Skip to main content Skip to navigation

Malinowski in the Age of AI: Can large language models create a text game based on an anthropological classic?

Project Overview

The document explores the application of generative AI, particularly Large Language Models (LLMs) like ChatGPT and GPT-4, in enhancing educational experiences through the development of text-based games focused on anthropological themes. By utilizing interactive fiction inspired by Bronislaw Malinowski's 'Argonauts of the Western Pacific', the study illustrates how generative AI can significantly boost student engagement and create unique learning opportunities. However, it also addresses the challenges associated with this technology, notably the AI's limitations in delivering precise and comprehensive anthropological insights. Overall, while generative AI shows promise in making learning more interactive and enjoyable, it also necessitates careful consideration of its accuracy and depth of knowledge in educational contexts.

Key Applications

Text-based games based on anthropological classics

Context: Classroom environment for anthropology students

Implementation: Developed three prototypes of text-based games using GPT-3.5, incorporating iterative design and playtesting with senior anthropologists.

Outcomes: Promising engagement from participants, potential for teaching anthropological concepts, and fostering media literacy regarding misinformation.

Challenges: Difficulty in providing in-depth thematic understanding, susceptibility to misinformation, tendency for monotonic responses during extended play, and lack of detailed biographical information.

Implementation Barriers

Technical Limitations

Models struggle to provide in-depth understanding of anthropological contexts and are susceptible to misinformation.

Proposed Solutions: Incorporate a Retrieval-Augmented Generation (RAG) framework to enhance the knowledge base of AI models.

User Engagement

Monotonic responses after prolonged play can diminish user engagement.

Proposed Solutions: Introduce a turn tracker to diversify prompts and enhance gameplay experience.

Project Team

Michael Peter Hoffmann

Researcher

Jan Fillies

Researcher

Adrian Paschke

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Michael Peter Hoffmann, Jan Fillies, Adrian Paschke

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies