Hector Research Institute of Education Sciences and Psychology

INFER

Intelligent Feedback System for Observing Videotaped Classroom Situations

Symbolbild: Videoaufnahme einer Unterrichtsstunde

Our Goal

The INFER project develops and tests intelligent feedback systems for teaching practice for teacher training students and teachers. Using natural language processing (NLP) and machine learning (ML) methods, INFER automatically analyzes teaching videos and provides immediate feedback.

The Challenge

The ability to observe and reflect on classroom practices is a key element of high-quality teacher education. Analyzing video-based lessons helps connect conceptual knowledge with authentic practice.

However, systematically evaluating written reflections is highly resource-intensive and prevents timely feedback. This is where INFER comes in.

Our Questions

The project pursues three main objectives:

  1. Automated Assessment
    Develop and validate AI models to evaluate the quality of students’ video interpretations along the dimensions of describing, explaining, and predicting.
  2. Adaptive Feedback
    Implement a web-based interface that delivers individualized, real-time feedback and fosters students’ digital self-determination.
  3. Scaling and Transfer
    Integrate INFER into teacher education at multiple universities and prepare for broader use by both pre-service and in-service teachers.

Intervention Design

The project comprises three central phases:

  1. Phase 1 (AI Development & Validation):
    Training and validating ML models on student and teacher reflections of both staged and authentic classroom videos, comparing automated classifications with expert ratings.
  2. Phase 2 (Experimental Testing):
    Embedding INFER into a classroom management seminar at the University of Tübingen. Conducting a randomized controlled trial (RCT) to compare outcomes with traditional video-based seminars.
  3. Phase 3 (Scaling & Implementation):
    Transferring INFER to five German-speaking universities (Tübingen, LMU/TUM, Kiel, Jena, Basel). Providing manuals, training materials, and open-access resources.

Timeline

2021-2025
Preliminary work, e.g., data collection
2025-2026
System development and validation
2026
Experimental testing (pilot study)
2027
Public beta version
2027-2028
Scaling & Implementation

Project participants

The project is being conducted at the Hector Research Institute of Education Sciences and Psychology at the University of Tübingen.

Project management (Germany): 

National partners:

International partners:

  • Dr. Ha Nguyen, School of Education, University of North Carolina at Chapel Hill

The development of the INFER app is based on seed funding from the Universities of Tübingen and UNC-Chapel Hill (2025–2026).

In addition, national funding applications (e.g., DFG, Stiftung Innovation in der Hochschullehre) are planned to conduct basic research and secure long-term funding for scaling and implementation.