Factors Leading to the Successful Use of Voice Recognition Technology
Tanya Goette · 1998 · Proceedings of the Third International ACM Conference on Assistive Technologies (Assets '98) · doi:10.1145/274497.274532
Summary
This paper from Georgia College & State University presents results from a field study investigating why voice recognition technology (VRT) succeeds for some individuals with disabilities and fails for others. Drawing on expectancy theory and innovation-diffusion theory, the researcher conducted in-person interviews with 23 successful VRT users, 15 unsuccessful users (who abandoned VRT), and 5 prospective adopters who had not yet begun using VRT. Participants had disabilities including quadriplegia (30 of the subjects), intervertebral disk disorders, multiple sclerosis, muscular dystrophy, cerebral palsy, arthritis, and tendinitis, with an average disability duration of 20.33 years. The study used both quantitative measures (survey questionnaire analysed with principal components analysis, t-tests, and discriminant analysis) and qualitative methods (standardised open-ended interviews with cross-case analysis). Two types of VRT systems were in use: older limited-vocabulary systems requiring extensive voice training (30% of participants, from Votan, Command Corporation, etc.) and newer large-vocabulary systems designed for word processing (70%, from Dragon Systems or IBM). Success was measured through three dimensions: user satisfaction, level of use (hours per week), and achievements through use (percentage of possible tasks accomplished via VRT).
Key findings
Discriminant analysis revealed that the single most important factor distinguishing successful from unsuccessful VRT users was perceived benefits — a combination of relative advantage, compatibility, and ease of use (p=.000). This was followed by result demonstrability (knowing what VRT could accomplish before adopting) and using VRT for all computer access. Counter to the researcher's hypotheses, successful users had *higher* expected benefits (p=.020) and *higher* expected success (p=.007) than unsuccessful users — suggesting psychological factors and self-efficacy outweigh other variables. Successful users also expected higher complexity (p=.029), indicating realistic expectations about the learning curve. The majority of unsuccessful users (53%) abandoned VRT within three months. Qualitative cross-case analysis identified four patterns of failure: task-technology fit problems (53% of unsuccessful users — using the wrong type of VRT system for their tasks), inadequate training (40% — including insufficient hours, compressed training schedules, and workplace pressure to perform immediately), environmental problems (20% — background noise, privacy concerns), and disability limitations (33% — inconsistent speech patterns, particularly with cerebral palsy, and VRT being slower than alternative input methods). A critical finding was that almost none of the participants had tried VRT before purchasing it, and unsuccessful users consistently said a trial period would have revealed incompatibilities.
Relevance
This study provides one of the earliest empirical investigations of assistive technology adoption and abandonment, grounding the analysis in established information systems theory rather than anecdotal evidence. Its findings remain strikingly relevant today. The one-third abandonment rate for assistive devices (cited from a national survey) is still widely referenced in AT literature, and the factors identified — task-technology fit, adequate training, environmental suitability, and trial periods — continue to be the primary determinants of AT success. For practitioners recommending voice input or any assistive technology, the paper's guidelines are directly actionable: manage expectations realistically, match the specific technology to the specific tasks, provide training scaled to prior experience, and critically, allow extended trial periods in the actual use environment before committing to purchase. The finding about workplace accommodation is also important — employees need reduced workloads or extended deadlines during the VRT learning curve, or they will revert to previous methods. The paper anticipates modern AT assessment best practices that emphasise person-centred, contextual evaluation over one-time clinical assessments.
Tags: speech recognition · assistive technology · technology abandonment · motor impairment · user studies · assistive technology adoption · empirical studies