A meta-analysis on the effectiveness of anthropomorphism in human-robot interaction

Abstract

The application of anthropomorphic design features is widely assumed to facilitate humanrobot interaction (HRI). However, a considerable number of study results point in the opposite direction. There is currently no comprehensive common ground on the circumstances under which anthropomorphism promotes interaction with robots. This metaanalysis aims to close this gap. A total of 4,856 abstracts were scanned. After an extensive evaluation, 78 studies involving around 6,000 participants and 187 effect sizes were included in this meta-analysis. The majority of the studies addressed effects on perceptual aspects of robots. In addition, effects on attitudinal, affective, and behavioral aspects were also investigated. Overall, a medium positive effect size was found, indicating a beneficial effect of anthropomorphic design features on human-related outcomes. However, closer scrutiny of the lowest variable level revealed no positive effect for perceived safety, empathy, and task performance. Moreover, the analysis suggests that positive effects of anthropomorphism depend heavily on various moderators. For example, anthropomorphism was in contrast to other fields of application, constantly facilitating social HRI. In conclusion, the results of this analysis provide insights into how design features can be used to improve the quality of HRI.Moreover, they reveal areas in which more research is needed before any clear conclusions about the effects of anthropomorphic robot design can be drawn.

Crisis Ahead? Why Human-Robot Interaction User Studies May Have Replicability Problems and Directions for Improvement

“…2021) had sample sizes lower than 20. At the same time, the meta-analyses reported here show that effect sizes in HRI user studies seem to be rather small to moderate in most cases (Leichtmann and Nitsch, 2020a;Esterwood et al, 2021;Hancock et al, 2021;Roesler et al, 2021;Stower et al, 2021). Another hint for expecting rather low effect sizes is that a large portion of HRI user studies examined factors including anthropomorphism, attitudes, personality, gender, and others-topics that have been studied in context of the social sciences and are known to have rather small effect sizes (Richard et al, 2003).…”

“…Taxonomies have been excluded from this analysis. In total, eight meta-analyses had been used to identify patterns of concern in the literature: A meta-analysis by 1) Leichtmann and Nitsch (2020a) on personal space in HRI with a total of k = 27 studies including N = 1,299 participants, by 2) Stower et al (2021) on trust in child-robot interaction including k = 20 studies with N = 977, by 3) Ötting et al (2020) on the effects of design features on HRI at work using k = 81 studies with N = 2,245, by 4) Esterwood et al (2021) on personality effects in HRI with k = 26 studies and N = 1,611, by 5) Roesler et al (2021) on anthropomorphism in HRI with k = 78 studies and N = 5,973, by 6) Hancock et al (2021) on trust in HRI using k = 45 studies, by 7) Yuan et al (2021) on the effect of robot-assisted cognitive training including k = 53 studies with approximately N = 1,166 participants, and by 8) Mou et al (2020) with k = 40 studies on “robot personality”. We additionally analyzed a recent meta-analysis not yet on Scopus, but in press.…”

“…One of the most obvious problems concern issues of scientific reporting, that is original studies are not always transparent enough, fail to describe all important information, and violate the norm of completeness. Most meta-analyses mentioned that statistical details had been missing such as effect sizes, confidence intervals, or descriptive statistics (i.e., means and variances) ( Leichtmann and Nitsch, 2020a ; Ötting et al, 2020 ; Mara et al, 2021 ; Roesler et al, 2021 ; Stower et al, 2021 ), or were inconsistent ( Hancock et al, 2021 ) to such an extent that some studies had to be excluded for meta-analytical calculations. Lack of information is not only crucial to integrate knowledge in meta-analyses.…”

Variable Impedance Control Based on Target Position and Tracking Error for Rehabilitation Robots During a Reaching Task

“…And with respect to vIC-e, better performance of vIC was obtained in all respects. The results in this article also might demonstrate the assumption that a control strategy with humanlike characteristics could ensure accurate and smooth tracking performance, as well as safe and natural HRI (Roesler et al, 2021).…”

Designing Man’s New Best Friend: Enhancing Human-Robot Dog Interaction through Dog-Like Framing and Appearance

“…While most work defining human-likeness [ 51 , 52 ] or animal-likeness [ 50 ] has focused on the physical appearance and form of a robot, what comprises the overall assessment of a robot may include additional elements such as unique characteristics, personality traits, and behaviors. Recent work has begun to chart the relative contributions of these dimensions to outcome measures relevant to human-robot interaction [ 53 , 54 ]. Depending on the dog-like feature modeled on the x-axis (a combination of form, characteristics, behavior), the outcome measure (y-axis) may show a different pattern.…”