PERCEIVED EXERTION AND
DETACHABLE SMARTWATCH DESIGN

Company: Technicolor
Role: UX and Product PhD Intern (Project Lead)

PROBLEM:

Glanceability and quick access are key strengths of smartwatches, making them ideal for micro-interactions—simple tasks that can be completed in seconds. However, when users need to perform tasks that require sustained interaction, there are significant limitations. To empirically establish the upper limit of comfortable usage time, we conducted a study to measure user fatigue during prolonged smartwatch interactions. Our findings reveal that continuous use leads to noticeable fatigue within just a few minutes, setting an upper bound on effective smartwatch usage that must be considered in application and interaction design.

Recognizing these constraints, we explored innovative product designs that maintain the benefits of glanceability while enhancing usability for longer, more complex interactions. We propose a versatile smartwatch design that can be detached from the wrist and used in multiple configurations based on user context, preferences, and requirements. This detachable design transforms the smartwatch into a more adaptable device, offering improved interaction, display, and sensor capabilities. Through comprehensive need-finding research and user interviews, we identified key scenarios where a detachable smartwatch provides significant advantages. These insights informed the development of prototype applications, demonstrating how this flexibility can expand the range of use cases and improve the overall user experience compared to a traditional, always-worn smartwatch.

OUTPUT: 

1. Two published articles at ACM IMWUT [Quantitative Study] and [Product Research + Prototype evaluation]

2. Patents (pending)

Figure 1. The two poses in which our study was conducted, (A) sitting with elbows resting on armrest, and (B) standing with arm raised.

QUANTITATIVE RESEARCH methods
STUDY PROCEDURE

We conduct a within-subject study, which simulates typical users’ target selection tasks with a smartwatch in different poses over time. We simulate it using three different forms of abstract smartwatch input primitives: 

1. Touch: Participants tap a target once with their finger. 

2. Dwell: Tap and hold a target for a specified time (500ms). 

3. Swipe: Swipe a target in a particular direction (shown on screen) for a specified movement of at least 35 pixels. 

The participants perform the inputs in two different poses (Figure 1), which mimicks commonly held smartwatch usage positions: A) sitting with elbows rested; and B) standing with arm raised. We denote each combination of input and pose as a condition in our study. The participants completed eight 30-second trials in each condition. 

We measure participants’ self-reported exertion with the Borg CR10 scale [5], once before each condition to establish a baseline, and at the end of each 30 second trial. Participant provided their responses verbally and a researcher recorded their response on paper. We recruited 18 participants (8 female, 10 male) via word of mouth. The participants’ ages ranged from 22 to 35 (median = 25.5) and their smartwatch experience varied from non-users to power users.

Participants start the study in one of two poses (sitting or standing). The poses are counter balanced across participants and the conditions (input task) within each pose are randomized. Participants complete 8 trials of 30 seconds in a given condition with a 5-minute break after each condition to rest and to reduce any carryover effects. For each trial, participants select a target as many times as they can in the allotted time. Touch and dwell targets are solid circles with a 35 pixel radius, while swipe targets have arrows indicating a particular direction participants swipe in to select the target. If the participants accurately select the target, it disappears and the next target appears at a random position on the screen. We time-stamp and log all touch events, the count of accurate touches, and total touches during each trial.

DATA ANALYSIS

We collect our data using the ordinal Borg CR10 Scale. The data was not normally distributed according to a Shapiro-Wilk test (p<0.05). We transform our data using ART analysis and the associated ARTool. This transformation aligns the perceived exertion scores for each main or interaction effect and assigns them ranks. This procedure allows us to conduct a parametric three-way repeated measures ANOVA (INPUT x POSE x TIME) for each effect on the transformed data. Table 1 shows the results of our ANOVA.

Table 1. Table summarizing results of ANOVA with pose, input and time as independent variables and perceived exertion as the dependent variable after performing the ART procedure.

Figure 2: Boxplot and linear regression model for PE of subjects in standing pose for all three conditions: Single (left), Dwell (middle), and Swipe (right)

Usage Time: There is a significant effect of usage time on exertion. The raw exertion scores of participants show a consistent upward trend over time. Figure 2 show the boxplots of the raw (non-transformed) exertion values in each input primitive in the standing pose. On the transformed data, we conduct a post-hoc Tukey's pairwise comparisons at each time point and observe a significant difference (p<0.05) between all pairs of time points except between 3.5 and 4 minutes. Moreover, the least-squared mean value of exertion increases with increase in time. Together, it confirms that participants consistently reported higher perceived exertion with increase in usage time. 

We aimed to determine the point at which continuous smartwatch use becomes unsuitable due to user fatigue. Using the CR10 scale, where a score of 4 is considered "somewhat strong" exertion, we identified this as the threshold for significant usability concerns. We aimed to determine when continuous smartwatch use becomes impractical due to user fatigue. Using the CR10 scale, where a score of 4 is considered "somewhat strong" exertion, we identified this as the threshold for significant usability concerns. In our study, participants using the smartwatch in a standing position reached this fatigue level quickly—at just 2.5 minutes for Single Touch and Swipe interactions, and at 3 minutes for dwell interactions.

Overall Insight:

• • Regardless of the input method, users are expected to feel "somewhat strongly" fatigued after only 3 minutes of continuous smartwatch use in a standing position.

These results demonstrate that extended use of smartwatches for any input type can lead to significant fatigue, underscoring the need to optimize smartwatch interfaces and interactions for shorter, more efficient use cases.

DETACHABLE SMARTWATCH DESIGN

The findings of our quantitative study were used to inform a novel/futuristic detachable smartwatch design.

By providing the user with the ability to quickly detach the smartwatch from the wrist, they are free to take on different poses while working with the watch. For example, the smartwatch can transition from the smartwatch's default on-wrist interactive state to one more analogous to a phone where the device is held in one or two hands. 

The ability to detach the watchface offers several affordances that considerably increases the input space available on a smartwatch. To better understand how people may use it, we conducted an elicitation study to identify usage themes.

DISCOVERY SESSION:

We conducted a study with six participants with a goal to elicit usage themes; a common technique used to evaluate prototypes for futuristic designs. We provided the participants with a functional prototype that used magnets to connect the watch face to the wristband. We first encouraged the participants to talk freely of ideas and use-cases that they might have, and then provided them with prompts for affordances such as the ability to rotate and dock a watchface. The study was audio recorded and we conducted a thematic analysis of the participant feedback. 

Based on our thematic analysis, the following usage themes emerged: 

1. Contextual Mode Switch: The transition between different dispositions can act as a trigger to change usage or context mode. For example, if a watch is detached and merely resting on the table, it would adopt a "Do Not Disturb" mode. Thus depending on the context, the watch can perform different functions and/or take different profiles without explicit user intervention. 

2. Heterogeneity: When detached, the smartwatch is not a wearable device anymore, rather depending on the location, it's a suite of sensors to sense different activities on-body or in the environment.

3. Modularity: It is not necessary that the whole watch gets detached; parts of it can remain on the wrist. Different activities may require only a part of the watch e.g., accelerometer to track exercises) and thus only specific "modules" of the watch can be detached which may interact with each other. This theme encompasses the affordance of the interaction between the band and the watch. They are just two pieces of a larger device.

Based on these themes, several prototype applications were developed such as detaching the watchface to use as a phone, as a camera device, as a fitness tracker and a handheld gaming device. We conducted another study to evaluate the potential usefulness of a detachable watch via the affordances it enables. As a proxy for each feature of the detachable watch, we use functional prototypes of a subset of the proposed applications.

A video summarizing the problem and the developed prototypes can be seen below: 

RETRO VALIDATION:

12 participants (5 female, mean age=26) took part in our study. 7 of the 12 had either owned or previously used a smartwatch. The participants were recruited using word of mouth and snowball sampling. 

We demoed our applications to the participants and explained how the applications work. We advised the participants to focus on the concepts (and not on the finesse of the implementation). For each of the applications, the participants answered two questions: (1) "Is this useful?", and (2) "Is this enjoyable?" on a 7-point Likert scale. The participants were also encouraged to provide lower ratings if it was a feature they could not foresee using for any reason. We also captured free-form feedback at the end to better understand their preferences. The participants spent around 20~minutes exploring different applications before giving feedback.