← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

Towards Inclusive External Human-Machine Interface: Exploring the Effects of Visual and Auditory eHMI for Deaf and Hard-of-Hearing People

Wenge Xu , Foroogh Hajiseyedjavadi , Kurtis Weir , Chukwuemeka Eze , Mark Colley

DOI PDF Program page Page JSON

This is a solid, population-expanding CHI paper: it does not merely tweak an eHMI design, but re-frames the problem around Deaf and Hard-of-Hearing pedestrians and backs that move with both formative and experimental evidence. The contribution is strongest as a scoped empirical finding plus design guidance, not as a universal eHMI theory.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

This paper’s main value is that it shifts eHMI research from a generic pedestrian model to an explicitly inclusive one, with Deaf and Hard-of-Hearing people treated as a first-class user population rather than an afterthought. The abstract and contribution framing make the gap clear: prior work had not attended to DHH people, and the authors respond with a two-stage program that combines formative focus groups with a VR user study. That is a meaningful departure from common practice in this area, because it couples stakeholder-informed requirement gathering with controlled evaluation instead of assuming that existing eHMI concepts transfer unchanged. The novelty is not a new theory or a radically new interaction primitive; it is a population-specific empirical package that produces design requirements and measurable effects for visual versus auditory eHMIs. The validation is reasonably strong for the claims made: 32 participants, including 16 DHH, in a controlled non-signalised crossing scenario, with outcomes spanning trust, usefulness, perceived safety, gaze, and stepping behavior. Importantly, the results are differentiated rather than uniformly positive: both visual and auditory eHMIs improved subjective judgments, but only visual eHMIs changed several behavioral measures. That nuance strengthens the paper because it avoids collapsing perception and action into one effect. The limitations are also clearly scoped in the provided evidence: the study is tied to one controlled scenario, one UK setting, one experimental session, and a narrow set of modalities. So the paper should be read as a strong, well-bounded contribution to inclusive AV interface design, with the most defensible claim being that DHH-centered eHMI evaluation reveals both shared and modality-specific effects that generic pedestrian studies would miss.

What Changed

Canon before

eHMI work has largely centered on general pedestrian communication and signal design, with little explicit attention to Deaf and Hard-of-Hearing pedestrians or to comparing visual versus auditory eHMI effects for that population.

Departure from common sense

The paper directly contests the default assumption that eHMI design can be evaluated for a generic pedestrian audience by stating that prior work gave no attention to Deaf and Hard-of-Hearing people, then centering DHH participants in both formative and experimental evaluation.

Actual novelty

The paper’s novelty is a first formative focus-group study with DHH people and stakeholders, paired with a first VR user study involving DHH participants to compare visual and auditory eHMIs for AV-pedestrian crossing. That combination yields population-specific design requirements and empirical effects rather than only a generic eHMI comparison.

Evidence

The evidence supports a mixed-methods contribution: a formative focus-group phase to extract design requirements, followed by a controlled VR user study with 32 participants, including 16 DHH. The reported findings distinguish perceptual outcomes from behavioral outcomes: both visual and auditory eHMIs improved trust, usefulness, and perceived safety, while only visual eHMIs changed several crossing-behavior measures. The paper also explicitly states scope limits around a single controlled non-signalised crossing, UK-only testing, and a single experimental session.

Limits

Method limits

The study uses a single controlled VR scenario and a single experimental session, so the behavioral findings are tied to that setup rather than to broad real-world traffic conditions. The paper also evaluates only one singular visual eHMI and speech as the auditory modality, which narrows the design space covered.

Deployment limits

The results are most directly applicable to AV pedestrian-crossing interfaces for DHH people in controlled crossing contexts. Translation to other road layouts, other countries, other auditory modalities, or deployed vehicle systems remains uncertain.

Boundary conditions

Findings are bounded by the non-signalised crossing scenario, the UK setting, the specific visual and speech-based auditory eHMIs tested, and the VR-based experimental environment.

Manual check: No additional manual check is strictly required from the provided evidence, but the exact wording of the novelty claim should be verified against the full paper if a stricter first/first-ever assertion is needed.

Position in field

This sits at the intersection of inclusive AV interface design and pedestrian communication research, extending eHMI work from general audiences toward a previously underexamined DHH population and providing both formative requirements and scenario-based evidence.

Abstract

External Human-Machine Interfaces (eHMIs) have been proposed to facilitate communication between Automated Vehicles (AVs) and pedestrians. However, no attention was given to Deaf and Hard-of-Hearing (DHH) people. We conducted a formative study through focus groups with 6 DHH people and 6 key stakeholders (including researchers, assistive technologists, and automotive interface designers) to compare proposed eHMIs and extract key design requirements. Subsequently, we investigated the effects of visual and auditory eHMI in a virtual reality user study with 32 participants (16 DHH). Results from our scenario suggesting that (1) DHH participants spent more time looking at the AV; (2) both visual and auditory eHMIs enhanced trust, usefulness, and perceived safety; and (3) only visual eHMIs reduced the time to step into the road, time looking at the AV, gaze time, and percentage looking at active visual eHMI components. Lastly, we provided five practical implications for making eHMI inclusive of DHH people.