[This project was part of a Ph.D. thesis at XLRI (A leading business school in India) where I assisted a doctorate candidate during my undergraduation]
Using Artificial Intelligence to help build engaging conversations in a community
Problem
We are living in a fragmented society and often feel unheard and detached from the community we live in.
Many students in an academic setting do not speak when they have an uncommon opinion or fall into the minority group
Solution
Providing real-time feedback on utterances using an AI system that will push students to be indulged in more engaging and attached conversations and move away from detached ones.
Implementation
Data Description (A sample data is attached)
Utterances: Data collected from 25 students on what they say when they are in the classroom, when they are leaving, etc.
Themes: Which part of the conversation leads to a particular trait?
Traits: Behavioral traits representing attach, engage and detach
[Attach - Implantation, Flexibility, Power Equity, vitality, Dependence; Engage - Confidence, Competence, Relational stability, competence-based competition; Detach - Individuation, Interdependence, Cosmology, World view, Political Activism]
Data Collection
In a classroom of students, utterance data is collected. For example: what do they say when they enter the classroom, what do they say when they are leaving, before exams, and similar scenarios
Data Annotation
This data is then annotated reflecting the behavioral traits of each student as well as the group
Modeling & Evaluation
A machine learning classification model is trained to predict behavioral traits and evaluated based on the F1 score
Future Work
Insights from the data could be used to build an automated system based on real-time behavioral feedback
It can move group/community discussions in a positive direction (i.e., towards attach and engage instead of detach)
My Contribution(This project was a part of the thesis of a doctoral student I assisted back in my undergrad)
Bits and pieces of conceptualization, Assisting in data collection and annotation tasks, worked on ML model
Uniqueness
We often work with scrapped data from social media, this was my first exposure to working on real-life human behavior data collection