XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE Measurement of Moral Concern for Robots Tatsuya Nomura 1,4 1 Ryukoku University Otsu, Shiga 520-2194, Japan 4 ATR Intelligent Robotics and Communication Laboratories Keihanna, Kyoto 619-0288, Japan nomura@rins.ryukoku.ac.jp Takayuki Kanda 2,4 2 Department of Social Informatics, Kyoto University Yoshida-honmachi, Kyoto 606-8501, Japan kanda@i.kyoto-u.ac.jp Sachie Yamada 3,4 4 Department of Psychological and Sociological Studies, Tokai University Hiratsuka, Kanagawa 259-1292, Japan s-yamada@tokai-u.jp Abstract— We developed a self-report measurement, Moral Concern for Robots Scale (MCRS), which measures whether people believe that a robot has moral standing, deserves moral care, and merits protection. The results of an online survey (N = 200) confirmed the concurrent validity and predictive validity of the scale in the sense that the scale scores are successfully used to predict people’s intentions for prosocial behaviors. Keywords— moral concern, self-report scale I. INTRODUCTION Morality is one intrinsic human characteristic. People have an innate motivation to help others even if such action/decisions decrease their own benefit. Although such moral cognition is usually applied to human beings, people sometimes expand it to include such non-human entities as animals and nature, e.g., extending basic human right to the great apes [1]. Individual differences exist in moral expansiveness. A less morally expansive person restricts her moral concern to those entities she deems “close” (e.g., family), and a more morally expansive person extends her moral concern beyond more “distant” entities like animals. However, opposite situations also occur. Sometimes people avoid expanding their moral concern to include pets and robots, and mistreat them (e.g., [2]). Imagine a future scenario where robots serve various roles in our daily lives. Robot abuse might be a serious societal problem. In a store, robot clerks might be abused and fail to maintain the stores; robot workers might be cheated by their human co-workers and fail to receive appropriate work efforts from their employees; when a robot asks a human for help, it might receive scorn or abuse. For such future scenarios, we expect people to offer a minimal level of prosocial behaviors, not necessarily a great level of morality, instead of harm. We expect diverse moral relationships between individuals and robots, depending on such factors as personality, robot appearance and behaviors, and interaction contexts. In some contexts, we want to elicit more moral concerns to improve a robot’s treatment. In other contexts, we might want to decrease our moral concern so that users can easily manipulate robots as tools without being bothered by their well-being. Here the fundamental research question is how to measure moral concern for robots. Our research establishes a self-report measurement for this concept, i.e., moral concern for robots. HRI empirical studies commonly use scales (self-report questionnaires). This paper reports the development of a scale for the moral concern for a robot called Moral Concern for Robots Scale (MCRS). II. SCALE DEVELOPMENT To collect item pool for MCRS, we adopted nine items from the interview protocol in Kahn et al. [3] which asks about moral concern for the disposal/destruction and forced labor of robots, two items from the Feelings toward Nature Scale [4] which asks whether people feel negative emotions if nature is destroyed, and five items from the Thoughtfulness toward Friends Scale [5] which asks about prosocial behaviors toward friends. Moreover, we created four items that mention humans’ moral treatment and account for robots based on the language in the instructions and definitions of the Moral Expansiveness Scale [6], and eight items based on scenes of possible robot abuse. Finally, we prepared 28 candidate items for our prototype MCRS version. Then, we conducted a questionnaire-based survey with 121 Japanese university students (males: 66; females: 55; mean age: 20.1 (SD = 1.6)). In the survey, to provide a context for the answer targets, we first presented a scene where a robot worked in a city. Then we administrated a questionnaire, i.e., a prototype version of MCRS that consists of the above 28 questionnaire items. Each item was evaluated by a 7-point Likert scale (1: strongly disagree, to 7: strongly agree). We analyzed the collected data by conducting an exploratory factor analysis using principal component analysis and Promax rotation. A two-factor structure was decided based on a scree-plot and item consistency. Two subscales (factors), consisting of 21 items (first factor: 12 items; second factor: 9 items), were extracted based on factor loadings, the contents of the items, and the item analysis results in each subscale, which consisted of I-T correlation coefficients and α-coefficients. The cumulative contribution ratio of these two factors on the data was 47.3%, which is enough coverage. The Cronbach’s α- coefficients for each subscale were .912 and .876, which indicate good internal consistency. The first subscale, which is called the basic moral concern, consists of items that ask whether people have general moral concerns for robots (e.g., when they should be destroyed or suffer physical harm) and whether people spend their resources to provide better welfare for them (e.g., helping them and The research was supported by JST CREST, Japan.