Using AI Personas to Practice Correcting Misconceptions About Human Biological Variation

Dr. Hernandez, a postdoctoral scholar in Anthropology, received a Dartmouth Teaching with GenAI grant to design a structured assignment in which students engage directly with a GenAI model to correct scientific misconceptions about human biological variation.

The assignment was developed in collaboration with DCAL at the GenAI Teaching Institute. Students used chat.dartmouth.edu (Dartmouth Chat) to complete the work.

Learning Goals

  • Develop knowledge of biological and cultural factors shaping human biological variation
  • Build effective scientific communication skills
  • Critically evaluate and challenge harmful social discourse about human biological variation

The Assignment

Students created a GenAI persona (via a custom system prompt) that held a specific misconception about human biological variation. They then engaged with the persona in conversation, correcting the misconception using course readings and evidence. Students submitted their system prompt, full transcript, and a written reflection.

Class time was dedicated to scaffolding: students learned to write system prompts, tested multiple models, and selected the best fit before completing the week-long assignment independently.

What Worked

  • Simulating real dialogue with someone holding a misconception—safely and without discomfort
  • Pushed students to demonstrate mastery by constructing persuasive, evidence-based arguments
  • Introduced students to a genuinely new use of AI (system prompts, persona design)
  • Clear AI use policies reduced ambiguity and student anxiety

Student Survey Results

21 of 22 enrolled students responded (95% response rate). Items rated on a 1-5 Likert scale.

hernandez_anthroaiassignment_studentresponses.png

Student responses to eight survey questions. Top rated responses were Guidelines for allowed use were clear (4.86), resources to use GenAI were available (4.86), and used GenAI in a new way (4.71).

Student Voices

"It allowed me to practice my dialogue skills without feeling unsafe—there was no physical threat."

"It showed me a completely new way to use or think about AI. I feel like it was a valuable assignment overall."

Faculty Reflections: Building on What Worked

  • Increase AI persona complexity to better challenge advanced students
  • Offer more explicit onboarding to Dartmouth Chat for students unfamiliar with LLM differences
  • Consider multi-model comparisons as an extension activity
  • Reinforce AI literacy (model selection, web retrieval) as foundational scaffolding