Evaluating the Effectiveness of AI-Generated Personalized Learning Content for Improving Engagement and Learning Outcomes

AI Graphic
Date Completed
Abstract

In this research project, the aim is to evaluate the effectiveness of learning materials that, using Generative AI, are personalized to students' individual interests, learning styles, and skill levels. As an initial testbed that lends itself to doing a controlled experiment, they plan to research this concept in the domain of vocabulary learning. They will develop a custom online vocabulary learning app that allows students to enter their interests, learning styles, and skill level. The app will use this information in conjunction with generative artificial intelligence in the form of large-language-models (e.g. Open AI’s GPT3 and GP4) and text-to-image models (e.g. DallE) to automate the generation of learning examples (i.e. text and/or images) that are tailored to students’ individual preferences, skill level, and interests. The app will be developed using an iterative prototyping process, in order to uncover how textual prompts to these APIs should be designed to generate the desired outputs. Great care will be taken in ensuring the safety and guardrails of the application, to make sure the customized materials do not include sensitive, age-inappropriate, or controversial contents.

Researchers
Pattie Maes, Joanne Leong
Lab Name
Media Lab