Bridging accessibility gaps in self-service environments with mobile conversational Al: A usability study

PurposeThe rapid proliferation of self-service kiosks in sectors such as foodservice, transportation, and healthcare has improved efficiency but has also exacerbated accessibility gaps for older adults and people with disabilities.[1] In South Korea, where kiosks have tripled in number within just two years [2], inconsistent interface designs and limited accessibility features have led to frustration, anxiety, and task abandonment, particularly among older adults and users with visual impairments.Existing interventions, such as adjustable layouts or physical modifications of kiosks, often rely on retrofitting and lack long-term scalability.Given the limitations of kiosk-centered solutions, this study aims to evaluate whether a conversational AI agent embedded in users' personal mobile devices could offer a more consistent, intuitive, and scalable alternative for accessible self-service ordering.Method The interdisciplinary research team at KAIST has developed an Android-based conversational AI agent called "Chaltteok" (meaning rice cake in Korean), which supports voice, text, and Braille interaction.The system integrates speech-to-text, text-to-speech, LLM-based dialogue management, and real-time APIs for menu navigation and payment processing.A total of 13 participants were recruited for this usability study using purposive sampling (4 older adults, 7 blind persons, and 2 persons with low vision).Participants completed six ordering tasks in a simulated café scenario.Usability was assessed using the System Usability Scale (SUS), perceived task difficulty using the Individual Prioritised Problem Assessment (IPPA), and task performance through completion status and time.Results and Discussion Chaltteok achieved a high SUS mean score of 84.8, rated between "Good" and "Excellent," indicating strong usability across participant groups.Older adults