Blinded But Biased: Students Prefer Chatbot Until They Know It Is One

Area of Science:

• Nursing Education
• Artificial Intelligence in Education
• Statistical Support

Background:

• The integration of artificial intelligence (AI) into educational settings necessitates understanding student perceptions of AI-generated academic support.
• This pilot study focuses on evaluating Doctor of Nursing Practice (DNP) students' views on statistical assistance from various sources.

Purpose of the Study:

• To investigate how DNP students perceive and rate statistical help provided by different sources, including a large language model (LLM) chatbot.
• To explore potential biases in student evaluations of AI-generated academic support.

Main Methods:

• Seven DNP students submitted capstone project-related statistical questions.
• Responses were generated by a custom-trained LLM chatbot, a graduate assistant, and a professor, presented blindly.
• Students rated responses on helpfulness, satisfaction, and likelihood of use, and attempted to identify the AI source.

Main Results:

• The LLM chatbot received the highest average ratings for helpfulness and satisfaction when responses were evaluated blindly.
• Students consistently assigned lower ratings to responses they believed were AI-generated, indicating a potential bias against AI.

Conclusions:

• DNP students preferred AI chatbot responses for statistical assistance when the source was unknown.
• A bias against AI was observed when students suspected the source, which could impede the adoption of AI tools in academic support settings.
• Further research is needed to understand and address AI bias in educational contexts.