Keywords

1 Theoretical Context

1.1 Serious Virtual Environments

VR and AR systems are becoming an integral part of our world. It is no longer just about research environments, computer games or experimental interfaces. Like massively multiplayer online games and social networking sites in the last three decades, nowadays virtual reality develops rapidly driven by the users’ creativity and collaborations. We are dealing with mass social environments, entertainment and training applications, immersive apps improving the health and well-being of users, or standard applications with new, 3D interfaces. The growing number of users of virtual environments (VE) means that they more and more often contact each other through these platforms where they also meet their inhabitants - artificial agents of various forms. Representations of VR users are most often (but not necessarily) virtual humans, or at least humanoids (especially in the case of some environments that have a simplified, even cartoonish form). Also, virtual agents often take the humanoid form - in this case, we deal with artificial virtual humans. Advances in technology have made it sometimes difficult to recognize whether we deal with a human or an algorithm shaped as one. The VE users see in front of them just a virtual character with which they can interact, as in the real world. Even relatively simple computer interfaces can cause users to behave in a manner characteristic of interaction with another human being, for example prompting them to apply rules of politeness [13, 14]. When interacting with (artificial) virtual humans in VE, this is even easier to observe due to the variety of possible behaviours that can be simulated [11], which stands in line with the media equation theory [17] and the media richness theory [4]. Research on social influence and compliance to requests demonstrated that virtual humans in VEs influence real people. We know that people comply in VE as they do in real life, but the authors devote less attention to the mechanisms of this influence and the possible consequences. In this context, it is also worth paying attention to the development of augmented reality (AR), in which virtual elements are mixed with the real world and robotics, in particular humanoid social robots that slowly enter physical reality in the form of surrogates (which can be considered a physical equivalent to avatars in VR) and autonomous robots (analogues of agents in VR).

1.2 Social Influence Techniques

Social influence is understood as changing the attitudes, perceptions, emotions, and behaviour of others [1, 2]. Many classic experiments have been replicated in VE. For example, a) social facilitation/inhibition effects, that is, improving the performance of simple tasks and worsening the performance of complex tasks in the presence of virtual humans [15, 18]; b) effectiveness of persuasion (increased by the level of realism of the virtual human and gender compatibility with user) [10]; c) conformity (giving answers in line with virtual agents, i.e. giving incorrect responses more often when agents also gave ones compared to the situation in which agents gave correct responses) [12].

Most often, social influence related phenomena are shown through influence techniques [6]. Research on discreet attempts to influence others originated from works of Freedman and Fraser [9], who hypothesised that people who agree to fulfill a small request would be more likely to comply with a more difficult request. In the original experiments, people who agreed to sign a petition to keep California clean, to improve road safety, or agreed to put a small label in the window encouraging compliance with these petitions, were more likely than in the control group to agree to put up an ugly sign in front of their house with the words “drive carefully”. The authors named this foot-in-the-door technique (FITD). Another technique, assuming the reverse order of formulated demands proposed by Cialdini [3], is the so-called door-in-the-face (DITF). Its typical course is as follows: making a large, difficult request followed by a refusal that prompts people to comply if the requester withdraws the first one and makes another, much easier request. In the original study, participants were asked to give their consent to act as guardians of juvenile offenders, which required two hours a week for two years (the vast majority of participants refused to comply with this request). The smaller request that followed was to help during a two-hour trip with a group of those young people to the zoo. Participants who previously refused to fulfil a larger request were more likely to agree to a smaller one than in the control group.

Since both agreement and refusal to comply with the first request may, under certain conditions, lead to compliance, the third technique, which is a compilation of the FITD and DITF, was proposed. The foot-in-the-face technique (FITF) proposed by Doliński [5] also assumes the sequential formulation of two requests, this time with the same degree of difficulty. The response to the first request paves the way for the effectiveness of this technique. Those who agree to fulfil the first request are on the “agreement route” (as in the FITD), and those who refuse to comply with the first request are on the “refusal route” (as in the DITF).

Field experiments, used in classic research on the abovementioned techniques, make it impossible to accurately trace the process that takes place at the time when subsequent requests are formulated. Whereas conducting laboratory experiments allows both the use of repeated measures of self-report variables, continuous observation of behaviour and measurement of psychophysiological variables (e.g. [8]). Research in VE combine a high level of ecological validity of field studies with the precision of control and measurement offered by a laboratory setting. The use of VE makes it possible to recreate the natural behaviour of participants under controlled conditions, including the influence of personality traits, distortions resulting from the appearance and behaviour of avatars, both of the user and partners of social interaction. In addition, it is possible to continuously and precisely record the behaviour of participants (position in space, position in relation to other objects, motor skills) and to record psychophysiological variables (such as electrodermal activity, heart rate variability). Unlike field experiments, it is possible to reproduce the experimental situation for each participant accurately and embed self-report scales in the user interface. Moreover, VEs are able to transcend the spatial and temporal boundaries of the physical world, which further expands the possibilities for the researcher.

1.3 Social Influence Techniques in Virtual Environments

Eastwick and Gardner [7] used unstructured online virtual world There.com to test if the FITD and DITF techniques are efficient also in computer-mediated, virtual settings. It is a bridge between the previously described experiments in natural conditions and laboratory tests using VE. The experimental procedure, therefore, takes place in a VE, but the participants did not submit to it in the laboratory. Instead, they were virtual passersby, devoted to their usual activities. The experimenter, acting as an ordinary user, approached avatars standing alone and started a conversation with them, saying that he was doing a photo scavenger hunt. In the control condition, he then asked to teleport with him to another location and for permission to take a screenshot of the user’s avatar on a virtual beach. In the FITD condition, the moderate request was preceded by an easy-to-meet request: “Can I take a screenshot of you:)?” and in the DITF condition, a difficult-to-meet request, regarding teleporting and taking screenshots in 50 different places, which was supposed to take two hours. Both techniques proved effective. Significantly more participants agreed to a moderate request after agreeing to a small request or refusing a large request. Additionally, the influence of the experimenter’s avatar race in the DITF condition was noted. More people agreed to the request of a light-skinned avatar.

Fig. 1.
figure 1

Participants in virtual environments created in Garry’s Mod (Half Life engine; left) and Minecraft (right)

Full size image

Pochwatko et al. [16] conducted a similar study in laboratory conditions, which allowed for the continuous registration of the physiological arousal operationalized as an increase in galvanic skin response (GSR). They used a specially designed virtual environment, developed using the Half-Life game engineFootnote 1 (see Fig. 1 left). The participants were convinced that in the VE there are other people who perform miscellaneous tasks (searching for and collecting objects of various kinds, taking pictures of other users in marked places or marking boards with their initials). The draw at the arrival to the lab was manipulated in such a way that the participant was given the last of the above-mentioned tasks. Assigning participants a facade task was important as it put them in similar conditions to virtual passersby from previous studies. In order to fulfil the experimenter’s request, they had to stop performing their task, as did the participants in the study of Estwick and Gardner [7]. Requests were formulated by the avatar of the experimenter’s assistant or virtual agent. They concerned taking photos at the meeting point (FITD - easy request), several characteristic places on the map, which required a lot of time and travelling (DITF - difficult request) and taking a photo in one remote location (moderate request and control condition). The classic FITD and DITF effects were not replicated, but other interesting effects were obtained. A different pattern of physiological arousal was observed in the experimental and control groups. In the control group, participants who were less aroused complied, while in the experimental groups, those who were more aroused after the first request did so. In addition, participants agreed to fulfil requests formulated by an experimenter’s assistant avatar more often than by a virtual agent. It is also associated with less agent-induced physiological arousal after formulating the request. As previous attempts to replicate social influence effects in VE have yielded mixed results, we aimed at testing whether they would be effective in certain type of VE, which offers high ecological validity and control of the confounding variables. Thus, we chose a popular multiplayer game, for the experimental scene.

2 Current Studies: Foot-in-the-face (FITF) in Laboratory Virtual Environment

Participants. 259 people participated in the study (170 in the replication of the classic procedure with and without time delay between requests and 89 in conditions with induction of guilt). The research was conducted anonymously; participants were randomly assigned experimental conditions. The experimenter’s confederates were unaware of the hypotheses or the condition in which they were involved.

Procedure. In our current studies, we used a VE created in the online version of Minecraft to deal with the limitations caused by the need to create artificial facade tasks as in the environments previously created in the Half-Life game engine. In Minecraft, the tasks for players are natural and engaging (which is confirmed, for example, by the incredible popularity of the gameFootnote 2). Therefore, there is no need to impose facade tasks that are to simulate the situation of interrupting current activities by the requester. A “survival” mode was used in which participants have to build a simple shelter and gather as many resources as possible to survive in the virtual world (in the case of an experimental situation, the test lasted three days and three nights of game time - about 40 min)Footnote 3. Tests have shown that this task is feasible even for people who do not play computer games. At the same time, unlike the scenarios created in previous studies, the task is natural and so unstructured that it should not overburden the participants. The general scheme of the study was similar to that of Pochwatko et al. [16]. The requests, constituting the experimental manipulation, were adapted to the specificity of the environment. They involved taking photos and video clips in different places and during various activities. Participants were informed that the study was running on a public server (to keep control of the situation, this was, in fact, an isolated server in the local network). Numerous modifications were applied to simulate the presence of many players (false information about number of players when logging in, false messages about the actions of players, NPCs, etc.). The experimenter’s confederate pretended to be a vlogger shooting the footage for his YouTube channel (very popular activity for Minecraft online community). The final request concerned consent to make a clip in which the participant performs simple activities (cuts trees, builds a shelter or digs a tunnel).

As previously shown, FITD and DITF procedures are effective in VEs and lead to compliance. Regardless of whether the participant agrees or refuses to comply with the first request, the probability of consenting to the second request increases. The conclusions from the previous research were used to plan the FITF studies. The procedure used in FITD and DITF proved successful and was only slightly modified, in line with the FITF assumptions - the request sequence was changed: instead of the initial small or very large request, a request of the same degree of difficulty as the target request was used. As in the original study by Doliński [5], previously tested requests were used, which under control conditions are fulfilled by approximately half of the participants.

Additionally, for half of the participants, requests were formulated on the first and third day of game time, which was to promote the so-called “agreement route” or the FITD mechanism. In the case of the second half of the respondents, the requests were not separated in time, which was used to test the “refusal route”, i.e. the DITF mechanism. In another condition, an additional variable was introduced - inducing a sense of guilt if the first request was not fulfilled (see [8]). In addition to the participants’ behaviour, physiological responses were also constantly monitored. The purpose of using these measures was to verify the presence of physiological arousal at different stages of the procedure. Unpleasant arousal, or rather the desire to reduce it, is often mentioned as one of the mechanisms explaining the effectiveness of social influence techniques.

Apparatus. The virtual environment was presented using an Intel Core i7 2.3GHz PC with Nvidia GeForce GTX660, and BenQ short throw stereoscopic projector. GSR was recorded with a Biopac MP150 with an EDA100C amplifier.

Results. Under classic conditions (as in Dolinski 2011 [5], without inducing a sense of guilt), a number of differences in response to requests under specific conditions were observed. As assumed in the control condition, almost exactly half of the respondents agreed to the first request (49%). Both in the condition with postponed and immediate requests, significantly more people agreed to the second request after consenting to the first (agreement route), 85% and 86% respectively: postponed request \(\chi ^2\) = 9.74 p <.001, immediate request \(\chi ^2\) = 8.33 p <. 01. Significant differences were also noted on the refusal route (80% and 83% refusal) respectively: postponed request \(\chi ^2\) = 5.69 p <. 01, immediate request \(\chi ^2\) = 7.59 p <. 01 (Fig. 2). Contrary to expectations, if the first request was refused, significantly fewer people agreed to the second request. The classic FITF effect was therefore not observed. After joint analysis of the agreement and refusal routes, there are no differences between the number of participants agreeing to the target request in the experimental and control conditions (\(\chi ^2\) n.s.). Therefore, the interpretation of physiological responses is limited. Nevertheless, different arousal patterns were observed in the agreement and refusal routes. Among the participants that agreed to the first request, a significant increase in physiological arousal (GSR) was observed immediately after the request was formulated. The arousal decreased faster among those who agreed to the second request. On the refusal route, a significant increase in arousal after the first request was observed only in people consenting to the second request. This result is consistent with previous postulates regarding the presence of unpleasant arousal, which leads to compliance, as we observe the “arousal-consent” pattern. In the guilt induction condition, a compliance effect was observed in the guilt-free agreement route (\(\chi ^2\) = 3.99 p <0.05). No significant effects were observed in the remaining conditions. While there was a tendency to increase compliance in the overall results, the difference did not reach statistical significance. Similar to the previously obtained results, physiological responses in the no-guilt condition indicate an increase in arousal after the first request.

Fig. 2.
figure 2

Compliance rates in FITF experiments

Full size image

Limitations. The occurrence of certain factors limiting the scope of interpretation of the above results cannot be ruled out. First, unlike classical research, participants had to agree to come to the laboratory and use bulky equipment. Thus, The altered self perception effect could occur, i.e. perceiving oneself as a helpful person, which could have influenced their decisions. On the other hand, it may have led to the belief that they have already agreed to many things and do not need to agree to further requests, which would reduce the tendency to help. Another limitation may be the game’s survival mode. On the one hand, it provides an open and ecologically valid task, but on the other hand, it can over-draw attention of participants and make them refuse more frequently.

3 Conclusions and Further Directions

The abovementioned successful replications of the sequential request techniques effect and partial replication of the FITF effect prove the possibility of conducting research on social influence in VEs and in laboratory conditions, which allows for continuous measurements of physiological responses and inference about the effectiveness of the studied techniques on the basis of more objective indicators than in the case of field studies. Still, the specificity of these influences requires further research. The dynamic development of virtual reality technology allows us to hope that such research will soon be possible on a larger scale. The number of users of VR headsets is growing rapidly. It is accompanied by an increase in the quality of devices for wireless recording of physiological reactions. There are also mass virtual social environments. The combination of these three elements will allow us to conduct high-ecologically valid and forget about the limitations caused by the use of environments derived from games. Classic social influence effects are present both in games and VEs, as well as in real life. The use of games and VEs to study them offers the possibility of better control of the experimental situation but also brings limitations that we must be aware of as researchers. Sometimes a game is just a game, but sometimes it is becoming serious.