The role of prior exposure in the likelihood of adopting the Intentional Stance toward a humanoid robot

1.1 Intentional stance toward robots?

Given the robots’ mechanical nature, people should treat them as artificial systems that have been programmed to display specific kinds of behaviors. Following Daniel Dennett’s framework [4,5], people should adopt the Design Stance toward robots and, thus, interpret and explain their behavior with reference to the way that the robots were programmed to behave [4,5]. However, in some contexts, people are also inclined to attribute mental states to robots, in order to explain their behaviors. It would mean that, under certain circumstances, people might adopt the Intentional Stance toward robots [4,5]. The adoption of the Intentional Stance leads to treating robots as intentional agents whose behaviors result from mental (intentional) states such as beliefs or desires ([6,7]; for a review see ref. [8]).

For instance, recent evidence [6] showed that, when presented with a series of various behaviors displayed by a human or by a humanoid robot, participants ascribed intentional states to the humanoid robot to a similar degree as when they observed the human displaying the same behaviors. In a similar vein, Marchesi and colleagues [7] asked participants to rate whether the behaviors displayed by the humanoid robot iCub [9] were motivated by a mechanical cause (such as malfunctioning or calibration, thus referring to the Design Stance) or by a mentalistic cause (such as desire or curiosity, thus referring to the Intentional Stance). Results showed that, overall, people adopted the Intentional Stance toward the humanoid robot to some degree [7]. Interestingly, individual differences among participants seem to play a role in the likelihood of adopting the Intentional Stance, as some of them were more prone to choose the mechanistic explanations overall, whereas some others tended to choose the mentalistic explanations [10]. Thus, the authors argued that the likelihood of adopting the Intentional Stance might depend on several factors, such as individual differences in attitudes toward robots [11,12,13].

1.2 Individual differences in the tendency to adopt the Intentional Stance toward robots

Epley and colleagues conceptualized the role of individual differences in attitudes toward robots and presented a three-factors theory explaining whether, and under which conditions, people are likely to anthropomorphize them [14]. According to the authors, three psychological determinants are crucial for people’s tendency of anthropomorphizing robots: (i) the elicited agent knowledge, namely the amount of knowledge about the other agent that people have access to; (ii) the motivation to understand and explain the behavior of the other agent (i.e., effectance motivation); and (iii) the desire for social contact and affiliation. In line with this, the authors suggest that interpersonal differences in these determinants strongly predict people’s likelihood of anthropomorphizing the other agent (i.e., a social robot), and it would explain why anthropomorphism is so variable [14]. In a similar vein, also Fischer and colleagues underlined the role of individual differences in human–robot interaction (HRI), arguing that there is a considerable amount of interpersonal variation with respect to whether artificial agents are treated as social actors [15]. Recent pieces of evidence further confirmed the role of individual differences in attitudes toward robots [16,17,18]. For example, individual differences in ascribing human-like features to a non-human-like agent predict the degree of moral care and concern related to an agent, the amount of responsibility and trust, and the extent to which the agent can serve as a source of social influence on the self [13,18]. Moreover, differences in the perception of the trustworthiness of a robot resulted to be predictive of humans’ behavior toward robots in a human–robot team [19,20].

In sum, knowing that people’s attitudes toward robots (and, potentially, their tendency to adopt the Intentional Stance) can vary depending on context and individual differences, calls for addressing the question of what factors are at play in this phenomenon. In the present study, we chose to focus on the role of prior exposure to robots, which may be a likely candidate factor affecting the adoption of the Intentional Stance toward them.

Specifically, one hypothesis may be that the more people are exposed to robots, and thus, they gain knowledge about the way robots are programmed and controlled, the less likely they would adopt the Intentional Stance toward robots. In other words, acquiring notions about robots’ functionality and use would allow people to explain their behavior based on the assumption that it is the outcome of a designed system [21]. This hypothesis would be in line with Epley and colleagues’ theory of anthropomorphism [14], according to which repetitive interactions can decrease individuals’ likelihood to attribute mental states (i.e., adopting the Intentional Stance) toward a robot, since the acquired knowledge about it would weaken the motivation to understand its behavior. Notably, a recent study seems to support this view [22]. The authors used a gaze-cueing paradigm to investigate participants’ likelihood of attribution of intentionality as a function of the duration of the exposure to the robot’s repetitive gaze behavior. Results showed that a short exposure had a positive change in participants’ attribution of intentionality. On the contrary, long exposure did not increase the initial likelihood of attributing intentionality toward robots [22].

3.1 Participants

Two samples of Italian high school students were recruited during two different science dissemination sessions at schools. The first sample included students with a scientific–technical background (“Scientific–Technical” sample; N = 41 students from scientific high school, N = 17 from scientific–technical high school, and N = 20 from technical high school, for a total N = 80; N = 2 students did not explicitly state which school they were enrolled, and thus, they were considered as NAs). The second sample included students with an artistic background (“Artistic” sample: N = 56 students from an art high school). All students that took part in the study were adults (age range = 18–19). All participants gave informed consent and declared themselves to be of legal age by ticking the appropriate box in the online form before completing the IST. All participants were naïve to the purpose of the study.

The study was approved by the local Ethical Committee (Comitato Etico Regione Liguria), and it was conducted following the Code of Ethics of the World Medical Association (2013 Declaration of Helsinki).

3.2 Apparatus and stimuli

Students were invited to attend a theoretical lecture about robots’ functionality and use, held by the authors. Before and after the authors’ presentation, participants were asked to complete the IST [7], which was programmed using the online platform SoSci (https://www.soscisurvey.de/) [26]. At the end of the lecture, participants watched two videos. First, they watched a typical HRI in a laboratory context from participants’ point of view; subsequently, they watched the same experimental context from the experimenter’s point of view, controlling the robot through a WoOZ manipulation [23,24].

3.3 Procedure

Participants were asked to complete the first half of the IST before the lecture (“IST-Pre”), whereas the other half of the IST was sent and completed at the end of the presentation, specifically after participants watched the WoOZ videos (“IST-Post”). Notably, the IST-Pre was made accessible the day before the lecture session, whereas access to the IST-Post was available only at the end of the lecture. This was done to ensure that participants filled out the second part of the IST only after attending the lecture and watching the videos.

Given the specific requirements of our study, namely the need to provide the two parts of the IST at different time points and the need to match the anonymized answers of the same participant, we designed a custom solution using the tools provided by SoSci Survey. In detail, we created another survey (called “IST-init”) in the form of a Web Form asking only participants’ email addresses. By submitting this form, a custom script incrementally generated a unique identifier for each new request and sent an email to the specified address with two links, one to access IST-Pre and another to access IST-Post. These generated links have encoded the identifier of the participant, so they were unique for each participant.

It is important to point out that participants’ email addresses were used only to send the links; thus, participants’ sensitive data (such as email addresses) were not accessible by the experimenters in any phase of the study.

4.1 Overall analysis

Regarding the Scientific–Technical sample, we excluded from further analyses the two NA participants who did not declare which school they were enrolled in. Thus, the final sample resulted in N = 78. First, we were interested in investigating whether both factors, namely participants’ education type and acquired knowledge, affected the likelihood of adopting the Intentional Stance toward robots. Thus, we performed a Repeated Measures ANOVA, with the IST mean scores in the two sessions (i.e., Pre vs. Post) as the repeated measures factor, and the education type (scientific/technical vs. artistic) as the between-subject factor (“Artistic” sample, N = 56; Mean IST score_Pre = 34.47, SD_Pre = 20.89; Mean IST score_Post = 41.67, SD_Post = 24.99. “Scientific–Technical” sample, N = 78; Mean IST score_Pre = 34.97, SD_Pre = 19.77; Mean IST score_Post = 41.79, SD_Post = 24.11). Only the main effect of IST emerged as significant [F _{(1, 132)} = 10.04, p = 0.02, η² = 0.02] (Figure 3).

Figure 3

Mean IST scores, plotted as a function of Session (IST Pre vs. Post), separately for Education Type (Artistic vs. Scientific–Technical students). Bars indicate the 95% confidence interval.

4.2 Analyses of the artistic vs. technical groups

To further explore the impact of technical background on the adoption of the Intentional Stance toward a humanoid robot in adult high school students, we decided to select only students with a technical background. In other words, we excluded students who attended a purely scientific high school without a technical component. We compared their mean scores to the students with an artistic background education with a Repeated Measures ANOVA (“Artistic” sample, N = 56; Mean IST score_Pre = 34.47, SD_Pre = 20.89; Mean IST score_Post = 41.67, SD_Post = 24.99. “Technical” sample, N = 37; Mean IST score_Pre = 39.77, SD_Pre = 17.8; Mean IST score_Post = 53.86, SD_Post = 19.32).

Results showed both the within-subject main effect of IST scores in the two sessions (Pre vs. Post) [F(1, 91) = 13.84, p < 0.001, η² = 0.05] and the between-subject main effect of the education type (technical vs. artistic) [F _{(1, 91)} = 6.2, p < 0.0001, η² = 0.04] (Figure 4). No interaction emerged as significant [F _{(1, 91)} = 1.4, p = 0.23, η² = 0.006].

Figure 4

Mean IST scores, plotted as a function of Session (Pre vs. Post), separately for each education type (artistic vs. technical students). Bars indicate the 95% confidence interval.

4.3 Analysis of social and non-social factors of IST

In a recent work, Spatola and colleagues [28] explored the factorial composition of a shorter version of the IST, identifying two factors. Factor 1 identifies scenarios where the robot is involved in social interaction with a human (“Social” factor), whereas Factor 2 identifies scenarios where the robot is alone, and thus, the social component is absent (“Non-Social” factor). That being said, we were interested in assessing whether there was a difference, in terms of likelihood of attributing intentionality to robots (i.e., IST scores), between the scenarios in which iCub was depicted while interacting with a human (“Social” scenarios) compared to the ones in which the robot was depicted alone (“Non-Social” scenarios), and whether it might depend on participants’ education type (i.e., Scientific–Technical vs. Artistic). Therefore, we decided to subselect only the IST items identified by Spatola and colleagues [28]. Then, we explored the two factors with a Repeated Measures ANOVA, where the IST scores for the Social vs. Non-Social items (Pre and Post) were considered as within-subject factors, and the education type (Scientific/Technical vs. Artistic group) was considered as a between-subject factor. Given that the assumption of sphericity was not met, results are reported with the Huynh-Feldt correction for sphericity. Results showed a significant main effect of IST Pre vs. Post scores [F _{(2.313, 91)} = 39.48, p < 0.0001, η² = 0.2]. The main effect of education type was not significant [F _{(1, 91)} = 3.68, p = 0.06, η² = 0.01], as well as no significant interaction between education type and IST Factors (Social and Non-Social) emerged [F _{(1, 91)} = 2.24, p = 0.10, η² = 0.01] (Figure 5).

Figure 5

Mean IST scores for Social and Non-Social items, plotted as a function of Session (Pre vs. Post), separately for each education type (artistic vs. scientific–technical students). Bars indicate the 95% confidence interval.