Scholarlink

<p>This study investigated how perceived authorship—human therapist versus AI (ChatGPT 4)—influences evaluations of empathy, professionalism, and factual correctness in therapeutic conversations. A total of 84 graduate-level clinical psychology students were exposed to six conversation excerpts (three AI-generated, three human-generated) across three phases: (1) <i>Masked</i>, where participants did not know the source of each excerpt; (2) <i>Deceived</i>, where the origins were deliberately misrepresented; and (3) <i>Source Revelation</i>, where the true source was disclosed. Participants rated each excerpt on empathy, professionalism, and factual correctness using 5-point Likert scales. MANOVA and paired sample t-test were conducted to compare the Source effect, Phase effect, and the interaction between the two. A significant interaction effect between Source and Phase was observed. Also, AI-generated excerpts received significantly higher ratings than the real human transcripts on all three dimensions in the Masked and Deceived phases. However, in the Source Revelation phase, ratings for the human therapist transcripts increased substantially, surpassing those of AI-generated content, highlighting how transparency and attribution profoundly shape judgments of conversational quality.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Perceived Authorship and Conversational Evaluations: A Study on AI-Generated vs. Human Therapist Dialogue

  • 期刊论文

  • 发表日期：

    2025年7月12日

查看全文
  • Samridhi Pareek,
  • Gagan Jain

摘要

This study investigated how perceived authorship—human therapist versus AI (ChatGPT 4)—influences evaluations of empathy, professionalism, and factual correctness in therapeutic conversations. A total of 84 graduate-level clinical psychology students were exposed to six conversation excerpts (three AI-generated, three human-generated) across three phases: (1) Masked, where participants did not know the source of each excerpt; (2) Deceived, where the origins were deliberately misrepresented; and (3) Source Revelation, where the true source was disclosed. Participants rated each excerpt on empathy, professionalism, and factual correctness using 5-point Likert scales. MANOVA and paired sample t-test were conducted to compare the Source effect, Phase effect, and the interaction between the two. A significant interaction effect between Source and Phase was observed. Also, AI-generated excerpts received significantly higher ratings than the real human transcripts on all three dimensions in the Masked and Deceived phases. However, in the Source Revelation phase, ratings for the human therapist transcripts increased substantially, surpassing those of AI-generated content, highlighting how transparency and attribution profoundly shape judgments of conversational quality.