Children’s Beliefs About the Agentic Properties of Three Different Interactive Technologies

• In Event: Poster Session 11
  In Poster Session: PS 11 Topic - Social Cognition

Abstract

Research has demonstrated that we attribute agency to plants (ojalehto, Medin, & Garcia, 2017), animals (Weisman, Dweck, & Markman, 2017), and even inanimate objects (Heider & Simmel, 1944), and this begins early in life (Johnson, 1998). Recently, children are growing alongside another type of agent in the form of interactive technology, and readily engage with such agents as if they are animate beings (Kahn et al., 2011; Chernyak & Gary, 2016). In this study we aimed to assess 4-11-year old children’s (N = 128, Mage = 7.51, SDage = 2.27) beliefs about the agentic qualities of three interactive technologies (i.e. robots). We chose robots based on features that have been shown to influence agency attribution: Roomba can move autonomously but isn’t designed to be interactive. Amazon Alexa doesn’t have a body but can interact and communicate. Nao is humanoid thus can move autonomously and can be interactive and communicative. For each robot, children were asked to rate similarity to a computer versus a person (0-5 rating scale from “a lot like a computer” to “a lot like a person”). We also asked children to rate on a 4-item scale (0 = “Not at all” to 3 = “A lot”) whether they thought the robot had a series of emotions, thoughts, and biological functions. We also asked about the robot’s moral knowledge and moral intentions/choices. Lastly, we asked whether the robot deserves moral treatment.

Children said Roomba was most computer-like (M = 0.94/5, SD = 1.28) and didn’t attribute biological, emotions, thoughts, and moral knowledge to Roomba. Alexa was rated less computer-like than Roomba (M = 1.06/5, SD = 1.36) though not significantly so. Ratings for Alexa’s emotions and thoughts were higher than Roomba (ps < .009) except for feeling scared or neglect. Children rated Nao as significantly most person-like than the other two (M = 2.06/5, SD = 1.70; pairwise t tests, ps < .001). Emotion, thought, and biological functions were significantly higher for Nao than either Roomba or Alexa (ps < .009) except for thinking, where Nao and Alexa were equal, and getting hurt, where all robots were equal (see Fig.1).

Despite children’s beliefs that Nao had more emotional and biological qualities than Alexa, their beliefs about moral issues were more divided. Children equally rated Nao and Alexa as capable of knowing “good” from “bad” and also rated Alexa’s moral intentions high and equal to Nao. For Nao, children linked moral knowledge to person similarity, emotions and thoughts (except neglect), biological functions, and moral intent (ps < .05). For Alexa, however, only emotions and thoughts, feeling pain, and moral intent linked to moral knowledge (ps < .05; see Table 1). These findings highlight the centrality of interactive abilities in children’s attribution of human-like agentic qualities to robots. Children’s beliefs about the disembodied yet communicative Alexa underscore this point – children viewed Alexa as a thinking, moral agent, regardless of it’s lack of biological and personhood qualities. These findings have implications for children’s social cognitive development in an increasingly technological world.

Authors

• Teresa Flanagan, Cornell University
  Presenting Author

• Gavin Wong, Cornell University
  Non-Presenting Author

• Tamar Kushnir, Duke Univeristy
  Non-Presenting Author