Measuring the Perceived Social Intelligence of Robots

Abstract

Robotic social intelligence is increasingly important. However, measures of human social intelligence omit basic skills, and robot-specific scales do not focus on social intelligence. We combined human robot interaction concepts of beliefs, desires, and intentions with psychology concepts of behaviors, cognitions, and emotions to create 20 Perceived Social Intelligence (PSI) Scales to comprehensively measure perceptions of robots with a wide range of embodiments and behaviors. Participants rated humanoid and non-humanoid robots interacting with people in five videos. Each scale had one factor and high internal consistency, indicating each measures a coherent construct. Scales capturing perceived social information processing skills (appearing to recognize, adapt to, and predict behaviors, cognitions, and emotions) and scales capturing perceived skills for identifying people (appearing to identify humans, individuals, and groups) correlated strongly with social competence and constituted the Mind and Behavior factors. Social presentation scales (appearing friendly, caring, helpful, trustworthy, and not rude, conceited, or hostile) relate more to Social Response to Robots Scales and Godspeed Indices, form a separate factor, and predict positive feelings about robots and wanting social interaction with them. For a comprehensive measure, researchers can use all PSI 20 scales for free. Alternatively, they can select the most relevant scales for their projects. CCS Concepts: • General and reference → Empirical studies ; Measurement ; • Human-centered computing → User studies ; HCI theory, concepts and models;

Does what users say match what they do? Comparing self-reported attitudes and behaviours towards a social robot

“…f) Social Information Processing: Finally, Social Information Processing was measured using the 5-item social information processing sub-scale from the Tromsø Social Intelligence Scale [28], as well as the 'Identifies Humans' and 'Social Competence' sub-scales from the Perceived Social Intelligence (PSI) Scales Test Manual [6]. The combined reliability for the three subscales was α = 0.76.…”

“…A Perceived Social Intelligence Scale for Robots [6] has also been proposed, with two main dimensions; social competence and social presentation. However, this scale has so far only been validated for video evaluations of social robots, rather than real-life interactions.…”

Evaluation of Socially-Aware Robot Navigation

“…As a consequence, most prior works adopted an ad hoc approach, using custom questions to assess sociability (e.g., Vega et al, 2019a ). More recently, Perceived Social Intelligence (PSI) scales ( Barchard et al, 2020 ) offer an initial point for benchmarking the subjective construct of sociability. In order to productively advance socially-aware navigation, however, further research is required to develop comprehensive instruments specifically designed to measure sociability and higher-level social skills in the context of navigational interactions.…”

“…The Perceived Social Intelligence (PSI) scales proposed by Barchard et al (2018) ; Barchard et al (2020) evaluate 20 aspects of robotic social intelligence. For instance, the Social Competence (SOC) scale consists of four items: 1) social competence, 2) social awareness, 3) social insensitivity (reversed), and 4) strong social skills.…”

“…For instance, the Social Competence (SOC) scale consists of four items: 1) social competence, 2) social awareness, 3) social insensitivity (reversed), and 4) strong social skills. PSI scales have been used in previous evaluations of socially-aware navigation (e.g., Barchard et al, 2020 ); recently, Banisetty and Williams (2021) used the perceived safety scale from the Godspeed questionnaire in conjunction with PSI to evaluate how a robot’s spatial motions may communicate social norms during a pandemic via an online study. Additionally, it has been determined that robots using socially-aware navigation planners are perceived to be more socially intelligent as measured by PSI than those using traditional navigation planners ( Honour et al, 2021 ).…”