Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Humans prefer interacting with slow, less realistic butterfly simulations

Paige L. Reiter, Talia Y. Moore

TL;DR

This work investigates how to optimize butterfly-like robots for safe and engaging human interaction by examining how flight pattern cues influence willingness to interact. Leveraging literature-informed butterfly biomechanics, the authors create eight simulated butterfly videos in Blender/Unity that vary wingbeat frequency, flap-to-glide ratio, and wingbeat pattern, and assess user responses via a Qualtrics survey with $131$ participants. Results show a robust preference for slower wingbeat motion around $3$ Hz, with higher willingness associated with gliding-dominant trajectories and migratory-like flight, while wingbeat pattern plays a minor role; demographic factors significantly modulate responses. The study demonstrates that iterative, simulation-driven design—coupled with user feedback—can streamline the development of zoomorphic robots by identifying functional requirements prior to fabrication, and it situates these findings within the broader context of media depictions and the Uncanny Valley for invertebrates.

Abstract

How should zoomorphic, or bio-inspired, robots indicate to humans that interactions will be safe and fun? Here, a survey is used to measure how human willingness to interact with a simulated butterfly robot is affected by different flight patterns. Flapping frequency, flap to glide ratio, and flapping pattern were independently varied based on a literature review of butterfly and moth flight. Human willingness to interact with these simulations and demographic information were self-reported via an online survey. Low flapping frequency and greater proportion of gliding were preferred, and prior experience with butterflies strongly predicted greater interaction willingness. The preferred flight parameters correspond to migrating butterfly flight patterns that are rarely directly observed by humans and do not correspond to the species that inspired the wing shape of the robot model. The most realistic butterfly simulations were among the least preferred. An analysis of animated butterflies in popular media revealed a convergence on slower, less realistic flight parameters. This iterative and interactive artistic process provides a model for determining human preferences and identifying functional requirements of robots for human interaction. Thus, the robotic design process can be streamlined by leveraging animated models and surveys prior to construction.

Humans prefer interacting with slow, less realistic butterfly simulations

TL;DR

This work investigates how to optimize butterfly-like robots for safe and engaging human interaction by examining how flight pattern cues influence willingness to interact. Leveraging literature-informed butterfly biomechanics, the authors create eight simulated butterfly videos in Blender/Unity that vary wingbeat frequency, flap-to-glide ratio, and wingbeat pattern, and assess user responses via a Qualtrics survey with participants. Results show a robust preference for slower wingbeat motion around Hz, with higher willingness associated with gliding-dominant trajectories and migratory-like flight, while wingbeat pattern plays a minor role; demographic factors significantly modulate responses. The study demonstrates that iterative, simulation-driven design—coupled with user feedback—can streamline the development of zoomorphic robots by identifying functional requirements prior to fabrication, and it situates these findings within the broader context of media depictions and the Uncanny Valley for invertebrates.

Abstract

How should zoomorphic, or bio-inspired, robots indicate to humans that interactions will be safe and fun? Here, a survey is used to measure how human willingness to interact with a simulated butterfly robot is affected by different flight patterns. Flapping frequency, flap to glide ratio, and flapping pattern were independently varied based on a literature review of butterfly and moth flight. Human willingness to interact with these simulations and demographic information were self-reported via an online survey. Low flapping frequency and greater proportion of gliding were preferred, and prior experience with butterflies strongly predicted greater interaction willingness. The preferred flight parameters correspond to migrating butterfly flight patterns that are rarely directly observed by humans and do not correspond to the species that inspired the wing shape of the robot model. The most realistic butterfly simulations were among the least preferred. An analysis of animated butterflies in popular media revealed a convergence on slower, less realistic flight parameters. This iterative and interactive artistic process provides a model for determining human preferences and identifying functional requirements of robots for human interaction. Thus, the robotic design process can be streamlined by leveraging animated models and surveys prior to construction.
Paper Structure (16 sections, 5 figures, 3 tables)

This paper contains 16 sections, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Biological Template
  3. Simulated Model
  4. Appearance of butterfly model
  5. Simulation of butterfly flight
  6. Simulated parameters
  7. Experimental Methods
  8. Experimental Results and Discussion
  9. Human preferences
  10. Demographic Factors
  11. Open responses
  12. Depictions of butterflies in popular media
  13. Data collection
  14. Trends in animated butterfly motion
  15. Conclusion
  16. ...and 1 more sections

Figures (5)

  • Figure 1: Comparison of flight characteristics for real butterflies (purple and images), animations of butterflies in popular media (orange), and butterfly robots simulated for this study (green). The letters indicate the popular media source, corresponding to rows in Tab. \ref{['tab:popref']}. Note that in a) the character analyzed is a moth; all other examples are butterflies. The numbers indicate the video simulation, corresponding to trajectories in Fig. \ref{['fig:unityscreenshot']}. The preferred video of survey participants is indicated with a black arrow. Note that for simulated dataset, each specified combination of wingbeat frequency and proportion of gliding includes one aperiodic and one periodic datapoint. Images of the biological butterflies obtained from Yale Peabody Museum (Danaus plexippus), Muséum d'Histoire Naturelle de Toulouse (Pieris rapae), and London (Manduca sexta). The shape of each plotted data point indicates periodicity of the wingbeats.
  • Figure 2: A) Female Pieris rapae specimen (Photo by Sarefo (Wikimedia Commons) licensed under CC BY-SA 3.0). B) Simple butterfly model in the Blender environment, with the armature used for animations outlined in blue.
  • Figure 3: A screenshot from a butterfly simulation that was presented to participants, with the trajectories of all butterfly trials overlaid.
  • Figure 4: Survey participant willingness to interact varied with respect to different flight parameters.
  • Figure 5: Comparison of Interaction Willingness (IW) with respect to demographic factors. Asterisks indicate significance level $p < 0.05$. A period indicates significance level $0.05 < p < 0.1$. Academic affiliation and previous experience with butterflies significantly predicted mean IW scores for each participant. The range of IW scores for a participant decreased with age, but this trend did not explain the majority of the variation in the data. The y-axes were truncated at the maximum available score of seven.