Voice-Based Chatbots for English Speaking Practice in Multilingual Low-Resource Indian Schools: A Multi-Stakeholder Study

TL;DR

The paper addresses the scarcity of spoken English practice in multilingual, low-resource Indian schools by deploying a voice-based chatbot (ChatFriend) in four Delhi schools and studying its reception, engagement, and design implications over six days. Using a qualitative, interpretive approach with Day 1 real-time observations and Days 2–6 extended use, it collects perspectives from students, teachers, and principals to identify design tensions between empathetic, open-ended dialogue and curriculum-aligned scaffolding, as well as fragility of confidence under technical friction. Key contributions include a set of actionable design guidelines and a field-implementation checklist for resource-constrained multilingual settings, evidence on student affective trajectories, and a proposed pathway for integrating speaking-lab style modules into national platforms with lightweight analytics. The findings highlight the practical significance of mobile-first, low-friction interfaces, intelligible speech output, and teacher-facing analytics to support scalable, socially sustainable EdTech deployments in low-resource schools. These insights offer a roadmap for deploying AI-based educational technologies that balance learner motivation, administrator priorities, and infrastructural realities.

Abstract

Spoken English proficiency is a powerful driver of economic mobility for low-income Indian youth, yet opportunities for spoken practice remain scarce in schools. We investigate the deployment of a voice-based chatbot for English conversation practice across four low-resource schools in Delhi. Through a six-day field study combining observations and interviews, we captured the perspectives of students, teachers, and principals. Findings confirm high demand across all groups, with notable gains in student speaking confidence. Our multi-stakeholder analysis surfaced a tension in long-term adoption vision: students favored open-ended conversational practice, while administrators emphasized curriculum-aligned assessment. We offer design recommendations for voice-enabled chatbots in low-resource multilingual contexts, highlighting the need for more intelligible speech output for non-native learners, one-tap interactions with simplified interfaces, and actionable analytics for educators. Beyond language learning, our findings inform the co-design of future AI-based educational technologies that are socially sustainable within the complex ecosystem of low-resource schools.

Figures (2)

• Figure 1: ChatFriend Interface--- Student participants interacted with the ChatFriend web application on research team-provided Android tablets. The left panel displays available conversation topics (e.g., "Best Friend", "Favorite Sports"), and students select one to begin a session. ChatFriend initiated the conversation, and students responded by tapping/pressing the on-screen microphone button to record audio. The system replied with synthesized speech, while text transcripts of both student and system turns appeared in the scrolling chat history on screen.
• Figure 2: Student-by-day Affective Trajectory Heatmap