TL;DR
This paper explores how people synchronize their walking gaits while talking on mobile phones, even without visual cues. They found that vibrotactile feedback of a partner's footsteps can induce synchronization, and surprisingly, spontaneous conversation alone can also lead to gait alignment, suggesting a deep, unconscious connection between communication and movement.
What Problem Does This Paper Address?
Conversation is a fundamentally social activity, and humans naturally use a variety of nonverbal cues to build rapport and ensure smooth communication. When we switch to mobile phone conversations, many of these crucial nonverbal channels (like body language and visual cues) are lost. This can lead to a more formal interaction and potentially hinder the development of a strong connection. The researchers are interested in whether technology can augment mobile communication to restore or even enhance this unconscious synchrony, and if this synchrony can serve as a proxy for communication quality. Specifically, they investigate if people can synchronize their physical movements (walking gait) when their only connection is through voice and potentially haptic feedback, even when physically separated.
Methods
The core of this research lies in measuring and analyzing gait synchronization.
- Gait Phase Angle Estimation: Accelerometers embedded in mobile devices (HP iPAQ 5550 with MESH sensor backpacks) captured walking data. To extract gait phase, they used a Hilbert transform. This technique allows for the estimation of instantaneous phase and amplitude from oscillatory signals, which is crucial for analyzing rhythmic movements like walking. The data was filtered to ensure it was "narrow-band" around typical walking frequencies.
- Synchronization Detection: To quantify the level of alignment between two participants' gaits, they employed techniques from synchronization theory. This involved generating "synchrograms," a visual representation of how the relative phase between two oscillators changes over time. They also calculated a "synchronization index" (An,m) to provide a quantitative measure of the degree of synchronization for specific ratios (e.g., 1:1). A threshold of An,m ≥ 0.5 was used to define "synchronous behavior."
- Vibrotactile Feedback: A key experimental manipulation involved using vibrotactile feedback. Participants' devices could vibrate to signal either their own footsteps or their conversation partner's footsteps. This allowed them to investigate the impact of direct haptic information on gait alignment.
- Experimental Tasks: Three distinct communication tasks were designed to vary the level of cooperation and naturalness:
- Scripted Task: Reading a pre-defined script (Monty Python's "Dead Parrot" sketch).
- Image Description Task: Describing cartoon images.
- Spontaneous Conversation Task: Free-form conversation on any topic.
Datasets
Not explicitly detailed as "datasets" in the traditional machine learning sense. The study collected data from 22 undergraduate or graduate students (11 pairs). After excluding one pair due to data recording issues, data from 10 pairs was analyzed. The data collected included:
- Speech recordings (separate channels for each speaker).
- Acceleration data from mobile devices (logged via PDAs).
- Experimental instructions displayed on the PDA.
The study did not introduce a new benchmark dataset but rather collected novel data within a controlled experimental setup.
Key Insights and Ideas
- Unconscious Gait Alignment: The most striking finding is that people can synchronize their walking gaits even when talking on a mobile phone and physically separated, relying solely on auditory cues. This suggests a powerful, unconscious mechanism for interpersonal alignment that extends beyond direct visual or physical interaction.
- Vibrotactile Feedback as a Synchronizer: Providing participants with vibrotactile feedback of their partner's footsteps was sufficient to induce gait synchronization, demonstrating the potential of haptic technology to enhance remote communication.
- Spontaneous Conversation is Key: The spontaneous conversation task, where participants could talk freely, consistently showed higher levels of gait synchronization, even without vibrotactile feedback. This highlights the role of natural, engaging conversation in fostering deeper interpersonal synchrony.
- Interference Effects: When vibrotactile feedback was combined with spontaneous conversation, the effect on synchronization became less clear, suggesting potential interference between different sensory inputs and communication modalities.
- Gait Synchronization as a Proxy for Rapport: The paper proposes that gait synchronization could serve as an objective, measurable proxy for communication quality and rapport, offering a valuable tool for evaluating augmented communication systems.
Key Results
- Spontaneous Speech Dominance: The spontaneous speech task, particularly without crosstalk (vibrotactile feedback of partner's footsteps), yielded the highest proportion of time participants' gaits were synchronized (29% of the time, with An,m > 0.5).
- Crosstalk Nuance: While crosstalk was predicted to increase synchronization, the results were mixed. In the spontaneous speech task, synchronization was slightly lower with crosstalk (26.5%) than without. However, in the scripted and image tasks, crosstalk showed a marginal increase in synchronization.
- Task Differences: Significant differences in synchronization were observed between tasks. Spontaneous speech showed significantly more synchronization than both the image and scripted tasks, especially in the absence of crosstalk.
- Image Task Behavior: The image task, even without crosstalk, showed higher synchronization than initially predicted, possibly because participants treated it more like free conversation, engaging in joint laughter and jokes.
- Scripted Task Performance: The scripted and image tasks showed the least synchronization when crosstalk was absent, with only marginal differences between them.
Limitations and Future Work
- Limited Distance: The WLAN range of 100m constrained the distance between participants, meaning they were walking in a park at opposite ends, but not extremely far apart. While participants reported not being aware of their partner's walking, this proximity might still have had a subtle influence.
- Task Interference: The study noted potential interference effects when vibrotactile feedback was combined with spontaneous conversation, suggesting that the interplay of different sensory inputs needs further investigation.
- Subjective vs. Objective Measures: While the study provides objective measures of gait synchronization, further work is needed to empirically link these measures to subjective assessments of rapport and communication quality.
- Future Work: The authors propose a research program to characterize the link between alignment and rapport. They aim to explore how to improve communication quality by increasing synchronization and to understand the phenomenon of users pacing while on the phone. The increasing availability of accelerometers in mobile phones opens up possibilities for logging natural user behavior and correlating it with communication quality and duration. They also suggest exploring different mappings between sensors and displays for augmented communication.
Summary updated