Skip to main content Skip to navigation

WaLLM -- Insights from an LLM-Powered Chatbot deployment via WhatsApp

Project Overview

The document explores the use of WaLLM, a WhatsApp-based AI chatbot that employs generative AI to improve educational access, especially in developing regions. It emphasizes key applications such as engagement tools that include daily top questions and a leaderboard to foster user interaction. The findings indicate that users exhibit a strong interest in health and well-being topics, which reflects their needs and behaviors. While the chatbot demonstrates promise in addressing the digital divide and enhancing information accessibility, it also encounters challenges, particularly concerning usability and user trust in AI technologies. Overall, the document highlights both the potential benefits and obstacles of integrating generative AI in educational contexts, showcasing how such innovations can facilitate learning and information dissemination.

Key Applications

WaLLM - WhatsApp-based AI chatbot

Context: Educational tool for users in developing regions (Pakistan, Sudan, USA diaspora)

Implementation: Operational for over 6 months, utilizing multiple LLMs via WhatsApp.

Outcomes: Amassed over 14.7K queries from ~100 users; users found it reliable for health and well-being information.

Challenges: Digital divide, trust issues in AI responses, and high costs of technology.

Implementation Barriers

Access and Usability

High costs of using advanced AI models and digital illiteracy hinder access.

Proposed Solutions: Provide free access to AI models via familiar platforms like WhatsApp.

Trust in AI

Users may rely on inaccurate information from AI due to lack of verification.

Proposed Solutions: Enhance user interface design to encourage critical engagement and transparency.

Project Team

Hiba Eltigani

Researcher

Rukhshan Haroon

Researcher

Asli Kocak

Researcher

Abdullah Bin Faisal

Researcher

Noah Martin

Researcher

Fahad Dogar

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Hiba Eltigani, Rukhshan Haroon, Asli Kocak, Abdullah Bin Faisal, Noah Martin, Fahad Dogar

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