Scholarlink

<p>The integration of artificial intelligence (AI) into qualitative research promises to augment traditional methodologies, yet its comparative analytical power remains underexplored.&#xa0;This exploratory proof-of-concept study&#xa0;employs a triangulation approach to compare thematic analyses of semi-structured interviews with pathological gamblers in Uruguay—one conducted by a human researcher and another assisted by a generative AI model (GPT-4). Our findings illustrate convergence in identifying core themes, such as the illusion of control and the role of gambling in emotional regulation. However, significant divergences emerged in interpretive depth: the human analysis provided critical, contextualized insights into sociocultural and existential dimensions, while the AI offered systematic efficiency in pattern recognition and thematic synthesis. We demonstrate that a hybrid methodology does not produce contradictory results but rather generates complementary knowledge. Within the limits of this proof-of-concept, AI appears as a potentially auxiliary tool that can enhance the rigor and scope of qualitative analysis in mental health and beyond. A brief but important clarification: this is not a clinical or epidemiological study. It is a methodological proof-of-concept. We use the interview on pathological gambling as a substantive case to illustrate how human and AI-assisted qualitative analysis can be systematically compared. Due to the small sample size (N = 7) and the lack of reproducibility assessment (single GPT‑4 run), findings are preliminary and should be viewed as a procedural illustration rather than reliable evidence of human‑AI agreement. Replication with multiple runs and more diverse samples is required before any generalized claims can be made.</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

Human–AI triangulation in qualitative research: a methodological proof-of-concept using interview data on pathological gambling

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月28日

查看全文
  • Mayda Portela

摘要

The integration of artificial intelligence (AI) into qualitative research promises to augment traditional methodologies, yet its comparative analytical power remains underexplored. This exploratory proof-of-concept study employs a triangulation approach to compare thematic analyses of semi-structured interviews with pathological gamblers in Uruguay—one conducted by a human researcher and another assisted by a generative AI model (GPT-4). Our findings illustrate convergence in identifying core themes, such as the illusion of control and the role of gambling in emotional regulation. However, significant divergences emerged in interpretive depth: the human analysis provided critical, contextualized insights into sociocultural and existential dimensions, while the AI offered systematic efficiency in pattern recognition and thematic synthesis. We demonstrate that a hybrid methodology does not produce contradictory results but rather generates complementary knowledge. Within the limits of this proof-of-concept, AI appears as a potentially auxiliary tool that can enhance the rigor and scope of qualitative analysis in mental health and beyond. A brief but important clarification: this is not a clinical or epidemiological study. It is a methodological proof-of-concept. We use the interview on pathological gambling as a substantive case to illustrate how human and AI-assisted qualitative analysis can be systematically compared. Due to the small sample size (N = 7) and the lack of reproducibility assessment (single GPT‑4 run), findings are preliminary and should be viewed as a procedural illustration rather than reliable evidence of human‑AI agreement. Replication with multiple runs and more diverse samples is required before any generalized claims can be made.