Legend has it that Saint Francis of Assisi thought that, like humans, all non-human animals, had minds and he communicated with them (Zanker, 2007). It is not only saints who perceive minds in non-human animals but the rest of us often do so on a day-to-day basis in different animals, such as monkeys, dogs, cats, and so on. Mind perception is not limited to living creatures. A wide variety of artificial entities (e.g., interactive robots) and natural phenomenon (e.g., the north wind and the sun in an Aesop’s fable) are sometimes treated as having a mind. This does not, however, mean that people of different ages and cultures share the same type of concept of the mind and the same attitudes toward the mind. For example, some Japanese people, following traditional Japanese conventions, believe that material objects (e.g., dolls and scissors) that have been used for a long time, develop minds and these people often hold a memorial ceremony for these objects when they dispose of them. By contrast, this idea that “inanimate objects have minds” is hard to make sense of in traditional Christian culture because a mind is considered a special gift from God, and artificial entities are denied a mind. Hence, it is important to investigate how we perceive minds in living and non-living entities for the purpose of understanding the diversity of human cultures (Kraft, 1995).

Questionnaire assessments are mainly used in the study of mind perception. One of the landmark studies of mind perception, conducted by Gray et al. (2007), used a large-scale web survey to analyze mind perception styles in over 2000 respondents. In this study, subjects were asked to rate the degree to which each of the 18 mental capacities (e.g., capacity to feel hunger) was suitable for explaining each of the 13 target characters (e.g., adult males, infants, dogs, and gods) on a 5-point Likert scale. The result of the study was that two orthogonal mind dimensions, named “experience” and “agency,” respectively, were found by principle component analyses (PCA). The dimension of “experience” indicates the capacity to sense and feel emotions, whereas the dimension of “agency” indicates the capacity to plan and execute intentional actions. For example, according to Gray et al.’s survey, we perceive strong “experience” but not “agency” in babies and other animals; on the contrary, we perceive “agency” but not “experience” in robots and gods. The multi-dimensional views of mind perception have been confirmed by other studies and several derivative findings, leading to the agreement that this is a good framework for explaining psychological phenomena, such as the “uncanny valley” and cognitive distortions of psychiatric disorders (Gray et al., 2011; Gray and Wegner, 2012).

Although the two mind dimensions proposed by Gray and colleagues are insightful, it is not easy to apply the questionnaire they used to measure mind perception toward non-living entities, e.g., robots. In their questionnaire, subjects were instructed to rate the degree to which a mental (biological) capacity was matched with a character. For example, subjects were asked “To what degree does a robot have a capacity to feel hunger?” The questionnaire, however, does not differentiate between what impression we have of and what we know about an entity. The subjects’ rating may be heavily influenced by their prior knowledge about robots, e.g., knowledge that robots are not capable of feeling hunger. To differentiate between what we mean by “impression” and “knowledge” of an entity, imagine the case in which a person meets a human-like android and has a strong first impression that the android feels hunger. In this case, one may perceive a mind in the android. At the same time, however, one knows that the android is a machine and does not have the capacity to feel hunger. This knowledge may prevent a person from forming a spontaneous mind perception. Most items included in Gray and colleagues’ questionnaire are concerned with mental capacities that only biological entities have. Subjects arguably have knowledge that these capacities are not implemented in non-living entities, and are forced to answer the questionnaire in a biased way. Sometimes we inevitably form an impression of an object and we assume the behavior comes from a mind. Hence, we need to exclude the effects of prior knowledge to investigate mind perception toward non-living entities, insofar as mind perception is associated with impression. The questionnaire used by Gray et al. is designed to measure one’s general conceptions of mental capacities of entities. A different questionnaire may be appropriate for assessing our mind perception.

We suggest that a semantic differential scale method could reveal multidimensional aspects of mind perception (Bradley and Lang, 1994). This scale does not include questions concerning mental capacities and is capable of assessing subject’s non-verbal impressions of various objects, events, and concepts on the basis of how they rate the matching between multiple adjectives and entities. In this method, an adjective is paired with its antonym and the two adjectives are assigned numbers on a scale. For example, “cold” is paired with “warm,” and they are given 1 and 7, respectively. Subjects were asked to evaluate where an entity is placed on the scale. In a previous study, we found that mind perception varies along two mind dimensions by using a questionnaire that included 21 paired-adjectives (Takahashi et al., 2014). The questionnaire we used was limited in generality and target; it mainly focused on brain activities in mind perception. The sample size in this previous study was small (n = 20) and the mean age of sample (university students) was a little biased. More importantly, we did not show that the two dimensions we found (i.e., mind-holderness and mind-readerness) corresponded to the two dimensions of “experience” and “agency” introduced by Gray and colleagues. By showing that the former dimensions correspond to the latter, we propose that our semantic differential scale is an effective way to detect the dimensions of “agency” and “experience” of mind perception.

In the current study, we performed a large-scale web survey in subjects of varying ages to generalize the two dimensions of our questionnaire. In this survey, subjects evaluated seven target characters by both the questionnaires of Gray et al. (2007) and of Takahashi et al. (2014). The results of the study suggest that the two dimensions found in our questionnaire correspond to the two mind dimensions of “experience” and “agency.” Our semantic differential scale can therefore be regarded as an effective way to intuitively assess multidimensional aspects of mind perception without asking questions about evident mental capacities.

We performed a PCA for rating scores of the two questionnaires to identify dimensions of each. PCA is a statistical procedure for an orthogonal transformation to convert a set of original multidimensional data into small number of orthogonal factors called principal components (details are seen in Holand, 2008). This method is suitable to extract specific factors related to mind perception from multiple questions in questionnaires. We found two factors with eigenvalues over 1.0 in the questionnaire by Gray et al. (2007), a factor corresponding to “Experience” (eigenvalue = 11.4) accounted for 63.4% and a second factor, “Agency” (eigenvalue = 2.2), accounted for 12.3% of the variance (detail loads of questions in each component are reported in Table 1). Furthermore, three factors with eigenvalues over 1.0 were found in the questionnaire used by Takahashi et al. (2014) a factor named “Emotion” (eigenvalue = 11.6) accounted for 55.1% of the variance, a second factor named “intelligence” (eigenvalue = 3.0), accounted for 14.3% of the variance and a third factor (eigenvalue = 1.0), accounted for 4.9% of the variance (detail loads of questions in each component are reported in Table 2). Results of semantic differential scale methods are often compressed into three components by using factor analysis. However, there were no significant correlations between this third component and the other two components in Gray et al. (2007). Hence, we do not discuss the third component in this paper.

We plotted mean scores of the PCA of the two questionnaires (Figure 1). We found that the locations of all characters in the two dimensions were similar between the two questionnaires. We calculated the correlation coefficients between “experience” and “emotion” and between “agency” and “intelligence,” respectively. Both values are positive and stochastic (“experience” and “emotion” r = 0.80, p < 0.00001, “experience” and “emotion” r = 0.75, p < 0.00001) and we concluded that the two dimensions revealed by the semantic differential scale method were similar to the two dimensions revealed from questions about mental capacities.

Furthermore, we investigated whether subject’s ages were correlated with these PCA components in each target and we found there were no significant correlations between ages and any dimensions of mind perception.

In this study, we directly compared two different types of questionnaires used for assessing mind perception. One asked about mental capacities and the other asked about impressions of targets by using the semantic differential scale method. From our results, we suggest that the two mind dimensions “emotion” and “intelligence” revealed in our questionnaire correspond to the dimensions “experience” and “agency” in Gray et al. (2007) questionnaire. This means that we can assesses mind perception without asking questions about mental capacities but by using the semantic differential scale method.

We consider the semantic differential scale method useful, especially when it comes to assessing multidimensional aspects of mind perception for artificial agents such as robots that are difficult to attribute mental (biological) capacities to but that we sometimes feel have a mind. Various new types of robots are being developed that are expected to communicate with us as living partners (Imai et al., 2003). We also know that attributing minds to inanimate agents improves trustworthiness and empathy toward these agents (Riek et al., 2009; Harrison and Hall, 2010; Waytz et al., 2014). These social emotions are essential to creating a good rapport between human and artificial agents. Hence, when we develop a social agent, it is important to evaluate how people perceive the mind in the agent. As our questionnaire does not include questions about mental capacities, our questionnaire can be broadly applied to the assessment of mind perception for various agents in human-robot or human-agent interaction studies.

Our study can be considered a Japanese retest of Gray et al. (2007) mind perception survey. Although the results obtained by Gray et al. (2007) and our results are relatively similar, results in artificial, inanimate objects were different between these two surveys. In our study, scores of “agency” (intelligence) in robots and super computers are higher than those in adult humans. Contrastingly, the scores of agency in robots were underestimated in Gray et al. (2007, 2011). We hypothesize that Japanese people might attribute stronger agency (intelligence) to artificial entities when compared to westerners. In some cultures, including Japan, people tend to believe non-living things have a mind, even if these things cannot be attributed to evident biological capacities. Furthermore, regardless of culture, young children tend to attribute mental states to non-living entities. For example, many children treat stuffed animals as their friends (Moriguchi et al., 2015). Moreover, some children have invisible friends called “imaginary companions” and communicate with them much like with actual human friends (Moriguchi and Shinohara, 2012). These beliefs about minds in non-living entities are often called “animism” (Harvey, 2005). Animism is the cultural attitude toward nature and external objects. Hence, this concept is strongly linked to views of life and religions in various cultures and the assessment of mind perception is quite important to understand cultural difference of these views. However, the sense of animacy is intuitive feeling and this sense cannot be explained logically. Therefore, the questionnaire that directly asks about mental capacities might not be appropriate for the assessment of mind perception in animism culture. We believe that the assessment of mind perception by using the semantic differential scale method is an intuitive way to assess subject’s impressions and this method is suitable to assess mind perception universally regardless of cultural differences. Further, because the abilities for processing other’s mind are often distorted in various psychiatric disorders (e.g., schizophrenia), our intuitive method might be useful for assessing these patient’s symptoms (Matsumoto et al., 2015).

HT, MB, and MA designed the research. HT and MB performed the research. HT 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.