Exploring the Effects of Support Type and Task Difficulty for Virtual Assistants as Social Companions

Paper Title

Exploring the Effects of Support Type and Task Difficulty for Virtual Assistants as Social Companions

Publication Info

• Topic area: Virtual assistants as social companions in varying task contexts
• Keywords: Virtual assistants, emotional support, informational support, task difficulty, embodiment, mixed reality, user experience, social connection, cognitive tasks, physical tasks

Background and Problem

• Problem / challenge: Existing research on virtual assistants (VAs) lacks clear guidelines on how support types (emotional vs. informational) should adapt to varying task difficulties, task types, and embodiment styles.
• Significance: Understanding these dynamics is crucial for designing VAs that effectively serve as social companions, enhancing user experience and task performance.
• Motivation and related work: Prior studies have explored emotional and informational support individually but have not examined their combined effects across different task contexts. Embodiment has been studied in isolation, with limited investigation into its interaction with support type and task difficulty.

Solution

• Proposed approach: Two user studies investigating the effects of emotional and informational support across task difficulty (easy vs. hard), task type (cognitive vs. physical), and embodiment (non-embodied vs. embodied).
• Novelty:
  1. Analysis of interaction effects between support type and task difficulty.
  2. Comparison of support effectiveness across cognitive and physical tasks.
  3. Examination of embodiment's role in moderating support effects.
• Procedure and key techniques:
  • Study 1: Budget allocation task (cognitive), 2 × 2 × 2 factorial design manipulating support type, task difficulty, and embodiment.
  • Study 2: Exercise task (physical), same factorial design.
  • Measurements: Subjective ratings (e.g., task load, social connection), objective metrics (e.g., accuracy, completion rate), and user interviews.

Results

• Concrete findings:
  • Emotional support was preferred for easy tasks, fostering companionship and motivation.
  • Informational support was more effective for hard tasks, particularly in cognitive settings, improving task performance and reducing cognitive load.
  • Embodiment had no significant influence on user perceptions or task outcomes.
• Advantage over baselines: Demonstrated nuanced interaction effects between support type and task difficulty, providing evidence for situationally adaptive VA design.
• Experiments / evaluation:
  • Study 1: Cognitive task with 32 participants; measured task performance, social connection, and user preferences.
  • Study 2: Physical task with 24 participants; measured completion rate, engagement, and attentional behavior.
  • Statistical analysis: Mixed ANOVA with Aligned Rank Transform for non-parametric data.
• Limitations and future work:
  • Limited generalizability due to short-term, simplified tasks in laboratory settings.
  • Fixed embodiment design; future studies should explore diverse visual representations.
  • One-way interaction design; two-way interaction effects remain unexplored.
  • Support limited to verbal communication; multi-modal support should be investigated.

Summary

This paper explores how virtual assistants can adapt their support strategies to varying task contexts. Emotional support was found to enhance user experience in easy tasks, while informational support improved performance in hard tasks, particularly in cognitive settings. Embodiment did not significantly affect user perceptions or outcomes. These findings inform design guidelines for adaptive VAs that dynamically tailor their support to users’ situational needs. Future research should address long-term interactions, diverse embodiment styles, and multi-modal support to further refine VA effectiveness as social companions.