Keywords

1 Introduction

Although most research on affective computing focuses on software applications [27], there is an emerging interest in extending affective computing to applications utilizing physical materials [1, 33, 36]. In the past few decades we have seen the development of smart and programmable materials that grow, adapt and transform to respond to environmental conditions and accommodate functional needs [25, 35]. We have also seen a significant shift in response to human-computer interaction, through the development of tangible, embodied, and other novel interfaces that allow the coupling of physical and digital information through the direct manipulation of materials [17, 37]. In addition, we have seen programmable materials offering novel solutions in biomedical applications through their use as artificial muscles [3, 4, 29], and augmented physical and sensory experiences when incorporated in body suits [11, 24]. It is becoming more and more apparent that materials have qualities that can help us respond better not only to functional needs, but also to our health and emotional needs, enhancing our wellbeing.

Recent applications in affective computing have demonstrated that technological interventions utilizing rhythmic sensory stimuli, such as sound, vibration and light can aid in emotion regulation by affecting our physiological functions and interoceptive awareness [9, 14]. Interoceptive awareness, the awareness of the physiological state of our own bodies, is highly interrelated to our emotional awareness - our ability to process and express our emotions [10, 19]. Disruption in emotion processing and expression has been associated with various serious mental health and developmental disorders [19, 20]. The fact that sensory stimuli can have an impact on emotion regulation has been associated with the phenomena of physiological synchrony and emotional contagion [14, 15, 22] and demonstrates a promising path for non-invasive interventions for emotion regulation.

In this study we take advantage of the affective impact of material sensory properties to explore the effect of programmable materials as emotion regulation systems. We designed a controlled study to test how different paces of rhythmic haptic action (slight pressure and warmth), induced through a programmable Affective Sleeve that we developed, can affect a subject’s psychophysiological signals. In particular, we test whether the haptic sensations induced by the sleeve may influence breathing rates and changes in calmness.

2 Related Work

2.1 Breathing Regulation and Interoceptive Awareness

As indicated by meditation practices, cultivating awareness of one’s bodily signals is important for reducing anxiety and promoting calmness. Sensory stimuli can help promote and guide such awareness. For example, Canales-Johnson et al. show that auditory heartbeat feedback plays a significant role in maintaining homeostasis and regulating our behavior [8]. The benefits of controlled slow breathing can also be achieved through the aid of technological interventions. As demonstrated by Elliot et al., device-guided breathing has positive effects on graded blood pressure reduction [13]. Based on such findings, several mobile applications have been developed to provide guided breathing by providing visual, auditory or tactile stimuli [7, 30]. Material-based mindfulness interventions have also been proposed though specially designed furniture, that can focus one’s attention on different parts of the body through incorporated heating mechanisms, or guide one’s breathing through programmed lighting patterns [33].

Although cognitive access to our interoceptive processes has been associated with healthy emotion regulation mechanisms, interestingly, high levels of interoceptive awareness have been positively correlated with anxiety symptoms [12, 28]. Research shows that higher than necessary levels of interoceptive awareness can be a symptom of emotion dysregulation [19]. Aligned with such findings, studies have demonstrated that providing false interoceptive awareness (false heart or breathing rate feedback) through technological intervention can regulate the heart or breathing rate to reduce anxiety levels [9, 14]. For example, Ghandeharioun and Picard demonstrated that barely perceivable auditory and visual interventions on a computer, rhythmically oscillating to match relaxed breathing rates, can increase both focus on a work-related task and calmness [14].

The development of the Affective Sleeve takes into account research findings on interoceptive awareness and the regulation of physiological signals though technological intervention. We suggest that the Affective Sleeve can regulate breathing through rhythmic sensations of warmth and slight pressure to reduce stress levels.

2.2 Benefits of Touch in Emotion Regulation

Through studies on married couples, Holt-Lunstad et al. provided evidence of the positive effects of human touch on health, particularly for lowering anxiety levels [16]. The effects of human touch on health have also been achieved through technological interventions. For example, Sumioka et al. demonstrated that the stress-reducing effect of interpersonal touch can also be achieved through huggable devices that provide tactile sensations [34]. In the HCI field, Bonanni et al., discussed the design and therapeutic effects of wearable haptic systems that have the ability to record and play back human touch [6]. Other studies, like the one by Boer et al. have explored the aesthetic potentials of haptic wearable interfaces and their impact on sensory experience [5].

Related to the aforementioned studies, our goal is to utilize the Affective Sleeve to achieve therapeutic effects similar to those of human touch. In the case of the Affective Sleeve, the sensation of touch is defined by both the stimulus of warmth and the stimulus of pressure. Warmth alone has significant benefits for health. Lowry et al. provide evidence on how warmth may alter neural connections regulating mood and cognitive processes, including serotonergic circuits that affect stress-related disorders [23]. Pressure stimulus of the touch sensation also has significant benefits in health. Research demonstrates that pressure stimulus in massage therapy significantly decreases the perception of pain, improves mood and lowers anxiety levels [18, 21].

Finally, recent contributions to the HCI and affective computing communities have focused on valence and arousal measures of temperature changes and vibrotactile stimuli, demonstrating that such stimuli can have a distinct impact on psychophysiology [31, 32, 38].

Through the intervention of the Affective Sleeve, we plan to examine the impact of pressure and warmth stimuli on human psychophysiology. We aim to provide the beneficial effects of human touch in cultivating calmness and lowering anxiety levels. In particular, we plan to examine the impact of different paces of rhythmic pressure and warmth stimuli on psychophysiology.

3 Method

3.1 The Affective Sleeve

We designed the Affective Sleeve as a series of autonomous cuffs to explore the impact of different types and paces of haptic action on wearers’ psychophysiology (Fig. 1). Each cuff can be actuated independently and controlled for various durations of time to provide flexibility for setting haptic rhythms. The final prototype consists of six cuffs of adjustable circumference that cover an adult’s forearm.

Fig. 1.
figure 1

Prototype of the Affective Sleeve. Activation of each cuff occurs in sequence from the palm towards the elbow.

Full size image

The rhythmic haptic action of the Affective Sleeve was made possible by sandwiching 0.5 mm Nitinol wire (a shape memory alloy) between two insulating layers of thick felt in each cuff (Fig. 2). Although the wires took on an arc shape when sewn into the Affective Sleeve, they were “programmed” to return to a flat, linear shape when heated to temperatures above 45 ℃. Based on this property, we were able to pass current through the wires to transform the geometry of each cuff and cause a slight sensation of pressure and warmth on the forearm. When the electrical current is disconnected and the wires cool, they are forced back into an arc due to the geometry of the sleeve and stiffness of the fabric.

Fig. 2.
figure 2

The Affective Sleeve consists of a sequence of independent cuffs containing nitinol wire, a shape memory alloy, sewn between two layers of thick felt. Current passed through the nitinol wire activates the cuffs, creating pressure and warmth.

Full size image

We chose to use a shape memory alloy because it allowed us to simultaneously induce pressure and warmth by actuating a single material. However, the limitation of the shape memory alloy used is that it allows only for a one-way transformation. Thus, a mechanical solution was necessary to create a repetitive actuation with the wire. After testing various solutions with plastics, elastics and springs, we decided that designing the sleeve using cuffs made of a stiff felt fabric created the best actuation and reversible behavior, as well as the most comfortable wearable experience.

For this study, we programmed each cuff of the Affective Sleeve to be activated periodically from the palm toward the elbow. The time of the activation for each cuff was set in response to the duration of the wearer’s breathing cycle (inhale and exhale) after which it returned to the cool state. The pace of the rhythmic haptic action varied according to three test groups, as will be explained in the experiment procedure. The periodic activation of each cuff creates the sensation of warmth and pressure along the forearm. The temperature of the fabric in contact with the human skin ranges from 26 ℃ (off) to 38 ℃ (on) (Fig. 3).

Fig. 3.
figure 3

Diagram depicting the periodic change of temperature over time for each cuff.

Full size image

We intended the Affective Sleeve to be an example of emotion-regulating clothing, stimulating different locations on the body. Unlike other wearable solutions for emotion regulation, such as the EmotionCheck application by Costa et al., which provides steady vibration stimuli in one specific location on the wrist, the Affective Sleeve induces sequential tactile stimuli at six distinct locations along the forearm, creating a flow of warmth and slight pressure [9]. The idea of the flow of induced rhythmic haptic sensations along the forearm was inspired by the body scan technique in breathing meditation that transfers the focus of attention along the body while synchronizing with rhythmic breathing, as well as by techniques of deep pressure stimulation along the body in massage therapy [2, 18, 26]. These connections, while not deeply explored in this study, are important aspects of the sleeve’s design that could be further explored in the future.

3.2 Experiment Method and Hypothesis

To evaluate whether the haptic action of the sleeve has a positive effect on lowering anxiety levels, we designed a randomized controlled experiment to measure the difference between stressed-state and relaxed-state physiological signals. Given prior research findings on false interoceptive awareness and sensory stimulation we hypothesized that the pace of the sleeve would have a positive correlation with breathing rate. Our hypothesis was that slow and fast rhythmic haptic action of the Affective Sleeve would decrease and increase, respectively, participants’ psychophysiological symptoms of stress.

We recruited 18 healthy adult college students to participate in a study that would require them to wear the sleeve while taking a spatial cognition quiz in order to induce high stress levels. The participants were told that the quiz would measure the effect of the sleeve on their cognitive performance. We chose to use the quiz as a stressor because simulating an exam setup was relevant to students and could be easily controlled and reproduced for further studies.

To ensure that the measured effects were a result of the haptic action of the sleeve and not a result of the texture of the fabric or other factors, we randomized participants into three test groups: the Control Group (inactive sleeve, but participants were told it was active and barely perceptible), the Fast Group (fast rhythmic haptic action of the sleeve) and a Slow Group (slow rhythmic haptic action of the sleeve). All participants wore the Affective Sleeve for the habituation and performance tasks of the experiment. Based on our hypothesis, we expected that participants in the Slow Group would exhibit lower electrodermal activity (EDA), a noninvasive measure of the sympathetic nervous system’s “fight or flight” response, and lower breathing rates than those in the Control and Fast Groups. Also, we expected that participants in the Fast Group would exhibit higher increases in EDA and breathing rates than those in the Slow and Control groups. Such results would provide evidence that the haptic action of the sleeve has an impact on participants’ stress levels and associated physiological processes.

3.3 Experimental Procedure

The participants were told in advance that they would be compensated with a $10 gift card, and given an additional $50 gift card if they achieved one of the two highest scores on the spatial cognition quiz. We gave the opportunity for this additional compensation in order to make the task competitive, eliciting more stress and aspiration for good performance.

We decided to withhold the true aim of the study at the onset of the experiment to ensure that participants would not consciously manipulate their physiology or behavior in a way that would affect their breathing rate and skin conductance levels. Instead, the participants were told that they would participate in a study measuring the effect of the sleeve on cognitive performance.

In addition to the Affective Sleeve, participants were asked to wear an Empatica E4 wristband sensor to record EDA levels, as well as a Zephyr BioPatch chest sensor to record respiration rate and its variability (Fig. 4). Because participants would need to use their dominant hand for the performance task, the programmable sleeve was worn on the non-dominant hand and the E4 wristband was worn on the dominant hand.

Fig. 4.
figure 4

Experiment setup and procedure. Participants wore the Zephyr Biopatch on their chest, the E4 sensor on their wrist and the Affective Sleeve on their forearm.

Full size image

The experiment consisted of four phases: (1) establishing baseline conditions, (2) habituation task, (3) performance task and, (4) qualitative surveys (Fig. 5). In the first phase of the experiment, participants wore the E4 wristband and Biopatch sensor while watching a relaxing nature documentary for 15 min. to collect relaxed-state baseline conditions for breathing rate and EDA. Participants were not wearing the sleeve during this task. The baseline measurements allowed us to account for variations in skin conductance and respiration rates among individuals at rest.

Fig. 5.
figure 5

Participants watched a nature documentary to collect baseline metrics for breathing rate and EDA (left). They then completed the habituation and performance tasks while wearing the Affective Sleeve (middle). Finally, participants filled out qualitative surveys describing their experience (right).

Full size image

The sleeve’s action was programmed to be different for each test group. For the Control Group, the haptic action was inactive. For the Slow Group, the haptic action was programmed to match the average relaxed-state breathing rate of individual participants. For the Fast Group, the haptic action was programmed to be 25% faster than the average relaxed-state breathing rate of individual participants. The pace of haptic action for the Fast Group was chosen based on the material properties of the sleeve (how fast the material could actuate) and initial experiments exploring the perceivable difference in actuation rates.

In the habituation and performance tasks, participants wore the Affective Sleeve, the E4 sensor and the Zephyr Biopatch sensor. The habituation and performance task required taking a spatial reasoning quiz of multiple-choice format. The duration of the habituation task was 10 min. but participants were allowed to finish earlier if they completed all the questions in less than the allocated time. The participants were told that results of the practice quiz would not count towards the final score. The habituation task was necessary so that participants’ psychophysiological response would not be influenced by the novelty of the intervention.

In the performance task, we designed the quiz to be impossible to complete within the given 15-min. timespan. The average score among all participants was 36%, the median 33%, the highest 57%, and the lowest 18%. The quiz consisted of multiple-choice spatial reasoning questions. We chose a spatial reasoning test as a method for eliciting stress to make sure that all participants (coming from different departments) would be able to respond to the questions. We wanted the stress to be elicited not because of a lack of knowledge of a particular subject but because of cognitive demands and time pressure.

After the performance task, the participants were asked to remove the Affective Sleeve and sensors and proceed to the qualitative survey task and oral interviews. The participants first filled out a multiple-choice questionnaire about their experience of wearing the sleeve during the study. Second, the participants filled out the Perceived Stress Scale (PSS) questionnaire, which has been scientifically verified and widely used. The PSS was used as a post-study screening method that would allow us to evaluate whether very high stress responses and/or poor performance during the study were due to other circumstances in the participant’s life. While the mean for our participants (16.27) was much higher than the mean provided in the PSS (12.1 for males, 13.7 for females), the association of PSS results to quiz performance and physiological data for each individual in the study did not show any significant correlation, which supports that there was no prior stress bias or cognitive ability bias among the random group assignments. Finally, the study ended with a brief oral interview regarding the experience of the study.

The setup of the experiment remained the same throughout the study: each participant sat at a desk with a computer and remained seated until the end of the study. Participants were tested one at a time. The experimenters interacted with the participants between the phases of the study to adjust or remove the sensors if required. Otherwise, participants did not interact with anyone during the tasks of the experiment. The room was isolated from other distractions to the best extent possible in a common academic environment.

4 Results

4.1 Physiological Data Analysis

The 18 participants (aged 19–34) were all students at Massachusetts Institute of Technology (MIT); 16 were graduate and two were undergraduate students. Of the participants nine were female and nine were male. Out of the 18 participants, 15 were students in the School of Architecture and Planning, one was a student in the Mathematics department, one in Bio-electrical Engineering, and one in Materials Science.

For every participant, we computed the mean and standard deviation for the first three phases of the experiment (baseline, habituation, performance), excluding any data outside two standards of deviation from the mean. Data from the first three phases were scaled horizontally to fit within a 15-min., 10-min., and 15-min. window respectively. After removing noise, we normalized the data per participant (maximum over the whole session = 1, minimum over the whole session = 0) to eliminate any difference in individual ranges. We calculated the difference between the mean value of the performance phase and mean value of the baseline phase for each participant. We then calculated the average difference between the performance phase and baseline phase for each test group as a whole. Given that this pilot study had a small n and we did not have a normal distribution of data, we used the Kruskall-Wallis test to evaluate our results.

Results from the EDA data (Fig. 6) show that the mean change in EDA in the Slow Group was 0.095 µS less than the Fast Group and 0.078 µS less than the Control Group. However, with such a small n, we cannot reject the null hypothesis. Additional studies must be conducted to determine how the haptic action of the Affective Sleeve affects EDA.

Fig. 6.
figure 6

The average difference in EDA from baseline to performance tasks.

Full size image

The results from the breathing rate data are more promising (Fig. 7). A comparison of the mean change in breathing rates from relaxed to stressed conditions show that subjects in the Slow Group breathed 0.021 bpm faster than the Control Group and 0.142 bpm slower than the Fast Group. Given our small study, we cannot reject the null hypothesis for these results. However, when comparing the Fast and Control Groups, the data show that the Fast Group breathed 0.164 bpm faster than the Control Group, (p = 0.0105). This result suggests that the rate of haptic action of the sleeve correlates positively to breathing rate; that is, a faster haptic action is associated with a higher breathing rate.

Fig. 7.
figure 7

The average difference in breathing rate from baseline to performance tasks.

Full size image

4.2 Self-reported Data Analysis

Participants were given qualitative surveys designed to evaluate their experience of wearing the sleeve. Out of the 13 questions, 11 had a multiple-choice format providing a scale of 0 to 4—0 indicating negative, 2 neutral, and 4 positive. Specific responses associated with the scale were provided for each question. Two out of the 13 questions prompted written feedback. We compared the results of the three groups to gain insight into the experience of wearing the sleeve and its effectiveness for regulating anxiety.

When asked about the comfort of the sleeve, 50% of the Control Group reported that they were either neutral or uncomfortable, and only 16.7% reported they were very comfortable. In the Slow Group, 50% of participants reported that they felt a bit comfortable and 50% reported that they felt very comfortable. In the Fast Group, 66.7% of the participants reported that they felt a neutral level of comfort. These responses suggest that the pace of haptic action is negatively correlated with perception of comfort: the faster the pace of haptic action, the less comfortable it is. Additionally, the absence of haptic action is perceived as more uncomfortable than either slow or fast haptic action (Fig. 8).

Fig. 8.
figure 8

Participants’ responses to the question “Regarding comfort, wearing the sleeve felt…”

Full size image

To inquire whether the sleeve affected participants’ performance, we asked participants whether they felt that the sleeve was distracting or helped them focus. Of all participants, 66.7% reported that it neither helped them focus nor felt distracting, 16.7% of participants reported that it helped them focus a bit, 16.7% of participants felt that it was a bit distracting, and none felt it distracted them significantly. In addition to the self-reported data, the mean scores from the performance task for each test group were very similar: the Control Group averaged 32%, the Slow Group averaged 34% and the Fast Group averaged 34%. These results suggest that the participants’ performance was not greatly affected by the sleeve.

Regarding the warmth produced by the sleeve, 66.7% of the Control Group reported the warmth was neutral and only 33.3% felt it was a bit calming. In the Slow Group, 16.7% reported the warmth was neutral and 83.3% reported it was either a bit calming or very calming. From the Fast Group, 50% reported that the warmth was neutral and 50% reported it was either a bit calming or very calming (Fig. 9). Overall, participants in the Slow Group perceived the warmth of the sleeve as more calming than the Fast or Control Groups.

Fig. 9.
figure 9

Participants’ responses to the question “Was the sleeve’s warmth calming?”

Full size image

When asked to describe the sensation of wearing the sleeve, all participants in the Slow Group described the experience as positive. Responses ranged from simply “Pleasant,” “Warm and comforting,” to more imaginative such as “Felt like a cat was lying on my arm.” In the Fast Group only half of the participants found the experience positive; the other half described it as slightly discomforting, mostly due to the sensation of movement or pressure. In the Control Group, 83.3% of participants described the experience as neutral, reporting that they did not notice any warmth or movement. Only one participant in the Control group reported the perception of change of temperature of the sleeve, which may have been the result of placebo effect. Participants’ feedback regarding their experience of wearing the sleeve suggests an association between slow pace of haptic action (matching their resting respiration rate) and participants’ positive experience, and an association between fast pace of haptic action and participants’ slightly negative experience.

Finally, participants were asked if they would wear the Affective Sleeve in their everyday lives if it was proven to decrease stress levels (Fig. 10). Of the Control Group 66.7% were neutral and 33.3% responded “most likely.” In the Slow Group 33.3% responded neutrally or “probably not” while 66.6% responded either “most likely” or “definitely.” In the Fast Group, 50% of participants responded neutrally and the remaining 50% responded “most likely” or “definitely.” This and other survey responses suggest that participants in the Slow Group had a more calming and positive experience than the Fast and Control Groups and that the valence of participants’ experience while wearing the sleeve possibly biased their opinion of its application in everyday life, as the Slow Group was the most receptive to the idea of wearing the sleeve.

Fig. 10.
figure 10

Participants’ responses to the question “Would you wear a programmable sleeve in your everyday tasks if the sleeve proves to decrease your stress levels?”

Full size image

5 Conclusion

The results from the analysis of the breathing rate data demonstrate a positive correlation between the pace of the haptic action and the change in measured signals, indicating that the rate of activation in the Affective Sleeve may have a subtle impact on breathing regulation. These results suggest that a faster pace of haptic action along the sleeve may increase the rate of respiration. However, our quantitative analysis does not yet demonstrate that a lower pace of haptic action can lower breathing rates. Further quantitative analysis with a greater number of participants and a slightly different experiment design are necessary to prove or disprove this statement of causality.

As our current hypothesis assumes that lowering breathing rate reduces anxiety levels, additional studies will be necessary to associate the change in breathing rates with the change in anxiety levels. Also, based on the comparison of the mean change in EDA per group, we believe that with a larger group of participants we may begin to see statistically significant results in the EDA results. Future studies with a greater number of participants and possibly with different methods for inducing different kinds of stress (e.g. in a social context) need to be conducted for further conclusions.

The results from the self-reported data show a positive correlation between calmness and slow pace of haptic action, and a negative correlation between calmness and fast pace of the haptic action. The experience of wearing a sleeve producing slow periodic haptic action, at a rate matched to resting respiration rate, was perceived as more comforting and positive than the experience of wearing a sleeve with no haptic action or 25% faster-than-resting-respiration haptic action. The experience of wearing a sleeve producing fast haptic action was perceived as more discomforting than wearing a sleeve with no haptic action. The results from the self-reported data suggest that the slower pace of the sleeve’s haptic action may help promote calmness.

As the haptic action in this study combines the stimulus of warmth, the stimulus of pressure, and movement along the arm, future studies on the impact of each of these sensations on calmness independently would be useful. Because shape memory alloys require heat activation for their actuation, perhaps an alternative future sleeve design could include inflatable materials as one component and heating wires as another component, each controlled independently, to allow for individual testing of the warmth and pressure stimuli.

Finally, it is important to note that the results of the performance task indicate no correlation between overall performance and pace of haptic action, indicating that the haptic interventions did not distract participants from their ability to perform. Taking into account the results of both the physiological and self-reported data analysis, we can conclude that the effect of Affective Sleeve on emotion regulation seems promising, although more studies need to be conducted to explore the range of emotional influence that is possible.