Adaptation of a Cash Dispenser to the Needs of Blind and Visually Impaired People
Jens M. Manzke · 1998 · Proceedings of the Third International ACM Conference on Assistive Technologies (Assets '98) · doi:10.1145/274497.274518
Summary
This paper from ETH Zurich describes the software adaptation of an Olivetti cash dispenser (ATM) to make it accessible to blind and visually impaired people, conducted as part of Switzerland's Walk-up-and-use Technology initiative. The key constraint was that no hardware modifications were permitted — all accessibility improvements had to be achieved through software changes alone. The modifications included adding speech output using pre-recorded natural speech, redesigning screen graphics for better legibility, and implementing three alternative function menu designs. The adapted machine featured Braille labels beside command keys, two loudspeakers on top, and a "Skip" command key for navigating speech sequences. Three menu designs were tested: Solution A followed conventional ATM layout with functions linked to screen-adjacent keys; Solution B stressed the most common function (withdrawal without receipt) and provided larger fonts; Solution C offered a linear binary navigation structure where users could step through options sequentially. The study employed two complementary evaluation methods: a usability test with 10 blind and 5 visually impaired participants performing four banking tasks, and a heuristic evaluation by 5 HCI experts. The research is notable for its systematic comparison of these two evaluation methodologies in an accessibility context.
Key findings
The usability test and heuristic evaluation produced notably inconsistent results. For speech output, blind subjects identified 5 design failures while experts found 11, agreeing on only 6. For graphics, impaired subjects found 6 issues versus experts' 19, with no agreement on any single issue. The experts failed to anticipate physical interaction problems that were critical for blind users — they did not identify difficulties finding card, receipt, and money slots, glare from direct lighting, or reflections on the screen. Blind users were more concerned with speech content accuracy, while visually impaired users focused on privacy implications of audible output and speed of interaction. White text on blue backgrounds proved most legible for low-vision users, while bright white backgrounds caused problematic glare. A speech rate of approximately 120 words per minute was found appropriate. Users preferred the linear menu (Solution C) for its potential simplicity but found all menus too slow. The study demonstrated that hardware adaptations — such as tactile guidance to slots, repositioned slots, better key design with raised rectangular shapes, and controlled lighting — are essential alongside software changes for true accessibility. A key privacy concern emerged: users wanted the option to suppress automatic balance announcements.
Relevance
This 1998 study remains highly relevant as self-service kiosks, payment terminals, and ATMs continue to be redesigned for accessibility. The finding that heuristic evaluation by sighted HCI experts fails to identify many of the most critical barriers for blind users is a powerful argument for including disabled users in usability testing — expert review alone is insufficient. The disconnect between expert predictions and actual user difficulties, particularly around physical interaction and environmental factors like glare and slot location, underscores that accessibility evaluation requires lived experience perspectives. The privacy concerns raised about speech output in public settings anticipate ongoing debates about audio feedback at accessible kiosks and payment terminals. For practitioners, the paper offers concrete guidance on screen design for low vision (blue backgrounds, sans serif fonts, minimal information per screen) and speech interface design (short sequences, skip functionality, echo feedback, default options for frequent tasks). The study's core lesson — that software-only accessibility retrofits have fundamental limits when hardware is inaccessible — remains a critical consideration for organisations deploying public-facing technology.
Tags: ATM accessibility · self-service terminals · blind users · low vision · speech output · usability testing · heuristic evaluation · kiosk accessibility · public access terminals · tactile feedback