Perception and Evaluation in Human–Robot Interaction: The Human–Robot Interaction Evaluation Scale (HRIES)—A Multicomponent Approach of Anthropomorphism

Abstract

The evaluation of how (human) individuals perceive robots is a central issue to better understand Human-Robot Interaction (HRI). On this topic, promising proposals have emerged. However, present tools are not able to assess a sufficient part of the composite psychological dimensions involved in the evaluation of human-robot interaction. Indeed, the percentage of variance explained is often under the recommended threshold for a construct to be valid. In this article, we consolidate the lessons learned from three different studies and propose a further developed questionnaire based on a multicomponent approach of anthropomorphism by adding traits from psychosocial theory about the perception of others and the attribution and deprivation of human characteristics: the de-humanization theory. Among these characteristics, the attribution of agency is of main interest in the field of social robotics as it has been argued that robots could be considered as intentional agents. Factor analyses reveal a four sub-dimensions scale including Sociability, Agency, Animacy, and the Disturbance. We discuss the implication(s) of these dimensions on future perception of and attitudes towards robots.

The Intentional Stance Test-2: how to measure the tendency to adopt intentional stance towards robots.

“…. We conducted two studiesto develop a task that does not rely on first-order questions of evaluation (which is traditionally used in questionnaires) (Bartneck et al, 2009;Carpinella et al, 2017;Spatola et al, 2020b) but on a second-order inferences. The aim was to reduce standard bias relative to questionnaires (Choi & Pak, 2005;Furnham, 1986;Kalton & Schuman, 1982).…”

“…The purpose is to provide a test of the consistency of the scale to measure the theoretical dimension of intentionality. This new approach in task validation allows us to evaluate whether the measure provided by a task is internally consistent (Spatola et al, 2020b).…”

“…For each factor we proceeded as follows: all items were included in a scale reliability analysis to evaluate the reliability of the factor if an item is dropped to maximize the Cronbach's alpha of the factor (Cronbach, 1951;Unwin, 2013). We considered factor with an alpha less than 0.7 as nonreliable and dropped them to ensure the psychometric structure stability (Laher, 2010;Spatola et al, 2020a). From the 32 experimental items, 13 remained in the final matrix, χ 2 (78) = 1288.74, p<.001; KMO = .91, explaining 47.43% of variance (Figure 2) with 2 factors (Table 1).…”

“…To evaluate the structural stability of the present questionnaire we wanted to compare the consistency of the prediction of the scale with a training and test sample. The purpose was to evaluate whether the scale was reliable to reproduce its prediction (here in terms of cluster solution as a respondent profiles proxy) on different samples (for similar procedure see (Spatola et al, 2020a)).…”

“…The paper is structured as follows: in the first study, we built on Marchesi et al's material (Serena Marchesi et al, 2019) and psychometric validation methodology, to test the psychometric structure of our task. We also evaluated (1) the external validity of the task in relation to anthropomorphic attributions (measured by the Human-Robot Interaction Evaluation Scale (Spatola et al, 2020a) and (2) the influence of the predictive personality traits (Epley et al, 2007) that are the need for cognition (Cacioppo et al, 1984) and the need for closure (Roets & Van Hiel, 2011). In a second study, we produced a confirmatory factor analysis and tested the correlation between the task we aimed to develop and the attitudes towards robots (measured by the Negative Attitude toward Robots scale (Nomura et al, 2006a)); in addition, we tested whether the present task measure could predict anthropomorphic attributions to other robots.…”