What are you trying to do?
This project is developing an AI generated, student success textbook delivered on a web platform. The objective of this project is to demonstrate the following:
- AI’s human-guided ability to create a culturally responsive and culturally relevant textbook
- AI’s human guided ability to mash up static textbook sections on which broad learning objectives are based, and personalized textbook sections on which contextualized learning objectives are based
- AI’s human-guided ability to power personalized textbook content based on one or more user defined inputs (i.e. age, race, geography, grade level, sexual orientation, etc.)
How is it done today, and what are the limits of current practice?
Modern textbooks used in the United States are written almost exclusively from a Western, dominant society perspective and rarely provide content that meets the needs of learners from diverse backgrounds. Given the resources currently available, writing personalized learning material that spans the needs of diverse learners is impossible in traditional practice. Artificial intelligence provides a unique opportunity to meet students where they are, both literally and figuratively.
What is new in your approach and why do you think it will be successful?
Responsive texts and textbooks could reshape an entire generation’s relationship to education, and maybe help close some of the stubborn opportunity gaps educators are struggling to close.
Who cares? If you are successful, what difference will it make?
This is something everyone, everywhere cares about if they feel a vested interest in the broad education of their community members. In every community in the United States (and perhaps the world) educational texts and textbooks are inaccessible to many learners.
What are the risks?
There is risk that GPT-3 will introduce bias into educational texts in a way that is harmful, but this is a reason for (not against) the research and study of AI and its ability to solve human sociological and structural problems. We need to train AI to be better for the sake of learners everywhere.
How long will it take?
Estimated time to proof of concept is one month.
Estimated time to a minimal viable product is three months.
What are the mid-term and final “exams” to check for success?