The proliferation of various remote communication devices such as mobile phones in recent years has turned them into an essential part of our daily lives. To induce the positive experience in remote communication beyond the role of merely information exchanges, efforts have been made to explore the use of various additional sensory modalities in communication device such as visual information in video communication (Isaacs and Tang, 1994; Finn et al., 1997; De Greef and Ijsselsteijn, 2001).

In addition to visual information, the uses of tactile stimuli in communication devices to contribute the positive experience of users in remote conversation has recently gathered a great deal of attention. From a psychological perspective, it has been suggested that tactile sensations play an important role in building rapport in interpersonal communication (Gallace and Spence, 2010; Tatsukawa et al., 2016). Thus, there have been attempts to integrate tactile functions into various media (Chang et al., 2002; Prattichizzo et al., 2010). For instance, it has been reported that the level of mutual closeness experienced by participants during a video chat increased when they had artificial contact in the form of a handshake with a robotic hand (Nakanishi et al., 2014). Furthermore, an attempt has been made to attach a device to a smartphone to enable users to experience an artificial sensation of kissing during remote communication (Samani et al., 2012). However, it should be noted that most studies evaluating the serviceability of these devices were conducted using controlled, predetermined conversations. In other words, very few studies have explored the psychological and behavioral effects of remote communication devices with tactile sensations using conversational scenes allowing free utterances. Therefore, we posit that while the market explores commercial applications for remote communication devices with tactile functions, serviceability studies using unconstrained communication are necessary.

Hugvie is a human-shaped cushion aimed at providing users with the tactile stimulation of hugging during phone conversations in order to improve positive feelings (e.g., closeness and trust) toward each other (Sumioka et al., 2013, 2014; Yamazaki et al., 2016), and its serviceability has been demonstrated using free-conversation scenes (Figure 1). A psychological study reported that conversations with a female while hugging Hugvie stimulated male users’ feelings of closeness toward her (Kuwamura et al., 2014). Furthermore, Sumioka et al. (2013) compared the level of cortisol, a hormone positively correlated with stress levels, between participants using Hugvie and those using a basic mobile phone during an unconstrained remote conversation, and showed that the concentration of cortisol in blood and saliva decreased in those using Hugvie. This result is extremely compelling, as it suggests that Hugvie users were more relaxed based on a physiological indicator. These psychological and physiological data have suggested that using Hugvie has positive effects on mutual closeness during a remote conversation. In addition to these findings, we expect that Hugvie has a strong effect in the formation of various social attitudes. Especially, we hypothesize that Hugvie enhances the trust of users, because previous psychological findings have suggested that the sense of interpersonal touch is strongly linked to the formation of trust (Gallace and Spence, 2010; Walter et al., 2013). If we investigate this hypothesis, a good experimental task that allows free-conversation and enabled to quantify the trust is required.

Conversation games allow free-flowing conversations between players; however, controlled responses according to the game procedure can also be objectively observed. Hence, these games are useful to validate the effects of Hugvie on trust during a free conversation, as it possible to quantify controlled responses in the game. The present study utilized a conversation-based game known as Werewolf, which was modified to evaluate the level of trust between players. One group of participants used Hugvie to play the game with remote players, while the other participant group used a basic communication device. Changes in participants’ level of trust toward other players were compared between the two groups. Collected data involved controlled responses (vote behavior) during the game and responses to a questionnaire. One of the advantages of using this type of conversation-based game was that although participants could talk freely, we could easily gauge their state of mind based on the actions invariably required to play the game. Although there are always winners and losers in the regular version of Werewolf, the game rules were modified in the present research to include possible scenarios in which there was no enemy and all players would win if they trusted one another. This modification allowed us to evaluate the potential increase in mutual trust caused by using Hugvie.

Firstly, we verified that identities had been assigned equally in the two experimental groups. In the Hugvie group, “villager” identities accounted for 70%, “seer” for 17%, and “werewolf” for 13%. In the control group, “villager” identities were assigned to 62% participants, “seer” to 18%, and “werewolf” to 20%. There was no significant difference in the proportion of identities between the two groups [chi-square test, X²(2) = 0.452, p = 0.8, Cramer’s V = 0.069].

Subsequently, we compared the ratio of “No werewolf” votes between the two groups. Specifically, we calculated the ratio of participants who voted “No werewolf” in each round and group (Figure 4). The ratio of “No werewolf” votes was found to be higher in the Hugvie group than in the control group only in the 4th round [chi-square test, X²(1) = 2.51, p = 0.012, φ = 0.32, stochastically significant after the Bonferroni multiple-comparison correction]. Furthermore, we also compared the overall mean ratio of “No werewolf” votes between male and female subjects and we could not find the significant difference between two [independent t-test, t(22) = -0.801, p = 0.43, d = 0.38].

Additionally, we examined how the level of trust toward other participants changed during repetitive rounds. The choice of “No werewolf” votes was set as the dependent variable representing participants’ trust behavior in a round, with a value of 1 when a participant voted “No werewolf” in the round and 0 when the participant did not vote it. Subsequently, we estimated a linear regression model per subject predicting the change in the dependent variable according to round number (explanatory variable). The slope coefficient of this regression model indicates how the level of trust changed through repetitive rounds. Positive slope values indicate that the level of trust increases during repetitive rounds while negative slope values mean that it decreases. We compared the means of the slope coefficients between the two groups and found that the value in the Hugvie group was significantly higher than that in the control group [Figure 5, independent t-test, t(22) = 2.181, p = 0.04, d = 0.89]. This suggests that the level of trust toward other participants in the Hugvie group was relatively maintained in comparison with that in the control group.

Furthermore, we analyzed how the performance in a previous round affected participants’ trust behavior in the current round and found that the ratio of “No werewolf” votes after a round without a werewolf (i.e., a werewolf did not exist in the prior round) was significantly higher than the ratio of “No werewolf” votes after a round with a werewolf (i.e., a werewolf existed in the prior round) in the Hugvie group [chi-square test, X²(1) = 7.58, p = 0.0018, φ = 0.40, stochastically significant after the Bonferroni multiple-comparison correction], but not in the control group [chi-square test, X²(1) = 5.19, p = 0.311. φ = 0.15]. We also found that the ratio of “No werewolf” votes after a round without a werewolf was significantly higher in the Hugvie than in the control group [chi-square test, X²(1) = 7.58, p = 0.0018, φ = 0.40, stochastically significant after the Bonferroni multiple-comparison correction]; however, there was no significant difference in the ratio of “No werewolf” votes after a round with a werewolf between two groups [chi-square test, X²(1) = 5.19, p = 0.311. φ = 0.15]. These results suggest that participants using Hugvie tended to have a biased belief that no werewolf existed in the current round after a round without a werewolf, although the identities of participants in the current round were independent from those in the previous round (Figure 6).

The winning percentage for the villagers was 37.5% in the Hugvie group and 62.5% in the control group. There was no significant difference in winning percentages between the two groups [chi-square test, X²(1) = 1.41, p = 0.32. φ = 0.28].

There was no significant difference of participants’ impressions of other players between two groups in any items of pre-experiment questionnaire. This means that the first impression of other players were controlled between two groups. The comparison of pre- and post-experiment showed that only the Hugvie group displayed a significant decrease in the post-test score on trustworthiness [Figures 7, 8 paired t-test, t(11) = 4.08, p = 0.0018, d = 1.49, stochastically significant after the Bonferroni multiple-comparison correction].

The present research sought to verify the hypothesis that using Hugvie, a remote communication device combining sound with tactile sensations, maintains the level of trust that a user feels toward his/her conversation partner in an ambiguous situation. We used a game called Werewolf with modified rules for the purpose of this study, creating the possibility that no werewolf existed among the players. When playing the game remotely, the likelihood of players guessing that others were not werewolves was maintained when using Hugvie as opposed to using a basic sound-based communication device. These results support our hypothesis that using Hugvie maintains the level of trust of a user feels toward his/her remote conversation partner. We believe that using Werewolf, a game based on free-flowing conversation, allowed us to obtain such a result.

However, our study also showed that using Hugvie did not increase the villagers’ winning rate. In fact, the winning rate of Hugvie users was lower than that of the control group, suggesting that the level of trust that Hugvie users felt toward the other players was not based on reasoning. Research in social psychology has developed a scale to measure individuals’ general trust toward others (Yamagishi et al., 2015). The kind of trust that increased by using Hugvie may be general, rather than specific. The results of the present study also demonstrated that participants in the Hugvie group felt a significantly lower level of trust toward the other players after the experiment, which is presumably due to the villagers’ low winning rate. In other words, these participants may have regretted having trusted the others after seeing that it did not lead to the desired outcome, which subsequently lowered their level of trust. It can therefore be surmised that Hugvie’s positive effect on the level of trust toward others in remote communication may backfire if the situation becomes less agreeable, leading to users feeling regretful when remembering previous conversations. Consequently, this device may be more useful in natural daily conversations in which neither person has a vested interest. In a negative sense, our results might suggest that using Hugvie in remote communication leads to decreased thinking and might generate blind obedience to other persons. For example, in Japan, frauds targeting elderly people using mobile phones become a social problem. Thus, it is important to consider from an ethical viewpoint how Hugvie-like remote communication devices may alter our communication styles.

Hugvie is a remote communication device that provides the user with the sensation of a hug in addition to audio communication. However, the following question remains: to whom does the user attribute this sensation? For example, if two lovers are in conversation using this device, it is likely that they imagine each other as the one they are hugging, although the hugging sensation is not actually transmitted from their partner via Hugvie. In our experiment, however, the conversations involved three individuals; therefore, it was more difficult to attribute the hug sensation to a specific individual. The trust obtained through Hugvie may represent a generalized feeling caused by the physiological sensation of a hug during remote communication, rather than it being attributed to a specific individual. Research in embodied cognitive science suggests that the state of a person’s body significantly influences the way in which he/she views the outside world (Ballard et al., 1997; Ybarra et al., 2000; Tschacher and Bergomi, 2015). The results of this study suggest that the physical posture of hugging along with the sensation of being hugged might influence feelings toward remote conversation partners without necessarily attributing them to a specific individual. This might be likened to having a remote conversation with someone while holding a baby, a family pet, or a security blanket. Of course, this interpretation is too speculative at present and some researchers may doubt the effects of embodiment cognition (Ranehill et al., 2015). Hence, further detailed studies need to be conducted to reveal how Hugvie affects trust levels.

In this study, we must be careful to over-generalize the results, because sample sizes were not so large. Especially, many previous studies have reported that both gender and social relationships (e.g., friends or strangers) strongly affect the level of trust toward others (Croson and Buchan, 1999). In the current study we could not find a salient difference between male and female participants, with our preliminary analysis suggesting that no significant differences in the ratio of “No werewolf” votes existed between female and male participants. However, the sample size in the current study was too small to confirm a gender difference, although gender and social relationship proportions were matched between the Hugvie and control groups. Future research is needed to investigate how gender and relationships influence the effect of Hugvie in remote communication.

Finally, we would like to emphasize that using conversation games such as Werewolf is useful to verify the function and validity of remote communication systems, as they are enjoyable for participants in comparison with conventional experimental tasks and enable the analysis of natural utterances and emotional expressions generated while playing. Recently, there are many communication devices using various communication channels, such as tactile stimulation, smell, light, text characters, and emotional icons. Conversation games might provide an experimental paradigm to uniformly investigate how these various devices alter our social attitudes and emotions during remote communication.

HT, MB, HO, JN, HS, and HI designed the research. HT, MB, JN, and HS performed the research. HT and HS wrote this paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.