ARTFEED — Contemporary Art Intelligence

Study Shows Preference for Human-Like Agents in Mobile Serious Game

other · 2026-05-28

A study with 90 participants compared a highly human-like spoken embodied conversational agent (ECA) against a low human-like text-based agent in a mobile game about pre-decimal UK currency. The game featured two agent roles: an Instructor (Alex) and a Shopkeeper/Collaborator. Users interacted via voice and mouse. Quantitative data from usability questionnaires (CCIR MINERVA) and the Agent Persona Instrument were analyzed using paired t-test, repeated measures ANOVA, and multiple linear regression. Results showed a statistically significant preference for the highly human-like agent version with a large effect size. Qualitative findings further support this preference.

Key facts

  • Study involved 90 participants
  • Within-subjects design
  • Compared highly human-like spoken ECA vs low human-like text-based agent
  • Game developed in Unity about pre-decimal UK currency
  • Two agent roles: Instructor (Alex) and Shopkeeper/Collaborator
  • Data collected via CCIR MINERVA usability questionnaire and Agent Persona Instrument
  • Statistical analysis: paired t-test, repeated measures ANOVA, multiple linear regression
  • Significant preference for highly human-like agents with large effect size

Entities

Sources