Jennifer L. Madiedo
MIT EECS | Fairbairn Undergraduate Research and Innovation Scholar
Persuasion Through Character Relation in Story-Processing Systems
2018–2019
EECS
- Natural Language and Speech Processing
Patrick H. Winston
Genesis is an AI system that understands, tells, and composes stories using commonsense rules and higher-level concept patterns. One area where Genesis still needs development is persuasion. Currently, it can persuade readers towards a certain point of view via a technique of information removal. The goal of my project is to achieve persuasion through character relation by synthesizing new text to be inserted into the story. Genesis will be able to read a story, and given a specific character, relate it to a similar character from a past story by adding new text to the story. This relation can cause a transfer of qualities between the two characters, impacting readers’ perception and ideas of the current character. The success of this project represents a higher level of understanding achieved by Genesis and another step closer to teaching computers to think as humans do.
I am participating in SuperUROP because my past research with the Personal Robotics Group in the Media Lab piqued my interest in how children learn new ideas, and how we can use those same methods to teach programs to learn on their own. I aim to get experience on a long-term research project in my field of interest and learn more about the applications and ethics of intelligent programs.