A qualitative framework for understanding hearing-aid gain self-adjustment

Technical Summary: A Qualitative Framework for Understanding Hearing-Aid Gain Self-Adjustment

Problem Statement
Current hearing-aid personalization relies heavily on clinician-led adjustments, a process fraught with theoretical and psychometric flaws. Clinicians cannot perceive sound as patients do, and patients often struggle to provide reliable, consistent descriptions of complex acoustic environments or recall specific listening situations. Consequently, clinical personalization risks becoming a "futile, frustrating and protracted procedure." While self-adjustment methods (via apps or over-the-counter devices) offer an alternative, they frequently label resulting settings as "preferences" without interrogating the rationale behind user choices. Existing quantitative analyses of self-adjustment (e.g., duration, button presses) fail to capture the cognitive processes, internal criteria, and decision-making strategies users employ. Furthermore, the design of self-adjustment interfaces remains a significant human-computer interaction (HCI) challenge, where functional labels may confuse users or bias their interactions, while label-less interfaces may lack transparency. There is a critical lack of inquiry into how users navigate these processes and what drives their decisions.

Methodology
The authors employed a Think-Aloud Protocol (TAP), a qualitative HCI method, to investigate the cognitive processes of hearing-aid users during self-adjustment.

• Participants: 36 adult hearing-aid users (aged 61–82) with mild-to-moderate bilateral hearing loss.
• Procedure: Participants performed nine self-adjustment trials while verbalizing their thoughts concurrently. Each trial involved adjusting Frequency-Gain Response (FGR) to preference using three different stimuli (speech in quiet, speech in cafeteria noise, and music) and three distinct unlabelled interface controls:
  1. One slider controlling bass and treble in opposite directions.
  2. Two independent sliders for bass and treble.
  3. Three independent sliders for bass, mid, and treble.
• Data Collection: Audio and video recordings captured verbalizations and non-verbal behaviors (gestures, facial expressions).
• Analysis: The data underwent reflexive thematic analysis (Braun & Clarke, 2006, 2021). This inductive approach avoided predefined theories or intercoder agreement calculations, focusing instead on interpreting the procedural context of the adjustments. The analysis organized verbalizations into three temporal stages (before, during, and after adjustment) to identify recurring patterns and archetypes.

Key Contributions and Results
The study produced a qualitative framework defining user approaches to self-adjustment as a continuum between two archetypes: "Exploratory" and "Anticipatory."

1. The Exploratory Archetype:

   • Mindset: Driven by the question, "What are the best possible settings given the available options?"
   • Behavior: Users initially assess control functions by testing both directions of sliders (increasing and decreasing) to gain perceptual certainty.
   • Evaluation: Judgments are hedonic and affective, often comparing the current sound to the immediately preceding setting (e.g., "better/worse").
   • Process: Often involves sequential testing of sliders from left to right in complex interfaces.

2. The Anticipatory Archetype:

   • Mindset: Driven by the question, "How can I achieve the sound that I want?"
   • Behavior: Users begin by evaluating the initial sound to identify specific issues (e.g., "too shrill," "not loud enough"). Adjustments are guided by prior experiences and specific goals to resolve these identified issues.
   • Evaluation: Adjustments are targeted to find optima within an anticipated range. Users evaluate whether the change resolved the specific problem identified earlier.
   • Process: Often complaint-driven; if the initial sound is acceptable, users may crossover to an exploratory mindset simply because interaction is required.

3. Perceptual Descriptors and Interface Interaction:

   • Participants' verbal descriptors (e.g., "tinny," "clear," "bassy") generally aligned with expert-based data and the direction of gain changes (e.g., attenuation described as "quieter" or "less bass").
   • Users often struggled to find precise vocabulary, relying on gestures or expressing frustration, particularly with the high-frequency slider in the three-slider interface.
   • Trade-offs: Significant trade-offs were observed, particularly with speech-in-noise stimuli, where users had to compromise between clarity and background noise levels.
   • Reactivity: Comparing the TAP session to a previous session without TAP, the study found that concurrent verbalization led to a statistically significant increase in low-frequency gain (3–4 dB), suggesting that verbalizing a desire for "bass" may influence the final adjustment (reactivity).

Significance and Claims
The paper claims that this qualitative framework is the first to utilize concurrent TAP to map the cognitive strategies behind hearing-aid gain self-adjustment. Its primary significance lies in:

• Redefining Self-Adjustment: Moving beyond the label of "preference" to understand the dynamic, evolving goals and strategies (exploratory vs. anticipatory) users employ.
• Validating Active Control: The study suggests that active control over sound (as opposed to passive pairwise comparisons) promotes the operationalization of sound descriptors, allowing users to better understand the relationship between acoustic changes and perceptual outcomes.
• Clinical and Design Implications: The framework offers a tool for clinicians to understand patient vocabulary and decision-making styles, potentially facilitating "shared decision-making." For device designers, it highlights the need for interfaces that accommodate both exploratory users (who need to test ranges) and anticipatory users (who need to solve specific problems).
• Ownership: The authors posit that engaging users in this process, even if it does not drastically change the final FGR settings, promotes greater ownership and understanding of the personalization process.

The authors remain modest regarding their claims, acknowledging limitations such as the subjective nature of reflexive thematic analysis, the potential for reactivity due to verbalization, and the fact that participants had prior exposure to the interfaces, which may influence the generalizability to users without prior exposure. They conclude that while TAP is feasible, future research must apply these methods in realistic, self-guided environments to further develop the model.