Exploring Human Perceptions of AI Responses: Insights from a Mixed-Methods Study on Risk Mitigation in Generative Models

2025-12-01

AI Summary

Generative AI frequently produces hallucinations and harmful content, yet human awareness of existing risk-mitigation strategies remains limited. Method: We conduct a mixed-methods within-subject experiment, integrating quantitative scoring with qualitative feedback to multidimensionally assess AI responses across factual accuracy, fairness, harmlessness, and relevance. We further propose novel evaluation metrics that jointly preserve semantic fidelity and capture contextual sensitivity. Results: Native language background and AI domain expertise significantly influence human judgments, underscoring heightened sensitivity to linguistic nuance. Our proposed metrics effectively bridge the gap between human evaluations and conventional automated metrics—demonstrating strong alignment with human judgment while retaining computational tractability. This work establishes a reproducible, interpretable, and human-centered framework for evaluating AI response quality in human-AI collaborative settings.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

Abstract

With the rapid uptake of generative AI, investigating human perceptions of generated responses has become crucial. A major challenge is their `aptitude' for hallucinating and generating harmful contents. Despite major efforts for implementing guardrails, human perceptions of these mitigation strategies are largely unknown. We conducted a mixed-method experiment for evaluating the responses of a mitigation strategy across multiple-dimensions: faithfulness, fairness, harm-removal capacity, and relevance. In a within-subject study design, 57 participants assessed the responses under two conditions: harmful response plus its mitigation and solely mitigated response. Results revealed that participants' native language, AI work experience, and annotation familiarity significantly influenced evaluations. Participants showed high sensitivity to linguistic and contextual attributes, penalizing minor grammar errors while rewarding preserved semantic contexts. This contrasts with how language is often treated in the quantitative evaluation of LLMs. We also introduced new metrics for training and evaluating mitigation strategies and insights for human-AI evaluation studies.

Problem

• Investigates human perceptions of AI risk mitigation strategies
• Evaluates mitigation effectiveness across faithfulness, fairness, and harm removal
• Examines how user background influences assessment of AI responses

Innovation

• Mixed-method experiment evaluating mitigation across multiple dimensions
• Within-subject study comparing harmful and mitigated AI responses
• Introduced new metrics for training and evaluating mitigation strategies