Opportunity
AI systems can help extend access to support when and where trained professionals or human expertise are limited or unavailable. This includes interpretation, teaching, and guided interaction, where AI tools such as captioning systems, virtual tutors, and social robots can provide immediate, on-demand support. In low-resource settings, including low- and middle-income countries, these tools may help reduce service gaps where human assistance is scarce. These systems can also provide more consistent access outside of standard hours or geographic limits, which can create flexibility to support participation in education, services, and daily activities.
Limitation
While these tools may improve access, they can also risk becoming a substitute for investment in human expertise rather than a supplement to it. AI systems may not provide the same quality, cultural understanding, or responsiveness as human experience and expertise, and may be less effective for people whose needs fall outside common patterns. Overreliance on AI for access can create unequal conditions, where some people faced with barriers are expected to use automated support while others receive individualized human assistance.
Illustrative Examples
Sign-language and captioning AI to address interpreter and captioner shortages
AI sign-language and captioning tools (i.e., sign-language recognition applications, avatar interpretation software, translation applications, automated captioning) can extend access where qualified interpreters and captioners are scarce, addressing shortages particularly in low- and middle-income countries.
Caution
AI tools for sign languages are not widely trusted or accepted in Deaf communities. These tools often simplify sign languages and do not fully capture meaning, context, culture, or the nuances and richness of sign languages. Many systems are built with little or no involvement from Deaf people.
There is no single, universal sign language. There are more than 300 distinct sign languages worldwide. Tools trained in one sign language usually do not work well for others, and many sign languages, especially those used by smaller communities, are missing or poorly represented in training data.
The use of sign‑language AI tools should be guided by Deaf and Hard of Hearing communities. In high‑stakes settings such as healthcare, legal, or education contexts, these tools should clearly state their limits and must always be paired with access to a qualified human interpreter.
AI tutoring and training for persons with disabilities in low-resource settings
In low-income and under-resourced regions, AI-based tutors are sometimes proposed as a way to provide individualized learning support for persons with disabilities. These tools may be used in places where there are few trained teachers, disability specialists, or accessible educational services, and where one‑to‑one human support is limited or unavailable.
Caution
The quality, accuracy, appropriateness, and safety of instruction should be assessed. Relying on AI in these contexts may reinforce unequal systems by placing persons with disabilities in separate or lower-quality learning environments, while others continue to receive human-led, tailored support. Rather than treating AI as a substitute, efforts could prioritize policy reform, sustained public funding, and the development of inclusive, accessible education systems. This approach helps ensure that persons with disabilities are not required to rely on AI as a workaround. In addition, AI systems are often less effective for learners whose needs, communication styles, or learning patterns differ from those most commonly represented in training data.
Social robots and embodied tutoring
Physical AI robots can provide impactful benefits beyond the screen for individuals whose learning needs involve physical engagement, interaction, or real-time feedback. For some learners, including those who benefit from structured routines or sensory input, embodied interaction may support attention, motivation, and understanding. These tools can also provide consistent one-on-one support in settings where human resources are limited.
Caution
Cost, maintenance, and technical support requirements may put these systems out of reach in many low- and middle-income countries, limiting who can benefit from them. There is also a risk that robots are introduced without enough input from the communities they are meant to serve, which can lead to low adoption or abandonment. Robots should be used to complement, not replace, human caregivers and educators, who provide context, judgment, and emotional understanding that current AI systems may not be able to replicate.