that this relation was linear, F(1,21) = 351.24, p < .001, (Table 1). Another repeated measures designs were used to analyze the effect of condition on PE. Mauchly’s test indicated that the assumption of

sphericity had been violated, X ² (2) = 8.66, p = .013, therefore Huyn-Feldt statistics were used. There was a main effect for condition on PE, F(1.57, 42) = 5.07, p = .018, (Table 8). In line with expectations planned comparisons showed that PE in the VR condition was

significantly higher than in the 3D condition, F(1, 21) = 9.07, p = .007. Contrary to

expectations planned comparisons showed that PE in the 3D condition was not significantly higher than in the 2D condition, F(1, 21) = .85, p = .773, and that there was no significant difference in PE between the VR and 2D condition, F(1, 21) = .51, p = .483, (Table 1).

10 12 14 16 18 20 22 24 26 2D 3D VR Po int of i nd if ferenc e Condition

POI1 POI2 Presence PE

2D 3D VR 18,32 (16,85) 17,00 (17,86) 17,50 (21,24) 16,82 (3,94) 16,32 (3,95) 16,55 (4,80) 21,91 (4,97) 34,69 (7,80) 49,50 (5,31) 1,52 (1,43) 1,45 (0,99) 2,52 (1,7)

(15)

Figure 7. Average scores on presence for the three conditions with standard error of the mean, computed over

within-subject normalized data.

Figure 8. Average scores on PE for the three conditions with standard error of the mean, computed over

within-subject normalized data.

To analyze the covariance of presence and PE in the effect of EBDM condition on POI the score differences between VR-3D, 3D-2D, and VR-2D of presence, PE and POI1 were correlated using Pearson correlations. Contrary to expectations, there were no significant correlations between POI and presence or PE (Table 2).

0 10 20 30 40 50 60 2D 3D VR P erc ei v ed ef fort Condition 0 0.5 1 1.5 2 2.5 3 2D 3D VR P erc ei v ed ef fort Condition

(16)

0 10 20 30 40 POI1 POI2 P oi nt of ind if ferenc e HL HE Table 2.

Correlations and significance levels of score differences between the 2D, 3D and VR condition of POI1, Presence and PE.

Presence 2D-3D 3D-VR 2D-VR PE 2D-3D 3D-VR 2D-VR POI1 2D-3D Correlation 0,020 0,174 Significance 0,930 0,439 3D-VR Correlation 0,103 0,012 Significance 0,647 0,959 2D-VR Correlation 0,282 0,211 Significance 0,204 0,346 Delay Discounting

A mixed design was used to analyze the effect of group on difference between POI1 and POI2, in all three conditions. Mauchly’s test indicated that the assumption of sphericity had been met for the interaction between difference in POI, condition and group, X ² (2) = 1.51, p = .471. There was no interaction effect between group, condition and the difference in POI, F(2, 40) = .722, p = .483, indicating that there was no difference between POI1 and POI2 for any of the groups, in any of the conditions (Table 3). These differences are illustrated in Figure 9, 10, and 11.

0 5 10 15 20 25 30 35 POI1 POI2 P oi nt of i nd if ferenc e HL HE 0 10 20 30 40 POI1 POI2 P oi nt of i nd if ferenc e HL HE

Figure 9. Average scores on POI1 and POI2 for the HL and HE

group in the 2D condition.

Figure 10. Average scores on POI1 and POI2 for the HL and

HE group in the 3D condition.

Figure 11. Average scores on POI1 and POI2 for the HL and

(17)

Table 3.

Average scores on POI1 and POI2 and standard deviations (between brackets) for the HL and the HE group.

Explorative measures

Figures 9, 10 and 11 show that POI2 was lower than POI1 for all groups and all conditions. These differences are not significant, however, they are curious. One would expect these differences not all to be in the same direction. POI2 is on average lower than POI1 in all conditions, and there is less differences between conditions for POI2 than for POI1 (Table 1; Figure 12). These differences are not significant, but arouse the suspicion that over time POI would have regressed to an average of 0, meaning that over time participants would tend to neglect effort and make decision based solely on reward.

10 12 14 16 18 20 22 POI1 POI2 P oi nt of i nd if ferenc e 2D 3D VR POI1 POI2 HL (n = 12) HE (n = 10) 2D 3D VR 2D 3D VR 25,42 (18,65) 15,25 (15,24) 10,50 (14,58) 9,80 (9,44) 19,1 (21,25) 25,90 (25,47) 24.67 (20.92) 14,25 (13,06) 10,67 (14,67) 7,40 (9,00) 18,8 (24,11) 23,60 (28,64)

(18)

Discussion

This pilot study was aimed at exploring the applicability of VR in experimental psychology by investigating the effect of more immersive paradigms on PE and rates of DD. Numerous issues have been encountered setting up this study, leading to unsuccessful manipulations, and weak data. This however did offer valuable insight in the factors that come in to play in new paradigm like this. Therefore instead of discussing the applicability of VR in experimental psychology, we discussed the pitfalls that have to be considered while setting up an experimental paradigm like this.

Effort based decision-making

More immersive paradigms were expected to induce more presence in participants, leading to higher PE, expressing itself in decision-making behavior in which participants would choose low effort choice alternatives more often. As expected participants experienced higher levels of presence in more immersive paradigms, indicating that these successfully differed in level of immersion. Contrary to expectations there were no differences in making behavior between paradigms. Analysis of PE to explain the results on

decision-making behavior yielded mixed results. As expected participants experienced higher PE in the VR condition compared to the 3D condition. Against expectations, however, participants did not experience higher PE in the 2D condition compared to the 3D condition. These

differences in PE could not account for the results on decision-making behavior.

A possible explanation for the unexpected results on PE is that participants found the 2D condition mentally exhausting. This might have resulted in higher PE in the 2D condition, causing it to be equal to PE in the 3D condition. This however does not explain why there is no difference in decision-making behavior between paradigms.

Decision-making behavior was based on PR and PE. We expected PR to be equal between paradigms, and PE to be higher in more immersive paradigms. The joysticks on the Xbox One controller used to operationalize effort did not induce sufficient effort in

participants. This can be traced back to the high rate of non-responsive participants which completely neglected reward and made their decision solely based on reward. When

comparing differences in decision-making behavior over time, data aroused the suspicion that over time responsive participants also tend to neglect effort and make choices based solely on reward. This indicates that participants did not perceive sufficient effort for a successful manipulation of PE. This means that there was a floor effect for PE. Because of this floor effect PE did not influence decision-making behavior, leaving PR the only factor that contributed to decision-making behavior. PR was equal between paradigms, explaining why there were no differences in decision-making behavior between paradigms.

Future research should therefore be aimed at developing a valid operationalization of effort. This could be done by using two Logitech ATK3 joysticks (Logitech, 2005). These joysticks enable operation of the task with high action fidelity. They offer considerably more resistance than the joysticks on the Xbox One controller. It would also be possible to attach weights to the wrists, or use elastics to increase resistance to acquire the desired level of

(19)

effort. This study has taught that whatever the operationalization of effort is, it should be piloted extensively. Only by doing proper testing on multiple subjects one can get an estimate of the desired level of effort to be induced in participants to perform a successful

manipulation of PE.

To eliminate confounding factors between paradigms in future research it therefore is an idea to set up three conditions with the Oculus Rift, and have these differ only in level of immersion by reducing action fidelity. This could be done by leaving the VR condition the way it is with corresponding head movements in all directions, having the condition with middle level of immersion only allow 360 degrees horizontal head movements, and have the condition with least immersion allow no head movement.

Delay Discounting

Delayed effort was expected to induce DD in participants, leading to lower PE, which would express itself in differences in decision-making behavior in which participants would choose high effort choice alternatives more often. Immediate effort was expected to have the opposite effect, leading to higher PE, expressing itself in differences in decision-making behavior in which participants would choose low effort choice alternatives more often. These effects were expected to be stronger in more immersive paradigms. Against expectations neither immediate nor delayed effort induced DD in participants in any of the paradigms.

This might have been due to the repetitive nature of the tasks. We intended for participants to perceive delayed effort as being further away in time, and immediate effort to as closer by in time. However, participants are aware that there is also effort to be exerted in future trials. This awareness creates a temporal perception of effort to be exerted in the task as a whole, and not on separate trials. This perception prevents a delay of effort within one trial to actually be perceived as further away, since the participant already anticipates the effort that has to be exerted on the next trial.

Future research on the effect of more immersive paradigms on DD should therefore be conducted in a paradigm without repetitive exertion of effort. When effort is operationalized with hypothetical effort that has to be exerted in the future beyond the task, this repetition problem would be avoided. One could then look at the effect of presenting this hypothetical effort in paradigms with different levels of immersion, to assess its effect on DD.

Conclusion

This first time study has shown that more immersive paradigms induce higher levels of presence in participants. The next step is to explore whether these higher levels of presence have an effect on decision-making behavior. Adding different levels of immersion to a

decision making paradigm adds to the complexity of setting up an experiment. These complexities create extra pitfalls to consider when designing a paradigm. Such paradigms should therefore be tested and retested extensively before going operational, to avoid fatal errors.

(20)

We have demonstrated that it is possible to integrate immersion and presence in experimental psychology through the use of relatively accessible hardware. With VR making promising advancements in clinical psychology (Difede, Hoffman & Jaysinghe, 2014;

Cuthbert, Staniszewski, Hays, Gerber, Natale & O'Dell, 2014) it should not be long before VR becomes more accessible as a research tool. Hopefully this will inspire the research community to explore VR as a tool in experimental psychology. Perhaps one day leading to experimental psychology no longer being criticized for not measuring real behavior, through VR enabling us to simulate any kind of realistic behavior, and measure it.

“What is real? How do you define 'real'? If you're talking about what you can feel, what you can smell, what you can taste and see, then 'real' is simply electrical signals interpreted by your brain.” (The Matrix, 1999)

Literature

Araujo, D., Davids, K., & Passos, P. (2007). Ecological validity, representative design, and correspondence between experimental task constraints and behavioral setting: Comment on Rogers, Kadar, and Costall (2005). Ecological Psychology, 19(1), 69-78.

Bohil, C. J., Alicea, B., & Biocca, F. A. (2011). Virtual reality in neuroscience research and therapy. Nature reviews neuroscience, 12(12), 752-762.

Borg G. A. V. (1998) Borg’s rating of perceived exertion and pain scales. Champaign, IL: Human Kinetics, viii-104.

Bowman, D. A., & McMahan, R. P. (2007). Virtual reality: how much immersion is enough?. Computer, 40(7), 36-43.

Croxson, P. L., Walton, M. E., O'Reilly, J. X., Behrens, T. E., & Rushworth, M. F. (2009). Effort-based cost–benefit valuation and the human brain. The Journal of

neuroscience, 29(14), 4531-4541.

Cuthbert, J. P., Staniszewski, K., Hays, K., Gerber, D., Natale, A., & O'Dell, D. (2014). Virtual reality-based therapy for the treatment of balance deficits in patients receiving inpatient rehabilitation for traumatic brain injury. Brain injury, 28(2), 181-188. Difede, J., Hoffman, H., & Jaysinghe, N. (2014). Multimedia Reviews: Innovative Use of

Virtual Reality Technology in the Treatment of PTSD in the Aftermath of September 11. Doya, K. (2008). Modulators of decision making. Nature neuroscience, 11(4), 410-416. Epic Games. (2012). Unreal Engine 4. Retrieved from

(21)

Kirby, K. N., & Maraković, N. N. (1996). Delay-discounting probabilistic rewards: Rates decrease as amounts increase. Psychonomic bulletin & review, 3(1), 100-104.

Logitech. (2005). Logitech Attack 3. Retrieved from http://support.logitech.com Microsoft. (2013). Xbox One Wireless Controller. Retrieved from

www.xbox.com/en-US/xbox- one/accessories/controllers/wireless-controller

Oculus VR. (2014). Oculus DK2. Retrieved from https://www.oculus.com/dk2/

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.

Sanchez-Vives, M. V., & Slater, M. (2005). From presence to consciousness through virtual reality. Nature Reviews Neuroscience, 6(4), 332-339.

Schubert, T., Friedmann, F., & Regenbrecht, H. (2001) The experience of presence: factor analytic insights. Presence: Teleoperators & Virtual Environments, 10, 266-281. Soman, D. (2004). The effect of time delay on multi-attribute choice. Journal of economic

psychology, 25(2), 153-175.

Soman, D., Ainslie, G., Frederick, S., Li, X., Lynch, J., Moreau, P., ... & Zauberman, G. (2005). The psychology of intertemporal discounting: Why are distant events valued differently from proximal ones?. Marketing Letters, 16(3-4), 347-360.

The Matrix, 1999. [film, DVD]. Directed by Andy Wachowski and Lana Wachowski. USA: Warner Bros

Treadway, M. T., Buckholtz, J. W., Schwartzman, A. N., Lambert, W. E., & Zald, D. H. (2009). Worth the ‘EEfRT’? The effort expenditure for rewards task as an objective measure of motivation and anhedonia. PLoS One, 4(8), e6598.

Trope, Y., & Liberman, N. (2003). Temporal construal. Psychological review,110(3), 403. Tversky, A., Kahneman, D. (1992). Advances in prospect theory: Cumulative representation

of uncertainty. Journal of Risk and Uncertainty, 5, 297-323

Witmer, B. G., & Singer, M. J. (1998). Measuring presence in virtual environments: A presence questionnaire. Presence: Teleoperators and virtual environments, 7(3), 225-240.

Vos, S. H., Gunter, T. C., Kolk, H. H. J., & Mulder, G. (2001). Working memory constraints on syntactic processing: An electrophysiological investigation. Psychophysiology, 38, 41–63.

Referenties

GERELATEERDE DOCUMENTEN

The same steps were followed in order to build the criteria tree for the second analysis (deep-seated landslides susceptibility): a) large landslides scarps and bodies were identified

The neural blackboards allow the construction of (potentially novel) combinatorial structures based on (familiar) in situ concept representations, using forms of variable

Through a combination of experimental measurements and discrete particle simulations, we have investigated the influence of particle geometry on the segregative behaviors

In this study, cytochrome P450 CYP109A2 from Bacillus megaterium DSM319 was expressed, purified and shown to oxidize vitamin D 3 with high regio-selectivity..

If a mutation in the BRCA1 and/or BRCA2 gene is associated with a lower ovarian reserve, this may have a negative effect on success chances of mutation carriers undergoing IVF

Si mes services ici pourroit être de quelque utilité soit à vous, Monsieur, ou à Messieurs d’Imhof et de Bulow, je vous supplie de faire parvenir vos ordres à celui qui a

Furthermore, the increase of carrier traps and the mobility characteristics in our devices with high-k dielectrics are an indication of the existence of polaronic-related interaction

Ons denken, voelen en gedrag veranderen er door, passen zich aan en beïnvloeden haar op hun beurt – in de zorg en ook daar buiten.. Denk aan hoe social media ons gedrag