Author: rishita

Posted on

Social Synchrony in Learning Networks

Understanding how learner interactions shape collective learning outcomes

Learning is often viewed as an individual process, where success depends primarily on the learner and the content they consume. However, decades of research suggest that learning is also deeply social. Learners influence one another through discussion, collaboration, feedback, and shared experiences.

This project explores how interactions between learners shape collective learning outcomes. By combining ideas from learning sciences and network science, it investigates how learning communities evolve and how patterns of coordination emerge through repeated interaction.

Learning as a Network

The project models a learning community as a network, where learners are represented as nodes and interactions between learners form the connections.

This perspective shifts the focus from individual learning trajectories to the behavior of the community as a whole. It allows us to study how information flows, how learners influence one another, and how collective outcomes emerge over time.

Social Synchrony

One of the central ideas explored in this work is social synchrony—the tendency of learners to gradually align in capability and progress through interaction.

As learners exchange knowledge and experiences, differences within the community can reduce, resulting in more coordinated learning behavior and stronger collective outcomes.

The concept draws inspiration from foundational work in social learning, sociocultural learning, communities of practice, and networked learning.

Key Findings

The study demonstrates that learner interactions play an important role in shaping collective outcomes. Across simulated learning communities, repeated interaction was associated with the emergence of synchrony and reduced capability disparities among learners.

These findings suggest that understanding learner-to-learner relationships may be as important as understanding learner-to-content relationships when designing future learning systems.

Research Outputs

Publication

Modeling Outcomes-led Learner Behavior and Emergent Social Synchrony Ashashree Sarma, Sushree Behera, Srinath Srinivasa, and Prasad Ram. Proceedings of the Twelfth ACM Conference on Learning @ Scale (L@S 2025), Palermo, Italy. DOI: 10.1145/3698205.3733959

https://dl.acm.org/doi/abs/10.1145/3698205.3733959

Poster Presentations

  • RISE 2025
  • L@S & EDM 2025
  • WebSciX 2025

Media Coverage

Google Maps for Learning: IIIT-B Researchers Develop AI Navigator for Personalised STEM Learning https://www.thehindu.com/news/national/karnataka/google-maps-for-learning-iiit-b-researchers-develop-ai-navigator-for-personalised-stem-learning/article70089077.ece

Featured in The Hindu.

Foundations

This work draws inspiration from:

  • Albert Bandura’s Social Learning Theory
  • Lev Vygotsky’s Sociocultural Theory
  • Etienne Wenger’s Communities of Practice
  • George Siemens’ Connectivism
  • Complex Systems and Network Science

People

Project Lead

Ashashree Sarma
 Research Scholar, IIIT Bangalore
Research Profile: https://wsl.iiitb.ac.in/ashashree-sarma/
 LinkedIn: https://www.linkedin.com/in/ashashree17321/

Supervisors

Prof. Srinath Srinivasa
 Professor, IIIT Bangalore
Profile: https://wsl.iiitb.ac.in/srinath-srinivasa/

Prof. Sushree Sangeeta Behera
 Faculty, IIIT Bangalore
Profile: https://wsl.iiitb.ac.in/sushree-behera/

Posted on

Active Learning Map

Traditional education often treats learning as a straight line—a rigid checklist of lectures and exams that assumes every student progresses at the exact same pace. But in reality, true human learning is non-linear, unpredictable, and deeply personal, filled with sudden cognitive leaps and variable moments of assimilation. To capture this dynamic reality, my research introduces a data-driven framework centered around the Active Learning Map.

Instead of tracking simple grades, this system maps a student’s evolving knowledge state as a continuous trajectory through a high-dimensional vector space called Polylines. Using a radial mapping algorithm, these complex mathematical dimensions are compressed into an intuitive, two-dimensional coordinate system where each fundamental course topic radiates as its own axis from a central origin.

By integrating a Conditional Generative Adversarial Network (cGAN) to simulate how a student’s knowledge profile changes when they interact with a resource, we’ve built a highly accurate transition state model. Navigating this space is an intelligent, model-free Deep Q-Network (DQN) recommendation engine. Operating much like a GPS for education, the AI agent treats curriculum routing as a shortest-path problem, dynamically discovering personalized, accelerated learning trajectories to close the knowledge gap and guide students toward mastery in the fewest possible steps. By turning abstract cognitive growth into a visually interpretable and traversable landscape, the Active Learning Map lays the groundwork for a truly adaptive, outcome-centric classroom.

Posted on

Rishita Patel

Rishita Patel is an M.S. Research Scholar at the Web Science Lab (WSL), IIIT-Bangalore, under the guidance of Prof. Sushree Behera and Prof. Srinath Srinivasa. Her current research, titled “Multi-modal Content Generation for Navigated Learning”, investigates the applicability of diffusion-based text-to-image generation models in the educational domain—an area where factual accuracy, pedagogical alignment, and contextual relevance are paramount. She is working on fine-tuning state-of-the-art diffusion models , aiming to bridge the gap between generative AI’s creative power and the stringent demands of educational content.

Her research interests lie at the intersection of multimodal AI, semantic reasoning, and responsible content generation. Her work explores how models trained for general-purpose creativity can be adapted to serve domain-specific, high-precision use cases like STEAM education.

📎LinkedIn Profile