[Paper Review] Technology Acceptance Among Digitally Vulnerable Groups — AI Acceptance Tendencies Among the Hearing-Impaired

As artificial intelligence rapidly spreads across all aspects of society, an important question is being raised: Are all social groups equally benefiting from this technological innovation? This paper examines AI acceptance tendencies among hearing-impaired individuals — a representative digitally vulnerable group — and analyzes the factors that either facilitate or hinder technology acceptance.

The research adopts an extended Technology Acceptance Model (TAM) framework, adding variables such as social influence, facilitating conditions, and trust to the traditional factors of perceived usefulness and perceived ease of use. Survey data was collected from hearing-impaired individuals across various age groups and occupational backgrounds.

Key findings reveal several distinctive patterns. First, perceived usefulness was the strongest predictor of AI acceptance among the hearing-impaired. Participants who believed AI tools — such as real-time captioning, sign language recognition, and speech-to-text conversion — would genuinely improve their daily lives and professional capabilities showed significantly higher acceptance rates.

Second, trust emerged as a particularly critical factor for this population. Concerns about data privacy, accuracy of AI interpretations, and potential misrepresentation of their communications created substantial barriers to adoption. This finding suggests that AI developers must prioritize transparency and accuracy when designing tools for communication-dependent vulnerable groups.

Third, facilitating conditions — including access to training, technical support, and accessible interfaces — played a more significant role than in the general population. Many hearing-impaired users reported difficulty in independently navigating AI setup processes and troubleshooting errors, highlighting the need for more inclusive design and support structures.

The research also identified generational differences. Younger hearing-impaired individuals demonstrated higher acceptance rates and were more willing to experiment with new AI tools, while older participants showed more conservative adoption patterns, often citing reliability concerns and preference for established communication methods.

Social influence, interestingly, showed a bidirectional effect. Positive recommendations from trusted community members — particularly within the deaf and hard-of-hearing community — significantly boosted acceptance. However, negative experiences shared within the community also spread rapidly and created disproportionate barriers.

The implications for AI developers and policymakers are substantial. Technology designed for mainstream users often fails to adequately serve hearing-impaired individuals, not necessarily due to fundamental capability limitations, but because the needs, contexts, and trust factors of this population were not centered in the design process.

The paper concludes that achieving genuine digital inclusion requires moving beyond simply making AI tools "accessible" in a technical sense, toward co-designing with digitally vulnerable communities from the earliest stages of development. When hearing-impaired individuals are involved as active participants in shaping AI tools meant to serve them, both acceptance rates and actual utility are significantly improved.

This research contributes to the growing body of work on inclusive AI development and offers a framework applicable to other digitally vulnerable populations, including elderly users, those with visual impairments, and individuals with limited digital literacy.