TL;DR
This paper investigates how older adults (OA) perceive and benefit from a social robot's deictic gaze, a crucial non-verbal cue for natural communication. While the robot's gaze improved task performance for all participants, older adults showed less difference in their subjective perception of the robot's social attributes compared to younger adults, suggesting a nuanced interaction between age and non-verbal cue interpretation.
What Problem Does This Paper Address?
As social robots become more integrated into daily life, particularly for assisting older adults (OA), understanding how they communicate non-verbally is critical. Deictic gaze, where a robot directs its "gaze" to an object to guide attention, is a fundamental aspect of human communication. However, the ability to perceive and utilize deictic gaze naturally declines with age, potentially impacting social perception and interaction effectiveness. This research aims to bridge this gap by exploring how age-related differences influence the perception and utility of a social robot's deictic gaze in a collaborative task, with the ultimate goal of informing the design of more adaptive and effective human-robot interactions for older populations.
Methods
The study employed a 2x2 mixed experimental design. Participants were divided into two age groups: adults (A, >18 and <65 years) and older adults (OA, >64 years). The robot conditions were within-subject: a static robot (SR) that always looked at the camera, and a moving robot (MR) that used head movements to mimic deictic gaze by looking at specific ingredients. The task involved preparing a sandwich, guided by the robot's verbal instructions and visual cues.
To measure performance, the researchers recorded:
- Reaction Time (RT): The time between the robot naming an ingredient and the participant clicking it.
- Task Completion Time (TCT): The total time to complete a task (two sandwiches).
Social perception was assessed using questionnaires measuring:
- Mental Demand: Using the NASA-Task Load Index (NASA-TLX).
- Perceived Anthropomorphism, Warmth, Competence, and Discomfort: Using modified versions of the Godspeed Questionnaire Series and the ROSAS scale.
Data analysis involved robust mixed ANOVA tests to account for violations of standard assumptions, with a focus on interaction effects between age group and robot condition.
Datasets
Not discussed in the paper. The study utilized a custom experimental setup and collected data from participants directly.
Key Insights and Ideas
- Deictic Gaze is Universally Beneficial for Task Performance: The robot's deictic gaze (MR condition) consistently facilitated task performance, reducing both reaction times and overall task completion times for all participants, regardless of age. This highlights the inherent effectiveness of this non-verbal cue in guiding attention.
- Age Doesn't Diminish the Magnitude of Gaze Benefit: While older adults generally had slower reaction times, the degree to which deictic gaze improved performance was similar between younger and older adults. This suggests that while OA might be slower overall, the robot's gaze still provides a comparable boost in efficiency for them.
- Older Adults Perceive Robots as More Anthropomorphic: Regardless of the robot's gaze behavior, OA consistently rated the robots as more anthropomorphic than younger adults. This aligns with existing literature suggesting older individuals may attribute more human-like qualities to robots.
- Subjective Perception of Gaze is More Nuanced in OA: While the MR condition generally led to more positive social perception scores (warmth, competence) across the board, OA's subjective experience of anthropomorphism and discomfort seemed less affected by the deictic gaze compared to younger adults. This suggests that while OA benefit from the gaze functionally, it doesn't translate into as pronounced a shift in their subjective feelings about the robot's social attributes.
- Many Participants Don't Notice the Gaze: A significant portion of participants (42%) did not report noticing the difference between the static and moving robot conditions. Interestingly, within this unaware group, older adults were disproportionately represented, and only age effects (not robot condition effects) were found for task performance. This raises questions about how subtle non-verbal cues are perceived when not explicitly attended to.
Key Results
- Task Performance:
- A significant facilitation effect of the moving robot (MR) was observed for both Reaction Time (RT) and Task Completion Time (TCT) across all participants.
- Older Adults (OA) generally exhibited higher RTs than younger adults (A).
- The strength of the facilitation effect (how much faster participants were with MR vs. SR) was not significantly different between age groups for both RT and TCT.
- Social Perception:
- A significant main effect of age showed OA perceived robots as more anthropomorphic.
- The moving robot (MR) condition scored significantly higher than the static robot (SR) condition across most social perception scores (anthropomorphism, warmth, competence), except for discomfort.
- A significant interaction effect was found for anthropomorphism and discomfort, indicating these scores varied less between robot conditions in the OA group.
Limitations and Future Work
- Limitations:
- The study used a simulated robot (video) due to the COVID-19 pandemic, which might not fully replicate the experience of interacting with a physical robot.
- A significant proportion of participants did not notice the difference between robot conditions, which complicates the interpretation of results for that subset.
- The "moving robot" used head movements to simulate gaze, which is a simplification of human eye gaze.
- Future Work:
- Future research should incorporate more conditions, such as a human face or non-social signaling, to better understand the differences in perception between human and robot gaze cues.
- Investigating the impact of more subtle differences in deictic gaze and exploring how these cues can be adapted for different age groups is crucial for designing effective HRI.
- Further exploration into why a significant portion of participants, particularly OA, do not notice gaze differences is warranted.
Summary updated