Reproduce “The Placebo Effect of Artificial Intelligence in Human-Computer Interaction” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Thomas Kosch, Robin Welsch, Lewis Chuang, Albrecht Schmidt

Research from Utrecht, LMU Munich, and Chemnitz University of Technology demonstrates that system descriptions can induce a placebo effect in human-computer interaction, leading to increased user expectations and subjective performance ratings for adaptive AI systems even when no actual AI functionality is present. This effect highlights a potential bias in HCI evaluation methodologies, as objective task performance was not directly impacted by the placebo.

Key insights

  • System descriptions effectively manipulated user expectations regarding AI system adaptiveness and anticipated performance.
  • Induced expectations persisted as elevated subjective performance ratings after interaction, even without actual AI functionality or impact on objective task performance.
  • A positive correlation was observed between individual performance expectations and objective performance, particularly for the non-transparent AI placebo condition, suggesting that personal belief can subtly influence individual outcomes.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.