JMIR Publications

ABSTRACT

Background:

Socially Assistive Robotics (SAR) is an emerging interdisciplinary field that integrates robotics, medicine, cognitive and social sciences, and neuroscience to support humans across diverse contexts. Despite recent advances, the potential of SAR for people with Multiple Sclerosis (MS) remains largely unexplored, even though MS affects sensory–motor, cognitive, and affective domains where SAR interventions could offer substantial benefits.

Objective:

This study aimed to evaluate the perceived usability of a humanoid robot in people with MS (PwMS), compare their experiences with age-matched healthy subjects (HS), exploratorily identify potential predictors of usability, and investigate whether language could provide insights into participants’ appreciation of the robot.

Methods:

PwMS and HS completed four predefined tasks with Pepper, a social robot developed by SoftBank Robotics. Usability was assessed using the System Usability Scale (SUS). Attitudes toward robots, robot-related factors, and task-related effort were measured with the General Attitudes Toward Robots Scale (GAToRS), Godspeed questionnaire, and NASA-TLX, respectively. Psychological variables (e.g., personality traits from I-TIPI) and clinical characteristics for PwMS (e.g., cognitive functions via the app DIGICOG-MS, upper-limb difficulties with the AMSQ) were collected. Participants’ language during semi-structured interviews was analyzed using Natural Language Processing (NLP) to explore associations between SUS and clinical measures.

Results:

SUS scores were above the standard cut-off, with no significant differences between PwMS (M 81.0; SE 2.75) and HS (M 77.8; SE 2.51; P = .389), indicating high appreciation of the robot in both groups. PwMS reported greater negative attitudes towards robot introduction at a societal-level (GAToRS S-: P = .023) and perceived Pepper as more anthropomorphic and intelligent than HS (Godspeed ANTR: P = .007; INT: P = .023). Spearman correlations showed that ANTR was strongly associated with visuospatial (ρ = -0.632, P = .003) and relatively strong with both verbal (ρ = -0.504, P = .023) memory from DIGICOG-MS and with Agreeableness from I-TIPI (ρ = 0.491, P = .028). A first exploratory stepwise multiple linear regression showed that conscientiousness (I-TIPI), negative personal attitudes toward robots at a personal level (GAToRS P -), and perceived upper-limb difficulty (AMSQ) explained 64% of the variance in SUS scores (Adjusted R² = 0.568, P = .001). A second regression model indicated that the frustration dimension of NASA-TLX explained 48% of SUS variance (Adjusted R² = 0.451, P = .001). No significant between-group differences were observed in extracted linguistic features, including lexical productivity and sentiment-lexicon coverage, indicating a positive emotional tone and high acceptability in both PwMS and HS.

Conclusions:

This study provides the first empirical evidence on perceived usability of a social robot in PwMS, suggesting that psychological aspects, prior attitudes toward robots, and clinical limitations significantly shape user experience. Integrating clinical, psychological, and language-based measures offers valuable guidance for the development of tailored SAR interventions in rehabilitation scenario in MS.